Talk:Negative feedback/Archive 1

This is an archive of past discussions about Negative feedback. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page.

Archive 1

Archive 2

Archive 3

Archive 4

The link

The link:

Donella Meadows' twelve leverage points to intervene in a system

has been moved here, this isn't an esoteric topic, it refers to the engineering term.. Perhaps there ought to be a Esoteric Negative Feedbak page??? Dominick 14:23, 13 Apr 2004 (UTC)

eBay in See Also??

Is this a typo or does eBay have something to do with negative feedback...? Rohitbd 09:28, 23 September 2005 (UTC)

Er, deleted text?

Does anyone care to explain why a bunch of the text I added about negative feedback in gaming was deleted, the remainder was commented out, and my reference was removed? I'm not making this stuff up, the book specifically used the term "negative feedback" and the examples seem to make it clear that this kind of cybernetic system responds in a way that tends towards equilibrium. Was it removed because the article is focusing on a narrower meaning of the term, or what? I'm just wondering. Deco 22:38, 30 November 2005 (UTC)

I would say the meaning of the term, as this is primarily a mathematical term. You need a loop, and then to show the response is subtracted from the input, and I think that man in the loop systems don't really count except in a narrow set of cases. I commented it out to see if I can come up with better text and examples. From where I sit, many examples of negative feedback are really examples of negative outcomes. When you put the man in the loop, they can choose to make this positive or negative feedback. If you can somehow show me how a loop is closed thats fine. Thanks for asking here, lets work on the article. Dominick ^(TALK) 01:05, 1 December 2005 (UTC)

Well, it is true that I was using it in a less rigorous/precise sense. Maybe it would help to have a more general informal article about positive and negative feedback loops as encountered in daily life and also more precise technical articles for the formal terms used in theory. I don't know. Deco 01:55, 10 December 2005 (UTC)

You encounter negative feedback loops every day in your life. Every second, in fact every keypress on your computer. All use amplifiers stabilized by negative feedback. There are few positive feedback loops. Generally if you can build a unstable amplifier, that is made to give outstanding gain, you can stablize it with a negative feedback loop. One application of positive feedback is an oscillator. Generally positive feedback amplifier techniques are only used where it is not possible to make a more stable circuit, like at extreme frequency. Dominick ^(TALK) 21:40, 10 December 2005 (UTC)

This information needs to be more specific. What about negative and positive feedback loops in environmental science? -Matt

Cleanup Tag

Can the original poster of the cleanup tag please give a cogent comment on what additional changes they want within 7 days, or I shall presume all is now well and remove the tag. Bob aka Linuxlad 14:03, 25 January 2006 (UTC)

NFB Section

Do we relly need the blow by blow account in developing the final equation?--Light current 17:37, 30 June 2006 (UTC)

Simplified it a bit--Light current 17:48, 30 June 2006 (UTC)

V' doesn't reproduce well on my Browser (Linux Firefox IIRC) - compared with V it looks very similar. Bob aka Linuxlad 20:28, 30 June 2006 (UTC)

Very narrow

Negative feedback happens in economics (scarcity causes prices to rise which chokes off demand) and in living systems. If it wasn't abundant in nature the world would fall apart. Yet this article confines itself to amplifiers and capacitors and suchlike. It misses the opportunity to explore a topic central to life itself. Or is that done in another article? David Colver 16:21, 30 August 2006 (UTC)

feedback depends on what your talking about

In Ecology, Feedback, is when you have a state variable that is being affected by another variable that infact is the state variable it self, you get feed back, now negative feedback for instance would be competition which is a negative-negative interaction ( The species hurts its self more than it hurts its competator), and positive feedback would be mutualism a positive-positive interaction, ( species are mutualists for simply their own benefit, it just so happens that the service they are doing for themselves helps the other species, for instance bees and flowers," the bee benefits because it gets nectar, the flower benefits because it gets pollinated"). This is feedback only in terms of Ecology. These statements are fact based on Dr. Mike Rosensweig Phd at the University of Arizona, and Dr. Paveo-Zuckerman whom both work in the college of science in the department of ecology and evolutionary biology. Dr Rosensweigs research is known world wide as one of the key players in ecology, he is the student of the late Dr. McArther, he has worked with Gause', Kitchner, Jarred Diomond, and Huffacker. —The preceding unsigned comment was added by 69.252.128.148 (talk) 08:00, 15 December 2006 (UTC).

Fever

"I know right now you are thinking, "CAN I HAVE NON-CONFUSING FEEDBACK?!?!?" well if so, then think of it this way. If your body has a fever, your body will then try to make you have a normal temerature, so it fights it off."

This is not very encyclopedic. It uses the first person, contains a spelling mistake, and is completely inaccurate. Fever is not a disease that the body's thermostat attempts to 'fight off'. It is a change in the control temperature which makes the internal environment less hospitable to pathogens. Someone with a fever has a perfectly working thermostat that is 'intentionally' making them hotter. As the article on fever points out, it is completely distinct from hyperthermia, in which the thermostat mechanism fails or cannot cope.

I recommend immediate deletion.

--83.217.142.213 13:25, 29 March 2007 (UTC)

Negative feedback loops and feedback inhibition in biology

There seems to be little information on negative feedback loops and feedback inhibition in nature, e.g. in enzyme biochemistry. Negative feedback loop redirects here and feedback inhibition redirects to feedback. Neither article provide sufficient information for a topic that is covered in high school and intro bio classes. - tameeria 21:13, 5 May 2007 (UTC)

Amplifier feedback isn't that simple

Most amplifiers aren't unilateral, so the simple feedback equation from control theory doesn't really apply. I think there should be an article dedicated just to negative feedback amplifiers covering the various methods of analysis. Roger 23:27, 11 May 2007 (UTC)

Really? Are you really claiming that most amplifiers have significant reverse gain - significant in the sense that the simple feedback equation doesn't apply? If you had said 'most RF amplifiers' aren't unilateral, I might not be so surprised. Please provide justification for this assertion, Roger. Do you have some citations to support this? Alfred Centauri 04:01, 13 May 2007 (UTC)

No, I mean most aren't perfectly unilateral, and yes, even less so at RF. Feedback circuits are certainly not unilateral (e.g. a resistive divider). Another problem is loading from the feedback circuit on the amplifier which makes it difficult to partition an amplifier according the ideal feedback model. Of course all of this is not normally significant in opamp circuits, but nevertheless it is in general. If you have "Analysis and Design of Analog Integrated Circuits, 4th Edition" by Gray and Meyer, check out page 499. He mentions this and gives a more complete block diagram for feedback amplifiers (its also here: http://ardem.com/images/NatBlockDiag+Adj.jpg). Roger 05:07, 13 May 2007 (UTC)

I have the 3rd edition of this book so please refer me to a chapter heading. At any rate, while I agree with you that the simple feedback equation is an approximation except in the limit of an ideal amplifier, is it reasonable to say that it doesn't apply? After all, aren't most equations in EE approximations? Alfred Centauri 13:14, 13 May 2007 (UTC)

Look in the chapter on feedback, right before the discussion of the return ratio method (if thats in your edition). If you have access to IEEE, find the follow paper: "Striving for Small-Signal Stability" by Tian, Visvanathan, Hantgan and Kundert. They say "It should be noted that the ideal single-loop feedback network shown in Fig. 1 is not an adequate representation of a practical feedback network. In practice, the active path may not be strictly unilateral; the feedback path is usually bilateral, and the input and output coupling networks are often complicated." Their Fig. 1 is the classical feedback block diagram. Anyway, I just thought feedback amplifiers deserved a more thorough coverage. I'll work on an article when I get a chance. Roger 00:20, 14 May 2007 (UTC)

Regarding "stability"

Hi Alfred,

You recently reverted my change regarding the phrase "improves gain stability" as an advantage of negative feedback, as I misinterpreted what was meant by "gain stability". Consequently, I think this could be clarified in the article. This article already talks about stability in the BIBO sense, and all mentions of stability in the Operational amplifier article also refer to the BIBO sense.

Given that there is currently no article on "gain stability", perhaps we could clarify what is meant by this phrase; most importantly, that it's not the same meaning of "stability" that was used a few paragraphs previously? Oli Filth 00:36, 14 May 2007 (UTC)

Good point. I think the word 'stability' shouldn't be used here for the reasons you mentioned. Maybe something like "reduces gain variations arising from component tolerances, changes in temperature and aging". I dunno, that sounds too long. See what you can come up with. Alfred Centauri 01:25, 14 May 2007 (UTC)

Tags?

I have just moved the tag saying that this artical is unreferenced to the top. I thaught that it would be easier for cleaners to see.60.241.0.81 (talk) 20:24, 27 November 2008 (UTC)

Hormones

Negative feedback can refer to any mechanism for regulating hormones. Examples given in the article such as baroeflex give absolutely no citation, and they are scientific out there about sleep can affect biological rhythm cycles. These neurobiology are still in research and they should no reason, why people can't reference it, if the evidence it solid. --75.154.186.241 (talk) 22:37, 11 April 2009 (UTC)

I see no reason to include a link to a page about a pseudoscientific concept that makes no mention of the subject. Whilst Negative feedback might be an appropriate "See also" for the Biorhythm article, that doesn't imply that Biorhythm is an appropriate "See also" for the Negative feedback article. Oli Filth^{(talk|contribs)} 22:40, 11 April 2009 (UTC)

Negative feedback is merely a mechanism of about one of relevant topic to equilibrium, biorhythms....etc and hence they are technical. The topics merely give guidance on the topic, it is neither farfetched nor not plausible. Just because, you don't believe in protoscience and psuedoscience doesn't mean they shouldn't be include. Wikipedia is not about your personal viewpoint.

The see also merely serve as a navigation. Relating similar topic is not off-topic, they only better serve a purpose likewise to systematic and informatics and melt down the barrier of conditioning. If you are bias and don't believe in conceptual science than you know beforehand that scientific reasoning did derived from philosophical reasoning. For those information, if interested look into it yourself. --75.154.186.241 (talk) 22:58, 11 April 2009 (UTC)

No connection to this article is apparent at the Biorhythm article. As far as I can tell, the human who is supposed be affect by rhythmic influences does not then turn around and damp the influences themselves. There's no negative feedback. That link is so-o-o out of here. Binksternet (talk) 23:01, 11 April 2009 (UTC)

Changed to Biofeedback, for relevancy. --75.154.186.241 (talk) 23:34, 11 April 2009 (UTC)

Biology

This article should refer to the term in biology primarily and have pages such as Negative feedback (mathematics) as sort of subsiduary pages as this is where the primary use of the term comes from.What do i want to... (talk) 10:42, 2 December 2009 (UTC)What do i want to...

I can attest to the fact that the term "negative feedback" is quite widely used in electrical engineering and has been for quite a long time. See, for example, Harold Stephen Black. Care to provide a reference for your claim that the primary use of this term is in biology? Alfred Centauri (talk) 03:40, 4 December 2009 (UTC)

Article opening needs to be simplified

"Negative feedback (also known as degenerative feedback)[citation needed] occurs when information about a gap between the actual value and a reference value of a system parameter is used to reduce the gap.[1] If a system has overall a high degree of negative feedback, then the system will tend to be stable."

Now I'm no idiot but it seems to me that this intro is a little too complicated. For someone who has heard the term "negative feedback" or "negative feedback loop" and wants to find out what it means, this paragraph starts things off badly. Is there any way to simplify the summary of what a negative feedback loop is? — Preceding unsigned comment added by 213.115.10.98 (talk) 09:12, 4 October 2012 (UTC)

I agree. I had a good look at the article and have gone ahead and boldly rewritten the WP:LEDE. I hope you think it is an improvement, but like everything else on Wikipedia, if you can think of an improvement, please either suggest it here, or make a further change yourself. First, I tried to use very general language - not to assume from the start that we were talking about an engineering example of -ve FB. Second, I tried to use everyday language and avoid the well known technical terms - setpoint (SP), error (E), process value (PV) and manipulated variable (MV) - even though they stand behind what I wrote. Third, I tried to give a human sense of what -ve FB 'feels like' using descriptive words like stable, accurate and responsive. Lastly I tried briefly to summarise the main sections of the article, although they themselves are so brief that this amounted to little more than naming them in practice. --Nigelj (talk) 17:57, 4 October 2012 (UTC)

IMO the lede now more wordy, less focused, and uncited. Rewriting it so completely (and quickly) after a single suggestion seems a bit of a knee-jerk response, isn't it? On what basis did the previous poster think it was "too complicated"? Trevithj (talk) 01:15, 5 October 2012 (UTC)

I completely agree that the intro was too complicated and I am glad an editor was bold enough to rewrite it. IMO it is much better in this way. Now, if it needs some more improvement, please edit! Lova Falk talk 10:29, 6 October 2012 (UTC)

Please clarify. Why is a five-sentence uncited paragraph less complicated (and "much" better) than a two-sentence cited paragraph? Trevithj (talk) 07:00, 7 October 2012 (UTC)

Because the originial first sentence was incomprehensible to most people who don't already know what negative feedback is. The five sentences now are easier to understand. But I agree with you that there is a need for references. Lova Falk talk 07:11, 7 October 2012 (UTC)

Fair enough. I have reinserted the original citation. This involved rewording the 3rd sentence to match the source - but I tried to reword it as little as possible. I agree that original first sentence seemed cluttered and ugly. Trevithj (talk) 07:54, 7 October 2012 (UTC)

Great, thank you! Lova Falk talk 08:41, 7 October 2012 (UTC)

Better naming conventions for Negative and Positive Feedback

Every single time I try to explain Positive and Negative feedback to any layperson, they immediately begin by thinking "positive = good" and "negative = bad".

I don't have any sources to cite this...except that every time I hear and see feedback explained...writers always need to clarify that "positive" does not mean "good".

So this is what I propose:

We try to help shape the discussion on the Positive Feedback and Negative Feedback pages by providing alternative naming conventions.

For example, we can write: Positive Feedback (or amplifying feedback) is.... Negative Feedback (or stabilizing feedback) is....

This immediately helps a lay-person understand that positive is NOT good, nor negative BAD.

Unfortunately, I don't have any citations for this, since I don't think anyone cares to do a rigorous study on the amount of time wasted explaining that positive does not mean good, nor does negative mean bad.

And, I know some people feel like the terms "amplifying" and "stablizing" focus more on electronic feedback, but I personally selected these terms in describing economic feedback loops.

I honestly believe that these types of changes are both valuable to the public, and good.

Thoughts? EzekielChang (talk) 20:18, 31 December 2012 (UTC)

Hi EzekielChang. I share your frustration with this (see Feedback#Types_of_feedback for my attempt to clarify the mess). Yes, the terms "positive" and "negitive" are ambiguous - negative can mean decreasing, undesirable or detrimental, which are exclusive definitions. The difficulty is that in some contexts, saying "negative feedback is bad" is accepted usage. (I personally hate that, but my opinion doesn't count in a public document.) And "stabilizing" may be confusing, given that negatvie feedback can result in oscillation (eg Hunting oscillation) which some disciplines would consider unstable.

I second Nigelj's comments on the Positive Feedback talk page. Nothing erodes the enthusiasm for a subject like an edit war (voice of experience)! It was obvious your edits were in good faith, but this subject is not as straight-forward as it looks. Perhaps Arkalgud Ramaprasad's paper On The Definition of Feedback would be of interest.

But thanks for raising the point - an important one that needs to be continually looked at IMO. Best wishes: Trevithj (talk) 04:40, 1 January 2013 (UTC)

The clarifying terms would be OK if sourced; whether they're appropriate in the lead would be the next question. If this terminology is really an issue, there will be sources that say so; find them and then we can work out how to use them. Maybe this book and this one? But don't give them undue weight, as this is not a common usage, I think. Dicklyon (talk) 06:45, 1 January 2013 (UTC)

I've heard and used the terms regenerative (positive) and degenerative (negative) feedback. For example: Applications of Regenerative Feedback in Integrated Circuits and An Analysis of the Effect on Degenerative Feedback on a Two-stage Transistor Amplifier Alfred Centauri (talk) 14:07, 3 January 2013 (UTC)

Hi Alfred Centauri. This is in the context of electronics? That's the trouble - many different contexts have their own prefered synonyms. I've heard and used the terms reinforcing (positive) and balancing (negative) feedback in the context of Systems dynamics, for example. Are these alternative terms sufficiently generic to use in the lead? If in doubt, don't (IMO). Trevithj (talk) 08:07, 4 January 2013 (UTC)

@Trevithj, while the links I give are indeed in the EE context, it isn't a stretch to consider the terms "regenerative" (reinforcing) and "degenerative" (opposing) to be appropriate in a wider context. Alfred Centauri (talk) 02:19, 17 January 2013 (UTC)

"Reinforcing/opposing", maybe. "Regenerative/degenerative", not so sure. "Regenerate" and "reinforce" are different ideas. Someone familiar with the subject may consider these terms similar, but that isn't who the article is necessarily aimed at. Trevithj (talk) 08:43, 19 January 2013 (UTC)

I think one of the powerful things about the concepts of positive and negative feedback is that they explain and unify lots of different things, in widely differing fields, that separate individuals have struggled with trying to name, to grasp, and to understand. Suddenly they were given common conceptual diagram, terminology and even mathematics so that economists, climatologists, sociologists and others could suddenly learn from the experience of electronic and control system engineers, and vice versa. If at the same time we go searching out books that either date from a time before this cross-fertilisation, or those written, edited and published by people who didn't know there had been any progress, then of course we will find economists talking about amplification, new-age gurus talking about regeneration, people seeing negative feedback and talking about damping - none of whom have any idea either that they're borrowing jargon that does not apply, or missing a chance to use precise language that would describe their thought perfectly. I think that what I'm saying here probably comes under History of Science, and I wouldn't know where to look to find refs for it. --Nigelj (talk) 10:00, 19 January 2013 (UTC)

Very good point. Regards refs for History of Science, there are a few under Feedback#Types_of_feedback, especially Bennett (1979) and Mindell (2002). Trevithj (talk) 00:46, 23 January 2013 (UTC)

Proposed History section

Hi. In the past, there has been some confusion arising from conflicting definitions of terms in different disciplines relating to this topic. I propose introducing a History section near/in place of the Overview section, to give a more impartial context to the different treatments of the subject. This would mirror the approach in the main feedback page. ~~Please see my draft section here: Draft HISTORY section~~

This was partly inspired by a recent re-working of the Reinforcement page that helped clear up some ambiguity related to how "negative reinforcement" is used in psychology. Your ~~feedback~~ critiques would be greatly appreciated. Trevithj (talk) 06:29, 16 June 2013 (UTC)

Hi Trevithj! Great job, please add it to the article. One minor remark though, direct quotes are usually not done in italices but with {{Quote}}. With friendly regards! Lova Falk talk 07:19, 26 June 2013 (UTC)

Hi Lova Falk, thanks for the pointer - I have modified the quotes as suggested. If no-one else has any suggestions after a week or so, I will publish. Many thanks! Trevithj (talk) 09:47, 28 June 2013 (UTC)

History section added. There are possible duplicate uses of cited references, but will leave that for others/another day. Trevithj (talk) 04:33, 6 July 2013 (UTC)

Centrifugual vs. centripetal

The statement is "As their speed increases they swing up and outwards due to their ... acceleration.". Since centripetal forces act "towards the centre", and outward motion is away from the centre, "..." must be "centrifugal". Please note that I said as much in my first edit summary, "inward" is equivalent to "towards the centre". Regards Paradoctor (talk) 12:21, 14 December 2013 (UTC)

The problem is that there is no such thing as "centrifugal acceleration". In your first edit summary you said, "centripetal forces act inward, not outward", but the sentence is about acceleration, not forces. I can only repeat that there is no such thing as "centrifugal acceleration". Many say there is no such thing as a centrifugal force either: to make something rotate around a circle, you make it continually accelerate towards the centre. F=ma, so you need a continual force pulling toward the centre - the centripetal force, causing the centripetal acceleration. The sentence probably needs a complete rewrite, preferably based on a WP:RS reliable source, but it's definitely worse now than it was before, IMHO. --Nigelj (talk) 17:52, 14 December 2013 (UTC)

The term "centrifugal acceleration" is certainly in use, as a quick look at Google Scholar should convince you. Paradoctor (talk) 03:03, 15 December 2013 (UTC)

I think I provided a good fix, with link to centrifugal force (rotating reference frame). The upward and outward motion of the balls is a motion in the rotating reference frame, so is most easily understood and described in terms of the fictitious force in that frame, which is centrifugal. The fact that in an inertial frame the balls are undergoing a centripetal acceleration doesn't explain squat. Many good sources explain the ball governor with centrifugal force. When sources invoke centripetal, they're talking about the opposing force provided by a spring or gravity, which moves the balls back the other way (and provides the force to keep an equilibrium at whatever level). Like here. Dicklyon (talk) 23:19, 14 December 2013 (UTC)

Looks fine to me. If this still causes debate, one might consider removing the explanation, and just describe how the balls move. Paradoctor (talk) 03:03, 15 December 2013 (UTC)

Diagram and discussion are not compatible

The left-hand diagram that appears in this article is for a negative feedback amplifier where feedback is used to control the amplifier gain. The closed-loop gain is determined by the feedback loop that is readily controlled as 1/B, instead of by the open-loop amplifier gain A, which (so long as it is large) can have any value without changing the closed-loop amplifier gain.

However, the description in the accompanying text refers to comparison with a set-point value and adjustment of any error toward zero. That system is better described by the right-hand figure showing how car speed is maintained at the set-point speed limit despite road conditions. Brews ohare (talk) 16:52, 6 June 2014 (UTC)

When I studied feedback and control at university, the diagram we used looked very similar to the one on the left, but the labelling was different. Instead of 'input' it said 'setpoint', the circle with the + in it had an x, the signal between the + and A can be labelled 'error', on the right it would say something like 'controlled variable' not 'output'. I think a relabelled version of what we have would be better as the alternative you propose is an example, not a general case, and it is rather muddled: What is 'road' doing there? That - at most - is a source of perturbations. The setpoint is 60, the speedometer is a sensor that measures the controlled variable. They have the driver doing the differencing nothing much doing the control algorithm, unless it's his boot. No, that's a mess, but the other is the right shape, just needs appropriate labelling. Maybe there is a modern source somewhere that has the current best-practice labelling we can copy? --Nigelj (talk) 18:34, 6 June 2014 (UTC)

http://www.controlguru.com/wp/p62.html This is pretty good. He has split 'A' into three boxes as he wants to define CO and MV as well. We're not teaching a course here, so they may not be needed. I can do an SVG in the end, once we decide what we want in it. --Nigelj (talk) 18:40, 6 June 2014 (UTC)
http://eleceng.dit.ie/gavin/DT009_3/Revision%20worksheet%20answers.htm Here's another. Three blocks for A again. --Nigelj (talk) 18:43, 6 June 2014 (UTC)
http://www.electronics-tutorials.ws/systems/closed-loop-system.html There's several on this page. The first one with G and H in is the simplest, and it's called 'Typical Closed-loop System Representation', which is good. --Nigelj (talk) 18:48, 6 June 2014 (UTC)

The current image is fine for this article. Things like the proposed image are not as clear to general readers as those who understand feedback might expect. Johnuniq (talk) 00:13, 7 June 2014 (UTC)

The current figure is not 'fine'. As pointed out, the negative feedback amplifier is different in concept entirely from the idea behind maintaining a system near its set point, for example, using a thermostat to keep a house at temperature. Brews ohare (talk) 03:32, 7 June 2014 (UTC)

A defect of the diagrams linked by Nigelj that I looked at is that they do not include a disturbance from outside the system that is responsible for causing a system reaction. That omission means that these diagrams don't apply to thermostatic control of house temperature, because there is no outside disturbance like a change in weather that would cause a temperature shift and force the thermostat to turn the furnace on. Brews ohare (talk) 03:44, 7 June 2014 (UTC)

In the right-hand figure, a change in grade of the road is an outside disturbance that affects the car speed, introducing a deviation from the speed limit that causes the driver to adjust the throttle to maintain the car at the speed limit. Brews ohare (talk) 04:03, 7 June 2014 (UTC)

While the paragraph with "systems controlled by a negative feedback loop" is not referring to the figure, the figure itself is fine. If someone wants to clarify the lead text, that would be OK, too. Dicklyon (talk) 06:36, 7 June 2014 (UTC)

Any 'clarification' should add the example of the negative feedback amplifier, and avoid dropping the more common application of feedback to maintenance of a set-point, as found not only in control theory but in biology where it is called homeostasis. Brews ohare (talk) 13:27, 7 June 2014 (UTC)

Here is a block diagram version of the set-point control using negative feedback. Brews ohare (talk) 15:01, 7 June 2014 (UTC)

I understand the points made here, but I maintain that this is the best format for a general diagram. There is no doubt that the existing diagram shows negative feedback around an electronic amplifier, which is an example if negative feedback, as is the road example, and the generalised version of it above from Brews. The problem is that none of these actually represent negative feedback, in the sense that we could say, 'if the situation is not a version of this, then it is not a situation in which there is negative feedback'. Such a diagram is possible, as linked in this comment. Perturbations and disturbances are important in controlled systems, but they are not required for negative feedback to exist; they are not fundamental, like a setpoint, a measured error, and a feedback path are. Effectors, and measuring elements, are also common in engineered feedback systems, but there will be, for example biological homoeostatic systems, where they cannot be found or distinguished. There is no need for there to be a general or a canonical diagram in the article, but we should admit that the op-amp diagram is not it, and neither would these be. The one I link in this comment is, and it can easily be redrawn with a suitable licence. --Nigelj (talk) 15:55, 7 June 2014 (UTC)

Hi Nigelj: We can discuss diagrams, of course. Let's look at your linked diagram here. Perhaps you could outline just what is going on here? In particular, what corresponds to the disturbance upsetting the equilibrium of the system?. It looks to me like there is none. The "input" in this diagram is the set-point, and the "error' is the difference between the "input" and some feedback-modified version of the system output. Supposing the "error" between set-point and feedback is driven to zero, it does not appear that the system "output' will be at the set-point. Also, there is nothing in this diagram to explain why the system is out of equilibrium in the first place. Maybe you can complete a word description of how this diagram represents the regaining of equilibrium from a disturbance using feedback? Brews ohare (talk) 16:13, 7 June 2014 (UTC)

Your link to control guru is equivalent to the block diagram above. Brews ohare (talk) 21:04, 8 June 2014 (UTC)

In the block diagram I've provided, notice there are two summation points, not just one. One is to determine the departure of the system status from the set-point producing the error signal. That is accomplished by a thermostat, for example. The second summation point combines the feedback with the disturbance to provide the net input to the system. It is this modification of input (heat from the furnace + sun load) that causes a change in system status, bringing the system closer to the set point.

Depending upon interpretation of the feedback elements, the input summation may be taken to be a simple summation or a difference. For example, for the case of a house furnace, the input is the sum of the sun load and the furnace heat, and the negative feedback is decided by the controller that increases or decreases the effector output depending upon the sign of the error. If the temperature is too high the furnace is turned down, if it is too low, the furnace is turned up. Brews ohare (talk) 16:29, 7 June 2014 (UTC)

I've found a number of diagrams with two summation points here that are re-arrangements of the one I have provided using one summation point and one difference point. Brews ohare (talk) 16:40, 7 June 2014 (UTC)

I modified the summation polarities to agree with this source. Brews ohare (talk) 16:48, 7 June 2014 (UTC)

Here is another similar diagram (Figure 3.31 (b)) in which the summation of disturbance and the feedback from the 'effector' ('actuator' here) is implied instead of explicit. Brews ohare (talk) 17:11, 7 June 2014 (UTC)

I've modified the introduction to include both applications. Brews ohare (talk) 15:41, 8 June 2014 (UTC)

'sloppy figure' ?

This edit replaced the first of the two images at right with the second, with the edit summary 'replace sloppy figure'. I don't see how the image is sloppy. In fact it's a pretty faithful vector conversion of the second, so should be used in its place, per the note on the older file's page. As an SVG it scales better (so looks much better as a thumbnail), is more easily adapted/edited and is better suited for print. If there are problems with it (and I can't see any) they should be fixed in the image, easily done as its an SVG. But I don't see any problems and the edit did not identify any.--JohnBlackburne^words_deeds 16:08, 9 June 2014 (UTC)

A matter of aesthetics, of course. I refer to the blurry bold labeling in the svg diagram and use of a hyphen instead of a minus sign for the negative input designation. Brews ohare (talk) 16:24, 9 June 2014 (UTC)

I've tweaked the minus sign so it looks like one instead of a hyphen. I don't see any excessive blurriness, just normal anti-aliasing as the thumbnail is generated from the svg. The font is a heavier weight but that's clearer at a small size. It's also more consistent between the β and other characters. The anti-aliasing is the main problem with the PNG: The different lines have uneven weights in the thumbnail, some of it a side effect of re-rasterising a bitmap graphic, some of it's as the horizontal and vertical lines aren't all horizontal and vertical in the PNG. The PNG also has uneven padding around the edges, with little at the top and none at the bottom, something that was fixed in the SVG.

As I mentioned already SVG is much better for vector diagrams. Where there is a problem with a vector graphic it's better to fix it in the graphic, not replace it with an older PNG, especially when the SVG incorporates other improvements from the original PNG.--JohnBlackburne^words_deeds 17:20, 9 June 2014 (UTC)

I've restored it; the -/− issue is fixed, 'blurry bold' is just normal anti-aliasing, while the PNG has a number of other issues (border padding, inconsistent fonts, lines not being horizontal, re-rasterising artefacts) fixed in the SVG.--JohnBlackburne^words_deeds 18:45, 9 June 2014 (UTC)

The change to a minus sign has improved the SVG file.

The various drawbacks you cite for the PNG diagram don't seem to affect its quality, which still is better than the SVG file, and the 'anti-aliasing' of labels in SVG that you mention still leads to fuzzy labels. The advantages of SVG files listed here appear over-rated to me. The idea that SVG files are easier to change than others is true only sometimes, the smaller storage space doesn't seem like a big concern for WP, the idea of better thumbnail presentation for SVG is simply bogus in the present case. Brews ohare (talk) 13:06, 10 June 2014 (UTC)

Your contention that the PNG is better is based on purely subjective criteria, there is nothing wrong with the SVG now. To my eyes the PNG is worse, uneven line weights being perhaps the worst problem. The lines in the PNG aren't horizontal or aligned where they should be, which generates visible artefacts as they're re-rasterised.

Being able to edit SVG files is perhaps their biggest benefit. Want to change a SVG file? Just download it and open it in an editor, to e.g. move a line, change some text. You can't do this with a PNG as the elements aren't editable, you need to recreate it, which is often much more work. In a diagram with text overlaying graphics you might need to redo the entire graphic to change a single symbol, or all the text when the diagram needs changing. If text needs changing but you can't match the fonts or font settings used in the original file you may need to redo all of it, and any graphics it overlaps.

You should really get up to speed with SVG editing. Most people use Inkscape, a free and easy to use graphics program. If you're able to create graphics in another package you should find Inkscape easy to pick up. Once you're using it you become much more productive as it becomes much easier to adapt and re-use existing images, both your own and other editors'.--JohnBlackburne^words_deeds 13:36, 10 June 2014 (UTC)

I suppose that changing a label in an SVG file is easy, but where is the file for this diagram? I'm not familiar with how to find SVG files. As for the ease of using inkscape, I suppose it can't be any worse than learning a programming language. But is it really easier than using Excel to draw a PNG file? The instructions here make it appear that one is basically doing the same drafting one would do in Excel, but with a more limited set of tools?

Where are these artifacts introduced by misaligned lines in the PNG diagram? I don't see any. You put a lot of weight upon the ability to revise, but very often no revision is necessary, as in this case it seems. Brews ohare (talk) 16:26, 10 June 2014 (UTC)

The file is here: File:Block Diagram for Feedback.svg. If you click on the main image it displays the raw image which in this case is SVG. You can then download it using e.g. your browser's 'save' function. Or right-click on the image, or on one of the datestamps at the bottom (preferably the most recent one as it's current) to download.

I haven't used Excel for a very long time and never used it for drawing, but I suspect it's like the more primitive drawing programs I was using that the same time like MacDraw, ClarisWorks. Inkscape is much more capable so has many more features, maybe not all obvious but given time there's little you can't do. The image at right for example. I imagine it would be hard to reproduce in Excel. See commons:Category:Created with Inkscape for many more.--JohnBlackburne^words_deeds 16:46, 10 June 2014 (UTC)

Lede diagram(s)

The recent edits to the lede diagrams touch on an older problem with the lede: it should give as generic an overview of the subject as possible, but actually begins with a very specific and not very intuitive example - the amplifier. While this is understandable from a historical perspective, it is a misleading example, because it implies:

the measured variable to be regulated is the main output
the mechanism of control is to alter the main input

It isn't hard to come up with examples of control systems that don't fit those assumptions: such as cases where internal states are regulated by changing output (body temperature/sweating, reservoir levels/floodgates). Looking at the alternative diagrams in the Diagram and discussion are not compatible section above, all depict the regulation action being control of a main input, and all but one imply measuring of the process output. They are therefore not generic.

I'm also unconvinced that the second diagram shows a "different use of feedback" as stated. The alternative diagrams all show an explicit reference, or desired setpoint. The amplifier has one too, but it is implied rather than explicit (something like 'stable gain', or 'input:output ratio'). This makes the example more difficult to comprehend.

As a possible alternative, I have mocked up this diagram, based on several diagrams in William Ross Ashby's Introduction to Cybernetics. It has much in common with the second block diagram in the lede, but doesn't make assumptions about what the regulator is actually measuring (any one of the blue arrows will do). Ashby makes the point for instance that some control systems can anticipate a disturbance - such as a human being walking over uneven ground.

Without deprecating the good work put in so far, I propose that the "ideal"^{[citation needed]} negative feedback amplifier diagram be moved to lower in the article, and a more generic diagram take its place. Trevithj (talk) 09:45, 11 June 2014 (UTC)

This article is about feedback, which is "after-the-fact" response, unlike feedforward, which anticipates a need for action. It also is restricted to negative feedback, which has to do with maintenance of system parameters, So the feedforward 'human walking' example may be outside the topic. The set-point maintenance diagram, on the other hand, includes maintenance of reservoir levels and body temperature. The possibility of changes in variable form, for example, from temperature as a status indicator to heat, or shivering or sweating as a system response are included.

I think your valid objection could be that the example diagram shows only one kind of disturbance met by only one kind of feedback. So it could be generalized to allow multiple disturbances and multiple responses. That could be confusing, and perhaps a more complicated example could be handled in the body text.

Your diagram also is confusing as it doesn't really indicate how regulation takes place, and doesn't show the system itself that is to be regulated. Can you provide a source where the diagram appears and is discussed? Brews ohare (talk) 14:18, 11 June 2014 (UTC)

Thank you, I do know what the article is about. The source (as stated above) is W. Ross Ashby, "An introduction to cybernetics", Chapman & Hall, London, 1956. There are three diagrams, pg 222-223. The proposed diagram is a merging of these three, with somewhat clearer labeling of the parts (using Ashby's terminology). The "desired states" box is mentioned in the text, so I added it to be more in line with your block diagram and the concept of a setpoint.Trevithj (talk) 09:35, 12 June 2014 (UTC)

I did look at Ashby but did not find it helpful in following your diagram. Perhaps you could provide a verbal accompaniment for your diagram and explain further how it expands the simple examples of the Intro? Brews ohare (talk) 13:39, 12 June 2014 (UTC)

For example, the first figure on p. 223 of Ashby is the customary feedback system, but omits explicit indication of the calculation of the error signals in block E that are fed to the regulator block R that indicate departure of the required variables E from their desired values. Other than this omission which is made clearer in the set-point diagram of the intro, and the lack of a division of the return block R into an actuator and an effector, this diagram is the same as that in the Intro. I don't think this more macroscopic depiction of the system is clearer to the reader than the provided diagram, as it requires more words to fill in what is going on. Brews ohare (talk) 14:30, 12 June 2014 (UTC)

Ashby's block T is defined via the remark: "The disturbances D threaten to drive E outside the set η [the set of states that must be maintained]. If D acts through some dynamic system (an environment) T, ..." (p. 219); it appears that T is the 'receptor' or 'system' (your 'environment' block) that combines the disturbance with the feedback, and the sensor registering the resulting effect upon the status variables either is implicit in T or implicit in E itself. Brews ohare (talk) 14:44, 12 June 2014 (UTC)

With these remarks I interpret your diagram as follows: the disturbance D acts on the environment T, but here is generalized to also cause variations in the regulator itself. The environment T now not only combines the disturbance and the feedback correction, but acts backward upon the regulator itself. The environment T is monitored via several 'essential variables' whose values are implicitly measured and sent to the regulator, which then implicitly computes the errors using the desired values provided by the bottom block. This error information (not explicitly shown) is used by the regulator to create a feedback response (again, by means not made explicit) and forwarded to the environment T. Now the only changes this diagram makes that extend the simple diagram of the Intro are (i) the allowance for regulator disturbances as well as input disturbances, and (ii) the bilateral coupling instead of unilateral coupling between the regulator and the environment and (iii) the allowance that there may be several status variables involved, not just one. The other differences from the diagram provided amount to lumping various functions together in single blocks instead of making them explicit. We might try to compare the explicit indication of various stages in the process as in the provided diagram with a word explanation of a more abstract and more macroscopic diagram that really offers no conceptual advantages in understanding the process. Brews ohare (talk) 15:25, 12 June 2014 (UTC)

I repeat my original point - a lede should provide a generic overview of a subject. Your points above criticize Ashby's diagrams for not being explicit - but that is the whole point, they are generic. We should be providing an introduction to the subject, not details of specific implementations. At present the lede is almost twice as long as it was a few weeks back, and contains more technical jargon. A new reader could be forgiven for believing that negative feedback only applies to electronics. We have lost sight of the audience here.Trevithj (talk) 19:38, 12 June 2014 (UTC)

You are a bit unfair in charging that the length has increased seemingly for no reason, inasmuch as the length has increased because previously only the negative feedback amplifier was illustrated (with no text) and the text described set-point maintenance (with no figure). The changes introduced text to match the amplifier and a figure to match the set-point discussion.

An additional paragraph was added to explain exactly the point you raise here, namely that it is not electrical signals that are involved throughout. An example is provided (thermostatic control of house temperature) exhibiting a variety of forms, ranging from temperature sensing to electrical impulses to gas flow and to heat production. Brews ohare (talk) 05:34, 13 June 2014 (UTC)

The length has increased for a very clear reason - trying to be overly explicit on specific examples. That's why it needed an extra paragraph to explain that the "ideal" diagram didn't fit many examples. A lede should provide a generic overview of a subject.Trevithj (talk) 07:18, 13 June 2014 (UTC)

You might try drafting a beginning to this article based upon a general approach and see how it looks. If you are a master of exposition, you may succeed in getting the points across clearly and more succinctly both for mindsets preferring proceeding from the general to the specific and for mindsets preferring the reverse, without resorting to simple examples. Personally I think trying to combine the negative feedback amplifier example and the set-point maintenance example coherently using a single diagram like yours won't work. To draw out the differences in how feedback works in these two situations will require more verbiage and probably more diagrams than the present explanation. Most textbooks on feedback use separate discussions, which will make sourcing your discussion points difficult. Brews ohare (talk) 05:34, 13 June 2014 (UTC)

A general approach wouldn't be succinct about the points, would use simple examples, and wouldn't attempt to draw out differences of any sort. A general approach would (by definition) look at similarities. That is what Ashby has done, and I think he is better at exposition than either of us.

To be clear, I haven't re-written the lede based on my own opinions, because I can see that you have worked hard at the detail you have put in. Just because I disagree with its appropriateness and location doesn't mean I disagree with the content or the workmanship. I would like to move it though.Trevithj (talk) 07:18, 13 June 2014 (UTC)

Ideal as idealized

I have provided a source explaining the term 'ideal' in the caption as you requested. Brews ohare (talk) 16:25, 11 June 2014 (UTC)

Thank you. As I recall, Wai-Chen has a diagram labeled "Ideal negative feedback model", but he doesn't explain anywhere in the text or citations why it is "ideal". For all I know, it could be a typo. The reference does justify your addition of the term "unilateral" however.Trevithj (talk) 09:35, 12 June 2014 (UTC)

The quote provided clearly states how real amplifiers commonly differ from the one in the diagram. Brews ohare (talk) 13:39, 12 June 2014 (UTC)

Agreed. Which makes the diagram less than ideal, in that sense.Trevithj (talk) 19:38, 12 June 2014 (UTC)

The word 'ideal' in this connection means that inessential complications have been stripped away (like bilateral networks and feed around) to emphasize the critical features. Of course, complications exist and the whole diagram could be cluttered with if's and but's. However, that would not be helpful to getting the idea of the negative feedback amplifier across. As you know, idealization is meant to clarify by emphasizing the key points, and employing elimination of complications that may occur, but whose introduction impedes the conception. Brews ohare (talk) 00:35, 13 June 2014 (UTC)

I don't know that. I thought that "idealization" involved a standard taken as a model for imitation. In that sense, Wai-Chen may well offer an ideal for emphasizing critical features of a negative feedback amplifier. It doesn't follow that it is an idea for negative feedback in general. I also think that the place for "getting the idea of a negative feedback amplifier across" would be the negative feedback amplifier page, and not the lede of a more generic article.

For what its worth, I rather liked the earlier example of a driver on steep roads. That was way more intuitive than diagrams with β symbols.Trevithj (talk) 07:18, 13 June 2014 (UTC)

the use of 'ideal' as a role model is, of course, one meaning of the term. But in technical discussions that is not the meaning usually used. So, for example, an ideal pendulum ignores friction in its pivot and variation in gravity with the bob's position along its arc and probably also changes in tension and in the length of the string during its motion, and so forth. Brews ohare (talk) 15:38, 13 June 2014 (UTC)

A more generic description

Below is a possible generic introduction the the feedback article:

In his Introduction to cybernetics, Chapter 12: The error-controlled regulator, W. Ross Ashby discusses several general versions of regulation as applied to survival of an organism or satisfactory operation of an industrial plant. He introduces the notation D for disturbances that "threaten to drive E [the essential variables] outside the set η [the set of states that must be maintained]." Without regulation, the situation is summarized as "the diagram of immediate effects":

D → T → E,

where T represents the the dynamic system upset by the disturbances D leading to some particular values for the essential variables E. The problem of regulation is then summarized as:

"Given E, η, T, and D, to form a mechanism R, so that R and T coupled, act to keep E within η." (p. 220)

Of course, there are many ways that R might act upon T to counteract the disturbance D. It could , for example, directly observe D and formulate a response to feed to T. Or, it could monitor T instead and formulate a correction based upon T′s reaction to D. However, in error controlled regulation, the response R is based upon the deviation of the essential variables E from the necessary values η for E, and formulation of a response that reduces this error. A simplified block diagram for error controlled regulation proposed by Ashby is at the right.

Each block in this diagram is complex. For example, as part of either T or E some sensors must be included that detect the values of the variables E assumed to summarize the status of T. Some mechanism to input the set-point values η is needed, and a comparator to compute the errors E–η. And the response requires a controller that analyses the errors and decides on a strategy to deal with them. And finally, we need an effector, a mechanism to implement the controller's strategy through what usually is called 'feedback', and in the case where the feedback reduces the errors, negative feedback. Brews ohare (talk) 18:14, 13 June 2014 (UTC)

Very nice. That diagram is simpler, and seems to capture the main points. We have a component that is being disturbed (T), some variable that reflects the disturbance (E), and a 2nd component (R) that influences the 1st component using a comparison of the variable to a desired value(η). The components can be mechanical, organic, electronic or environmental systems or sub-systems. (Ashby uses the word 'machine' in a very generic way.)

The lede could then include some overviews of later examples, by labeling T → E → R (and assuming D and η are understood). Something like this:

mechanical, Steam engine → Speed → Governor
bio-chemical, Body → Internal temperature → Sweat glands
economic, Supply → Excess demand → Profit
electronics, Amplifier → Output amplitude → Gain control
mixed, Automobile → Speedometer → Driver

The complexity of the blocks/components can be elaborated on outside the lede - some subsections already contain links to this end.Trevithj (talk) 08:59, 14 June 2014 (UTC)

In fact, could we possibly simplify the diagram further by emphasizing the two-component idea? Something like what Ashby describes in pg 53/54, but slightly more enhanced. This would follow on from the feedback page's lede diagram - two components that influence each other (T and R) are shown as blocks. The other aspects (D, E and η) could be shown as labelled arrows, or mentioned only in the text. Something like this. Trevithj (talk) 09:21, 14 June 2014 (UTC)

Your diagram differs from Ashby's in changing the D & E blocks to labels. I suppose that a block is intended to indicate some mechanism, while labels are supposed to indicate information or action. Ashby actually uses only blocks and arrows, where arrows are translated by "is unilaterally coupled to" and blocks indicate system parts. Any thoughts? Brews ohare (talk) 14:31, 14 June 2014 (UTC)

Not sure if the E block counts as a 'part', but you are quite right about the arrows. He goes to some trouble to make that clear, as I recall. Although he defines feedback very basically as two parts unilaterally coupled (not his exact words) which is why I wanted to emphasize T and R.Trevithj (talk) 01:26, 16 June 2014 (UTC)

Consideration of a diagram involves the accompanying text as well: that text has been provided for the Ashby figure, and Ashby can be cited as a source for the text and for the figure. Brews ohare (talk) 16:15, 14 June 2014 (UTC)

That's a very good point. Okay, let the original diagram stand.Trevithj (talk) 01:26, 16 June 2014 (UTC)

It might be noted that Ashby's diagram for error-controlled regulation is not applicable to the negative feedback amplifier, where feedback is not used to control the state of the open-loop amplifier, but to defend the amplification of an input signal from variations in the open-loop gain, which are allowed to vary as they please with no attempt to contain them. The basic reason why the negative feedback amplifier is advantageous is that the gain is not affected by the highly variable open-loop gain but by the feedback network as 1/β and this feedback network is assembled from components that do not jeopardize control of the gain because these simple networks are composed of adequately stable components (for example, simple resistors). Brews ohare (talk) 16:29, 15 June 2014 (UTC)

I'm unclear. How does this not fit into the diagram? It sounds like the open-loop gain is effectively the source of disturbance, and that the actual gain is what is being controlled.Trevithj (talk) 01:26, 16 June 2014 (UTC)

The feedback "sets" the gain at 1/β but that is not a matter of adjusting the input to keep T so its essential variables are at some necessary values. In fact, the open-loop amplifier can do as it pleases, so long as it's gain is large. Brews ohare (talk) 02:31, 16 June 2014 (UTC)

The negative feedback amplifier works - in Harold Stephen Black's own overview - by "feeding the output back to the input in such a way as to throw away the excess gain". So T is the amplifier, E is the attenuated output (same as the β block), and R is the summing component. D is the open-loop gain, and (somewhat unintuitively) η is the input signal. The path R→T indicates the modified input to the amplifier. That works, doesn't it? It is more or less the same diagram as Black's with different labels: plus D is made explicit, and the amplifier's main output isn't relevant in Ashby's form.

We're replacing A→β→Σ with T→E→R. Trevithj (talk) 06:59, 16 June 2014 (UTC)

An interesting comparison, and a bit of a stretch. By identifying the monitored internal variable E as the β-block, one is suggesting somehow that the feedback network is related to what goes on inside the open-loop amplifier, taken to be T. That is not so, so yours is a re- interpretation of the Ashby diagram with different meanings. It might work, but it is not a description of error-controlled regulation. In practice, one designs the feedback network to have the β one wants, and you can use absolutely any old open- loop amplifier you like, and still have a gain of 1/β. There is no concern about monitoring what happens inside T. Other things may limit your choice of open-loop amplifier, such as bandwidth, but these considerations don't seem to play a role in this discussion. Brews ohare (talk) 16:17, 16 June 2014 (UTC)

Can something simpler be presented?

Ashby's bunch of blocks with letters in them is not an effective way to convey the concept. And the way Brews included it in the article with an extra 3 kB of bloat on top of the kilobytes he already added is unjustifiable. Let's see if we can improve the article without bloating it. Dicklyon (talk) 16:04, 17 June 2014 (UTC)

Hi Dick: The use of letters in the diagram is a compromise between a simple figure and the text needed to define the variables. How this is done is simply a matter of taste, and evidently Ashby and I don't share yours. That aside, the use of labels greatly simplifies the discussion in the text of set-point regulation. Using Ashby's notation and diagram also assists those readers who wish to consult Ashby's book for further discussion. Brews ohare (talk) 16:08, 17 June 2014 (UTC)

If you think, Dick, that you have a superior approach to presentation, let's see it. Simple reversion to the less general discussion previously present with a utopian belief that something better must exist is just work-avoidance. The notion that a general discussion of error-controlled regulation is "bloat" is odd, inasmuch as the general discussion is useful and was not previously presented. You also are ignoring the Talk-page discussion above with Trevithj. Brews ohare (talk) 16:25, 17 June 2014 (UTC)

Probably the best thing is return to something like the version that existed before all these changes. It's clear and concise, and an appropriate length for a lead. What's there now is far worse: bloated, badly written with added material that just doesn't belong in the lead.--JohnBlackburne^words_deeds 18:03, 17 June 2014 (UTC)

The version referred to by Blackburne is not only unclear, but incorrect, confusing the negative feedback amplifier with error-controlled regulation. The version insisted upon by Dicklyon without Talk-page support ignores several references, including that of Ashby and relies upon a particular example of set-point regulation without presenting Ashby's broader context. Both Dicklyon and Blackburne have decide to ignore entirely the Talk-page discussion initiated by Trevithj and carried through several threads above. This attempt at imposition of abridged and even incorrect material on unsupportable grounds like "bloat" with no attempt at Talk-page support is lamentable and contrary to WP policies. Brews ohare (talk) 20:29, 17 June 2014 (UTC)

I am not insisting on a version, but I dislike the style of the new less-informative figures and the wordy writing style of Brews (and it's unclear to me why Brews is editing here at all; did his indefinite ban end?). I'd like to hear from Trevithj what he thinks. If there's agreement about what the problem is and what sort of change would help, then I can work on that. I don't think it will require adding kilobytes to the lead, as Brews does. Dicklyon (talk) 00:46, 18 June 2014 (UTC)

There are a couple of problems, somewhat inter-related:

for historical reasons, Harold Stephen Black's diagram often makes its way into the lede. Black is credited with coining the term "negative feedback", so that is understandable.
his diagram is nice and simple, but is actually a deceptively specific example. It takes more explanation than we expect to generalize from negative feedback amplifiers to negative feedback. See above discussion, but in a nutshell: it isn't always the output being measured, nor the input being controlled.
Black's work was also quickly superseded by Harry Nyquist and Hendrik Wade Bode, who used a different and more generic definition of -FB, which is more along the lines that William Ross Ashby uses. Ashby is credited with coining the term "cybernetics", BTW - and his work is very readable - so he is also a credible source.

People are passionate about their points of view (me too) in this subject. See Ashby pg 53-54 for a good overview of that conflict (its an old one). But ridiculing other people's edits is never a useful approach. I agree that much of the present detail does need to be moved out of the lede, but that still presents us with the problem of what to put in its place. The version referred to by Blackburne is concise, but we are back to that diagram of Black's again (or near enough). Which uses letters and symbols too, by the way: I don't quite follow Dicklyon's aversion to Ashby's version. Or was it the admittedly poor description? We were working on that.

As for a generic definition: I submit that something along the lines of "two components that influence each other in such a way that a change in one is opposed by the other" is pretty close. "Components" can be any mechanical, bio-chemical, environmental or electronic system or sub-system you like. If we want to break it down into extra detail, it should be limited to: (1) the component/system being controlled; (2) the regulator component/system doing the controlling; (3) something being measured in the system by the regulator; (4) some reference value being compared to the measure; (5) maybe some source of disturbances - tho' probably okay to imply this.

I thought Ashby's diagram fit the bill. Thoughts? Trevithj (talk) 04:01, 18 June 2014 (UTC)

Ashby's diagram was too abstracted to convey meaning visually. Things like "R is the summing component" reduce a meaningful symbol (an adder) to an unfamiliar abstraction, putting all the burden on the description to make the diagram meaningful. Engineers will understand Black's diagram (is that really where it originates?) easily without description, and others will understand it as well as they are able with just a little description. So I suggest you take a cut at it with that in mind. And it's probably not most effective to start with the most general description; starting with a simple definite case will be much more helpful to the unfamiliar reader. Dicklyon (talk) 04:15, 18 June 2014 (UTC)

Ashby's diagram is too abstracted to convey detailed meaning visually. But that is the whole point of abstraction. Trying to fit a general model into a specific example is the wrong way to go. I'm happy to start with a simple example though. How about a common kitchen appliance - the thermostat-controlled fridge? When the compartment gets too warm, the refrigeration unit turns on. When the compartment gets cool again, the unit switches off. A change in one component is negated by the other.Trevithj (talk) 07:13, 18 June 2014 (UTC)

I edited the lead to clarify that the lead image, named "Ideal_feedback_model", is not intended to represent just an amplifier. That is one application. Dicklyon (talk) 03:20, 18 June 2014 (UTC)

Don't be fooled by the file name - this diagram is less than ideal. If you want to generalize it, remove the "output" arrow, relabel "input" as "reference", and drop any mention of "loop gain". Also, a triangle is often used to depict an amplifier component, so maybe A should be a box or circle, to match the other components.Trevithj (talk) 07:13, 18 June 2014 (UTC)

This image is that of a negative feedback amplifier made up of idealized unilateral blocks, and it does not represent some general negative feedback "system". A simple bit of algebra implied by this diagram relates output to input as O/I = A/(1–BA) ≈ –1/B (BA >> 1). There is no connection here to error-controlled regulation. There is no monitoring of an internal state and adjustment of that state as described very adequately by Ashby. I suggest that Dicklyon become educated about the feedback amplifier and error-controlled regulation. It also would be helpful if he read the exchange in the threads above between Brews_ohare & Trevithj that explain all this very carefully. Brews ohare (talk) 13:15, 18 June 2014 (UTC)

Dicklyon's interpretation of Ashby's block R as "the summing component reducing a meaningful symbol (an adder) to an unfamiliar abstraction" is a complete misinterpretation at variance with the text accompanying this diagram, which contains a paragraph indicating that "each block in Ashby's diagram is a composite" and goes on to point out how sensors, comparators, controllers and effectors are all involved in a more detailed view of what Ashby's blocks contain. Brews ohare (talk) 15:04, 18 June 2014 (UTC)

This version of the introduction has these merits not shared by the present version: (i) it is clear (ii) it is sourced (iii) it is general (iv) it is accurate. It also has figures large enough that one can read the diagrams. The use of Ashby's labels for the parts of the error-controlled regulatory system D, T, E, R in the diagram avoids cramming long labels into the system blocks. It is not an improvement to use long labels and then compress the figure into a small block so the reader has to expand it to read the labels. The labels also serve as a shorthand for the discussion in the text, just as is done by Ashby. Brews ohare (talk) 13:31, 18 June 2014 (UTC)

These two figures can be compared to see the merits of the use of short labels instead of long ones. It might also be noted that the discussion of the diagram on the left provides the context for the special case illustrated on the right. Brews ohare (talk) 13:44, 18 June 2014 (UTC)

I suggest removing the "feedback" label on the arrow in the left diagram. Different disciplines use different terms here - some say E→R is the feedback. Others say the whole E→R→T path is the feedback. Easiest to leave it unlabeled. Trevithj (talk) 23:49, 18 June 2014 (UTC)

It seems likely that Dicklyon did not read this version carefully, but looked at Ashby's figure, jumped to incorrect conclusions about its content, and proceeded to engage in hit-and-run reversion without Talk-page engagement that might have clarified matters for him. This haste can be rectified and maybe something better will be forthcoming? Brews ohare (talk) 15:39, 18 June 2014 (UTC)

Brews, my experience leads me to avoid engaging in talk page discussion with you. I will leave this issue to you guys to work out, and we'll see where it goes. Dicklyon (talk) 04:35, 19 June 2014 (UTC)

Fine; I have restored your version of 05:17, 13 June 2014‎. Brews ohare (talk) 14:52, 19 June 2014 (UTC)

My version? That version has your figures, and 4 KB of your new bloat text, since the last version I touched before then. And it throws out my recent attempt to clarify. But whatever. Dicklyon (talk) 21:53, 19 June 2014 (UTC)

Feedback label in diagram

I suggest removing the "feedback" label on the arrow in the left diagram. Different disciplines use different terms here - some say E→R is the feedback. Others say the whole E→R→T path is the feedback. Easiest to leave it unlabeled. Trevithj (talk) 23:49, 18 June 2014 (UTC)

Trevithj: Can you provide some sources that refer to E→R as the 'feedback'? I think that is a poor usage. Ashby uses feedback as what is passed back to the system T. There is some ambiguity there, although in simple cases like heating a house, the 'feedback' has to be heat inasmuch as the disturbance is heat. However, the matter can be a bit vague. For example, in controlling the speed of a car, one could say the feedback to an increase in speed because of going down a hill is a reduction in fuel input, although that can only have an effect as a decrease in torque on the wheels from the engine counteracts the increase in torque due to gravity. Brews ohare (talk) 14:40, 19 June 2014 (UTC)

Trevithj: I found this figure and this figure. It looks as though the term 'feedback' depends upon the diagram one wants to use to describe the system. If so, perhaps the term feedback in the Ashby diagram is OK, but with a different diagram one would use different wording? Brews ohare (talk) 17:12, 19 June 2014 (UTC)

Thanks for tidying up the discussion. The first (fig 8.2) effectively is the Ashby diagram with different labels. Note that the author calls E "feedback sensor", so that would argue against labeling the R→T arrow. As for the second (fig 1.2), that is Black's diagram, with A split into K + H, and B represented as a plain line.Trevithj (talk) 08:51, 20 June 2014 (UTC)

If the label 'feedback' is removed, these matters can be explained in the text. However, for the cursory reader the label feedback in the diagram gives some idea of what the term means without too much thought. Brews ohare (talk) 14:47, 19 June 2014 (UTC)

Well, in that case I'd suggest applying the label to the entire E→R→T cycle. After all, remove any one of those, and there is no feedback.Trevithj (talk) 08:51, 20 June 2014 (UTC)

As Dicklyon has decided to be only an observer, can we decide whether or with what text Ashby's diagram is to be used in the article? Brews ohare (talk) 15:20, 19 June 2014 (UTC)

Trevithj: I hope not to bore you, but some further analysis of these two diagrams is needed. Both figures can be seen as rotations of the Ashby diagram that put the system input in the middle of the diagram as u(t) in Figure 8.2 and as u in Figure 1.2, referred to as the "input". Both also put the set-point information at the extreme left, labeled r(t) in Figure 8.2, and as "desired output" ȳ in Figure 1.2. As a result, the bottom link becomes the 'feedback', going to the 'feedback sensor' (block E of Ashby) as you noted for Figure 8.2 and going directly to the comparator generating the error signal ȳ–y fed to block K (the effector block) in Figure 1.2, not explicit in Ashby's figure but implicit either in block E or block R. Figure 1.2 does look like 'Black's diagram', assuming that to be the depiction of a negative feedback amplifier. However, it leads to the relation y/ȳ =KH/(1+KH) (Eq. 1.4, p. 3) which lacks the key aspect of Black's amplifier that for large gain K the gain is the reciprocal of the feedback (that is y/ȳ =K/(1+Kβ) → 1/β. So I'd take Figure 1.2 as a mess in which the disturbance D of an error-controlled system is omitted as well as Black's feedback block (or, if you like, a negative feedback amplifier with β≡1 - a buffer stage).

Are we in agreement? Brews ohare (talk) 15:45, 20 June 2014 (UTC)

I've added a modified version of Ashby's diagram that seems to allow both types of negative feedback with appropriate block interpretations. I am uncertain whether this second diagram is really useful - it is simply a parallel with that of Friedland Brews ohare (talk) 17:01, 20 June 2014 (UTC)

Outline

We can put forward a draft, sure. I favour a reverse-historical approach myself. Something like:

A broad overview of feedback as A⇔B^[1]^[2]^[3] but adding that "negative" is where change in one is negated by the other.^[4] Example: A=freezer compartment, B=cooling unit. [compartment warms]→[unit switched on]→[compartment cools]→[unit switched off].
Ashby's 1956 diagram, adding in an essential variable E, and a reference value η. Enhance the above example, where T=freezer compartment, R=cooling unit, E=internal temperature, η=desired temperature. Make the point that some sort of comparison is involved. ^[5]
Maybe* Black's 1934 diagram as an explicit example, where E = output amplitude, R = (β + summation), and η = raw input. The exact mechanism for control (cancellation of some of the input) should not be gone into in detail - it is a complex point to make, and best left to the other article.

*I'd rather not add this, but Black did coin the phrase "negative feedback", so it is historically important. Also widely cited. Reference list is below

Trevithj (talk) 08:56, 20 June 2014 (UTC)

Don't forget Karl Küpfmüller. See this book and this paper. I have copies of some of his books if you need a look. Dicklyon (talk) 05:06, 21 June 2014 (UTC)

It might be noted that this article is not about feedback but about negative feedback. As such, perhaps the attempt to place negative feedback in a wider context should be placed in a subsection "Connection to other forms of feedback"? It could there be differentiated from feedforward control and open-loop control, among other things. With regard to negative feedback, a general statement like those you quote above (for instance, "[negative feedback] tends to oppose what the system is already doing" could be made clearer and used as an introduction. The generality of negative feedback needs to be clarified as not being restricted to error-controlled regulation, treated, for example, in Ashby's Chapter 12: The error-controlled regulator. As one limitation, error-controlled feedback is most often treated from the stance of linear control theory, and nonlinear control is beyond its ken. I am unclear just how this will proceed. Based upon Figure 1.1 and Figure 2.1, error-controlled regulation is nonetheless the basis for much analysis - we just approximate the general case with a linear version and most of the discussion is how to choose the best linearized version. The question remains as to how to include the negative feedback amplifier in the general approach. I've provided a possibility in the figure above. Brews ohare (talk) 13:30, 21 June 2014 (UTC)

Re making the quote clearer - "Negative feedback describes the case where two components influence each other^[1]^[2]^[3] in such a way that change in one is opposed by the other".^[5]^[6]

Re including the negative feedback amplifier in the general approach: not sure why this is a problem. [amplifier output increases]→[input dampening increases]→[amplifier output decreases]→[input dampening decreases]. Perhaps mention that the undampened input signal is used as a reference point for the output^[5], and add a link to the negative feedback amplifier page for the exact details.

Reference list

^ ^a ^b William Ross Ashby, "Introduction to Cybernetics" 1956, pg53. "'feedback' exists between two parts when each affects the other..."
^ ^a ^b K.Johan Astrom and Richard Murray, "Feedback Systems", 2008. Page 1. "The term feedback refers to a situation in which two (or more) dynamical systems are connected together such that each system influences the other..."
^ ^a ^b Peter M. Senge (1990). The Fifth Discipline: The Art and Practice of the Learning Organization. New York: Doubleday. p. 424. ISBN 0-385-260-946. page75 "[Feedback is] any reciprocal flow of influence ... every influence is both cause and effect."
^ Norbert Wiener "Cybernetics, or control and communication in the animal and the machine". Cambridge, Massachusetts: The Technology Press; New York: John Wiley & Sons, Inc., 1948. page97 "Notice that the feedback ... tends to oppose what the system is already doing, and is thus negative."
^ ^a ^b ^c Charles H.Wilts, "Principles of Feedback Control", C.I.T 1960. pg1 "...actual comparison of [the essential variable] quantity with its desired value and utilizing the difference to reduce the error observed"
^ David A. Mindell (2002). "Between human and machine: feedback, control, and computing before cybernetics." Baltimore: Johns Hopkins University Press. "In analogy to James Watt’s flyball governor on a steam engine when the engine speeds up, the spinning balls slow it down, and when the balls spin slower, they speed up the engine. Hence the feedback is negative."

Trevithj (talk) 04:33, 22 June 2014 (UTC)

Trevithj: The negative feedback amplifier doesn't work that way. It is just an amplifier and amplifies the input signal by the gain 1/β If the signal increases by a factor of two (say) so does the output. There is no 'damping' effect. Brews ohare (talk) 05:17, 22 June 2014 (UTC)

The difference from set-point maintenance is that, while set-point regulation aims to counter system changes by counteracting the influence of disturbances, the negative feedback amplifier simply accepts these disturbances and makes them irrelevant altogether. Using the two diagrams above, for error-controlled regulation the disturbance block D would alter the output were it not that the system status is restored to the set-point values using feedback. In contrast, in the negative feedback amplifier, the disturbance D freely alters the system gain, possibly changing it fundamentally, with no attempt to rectify these changes using feedback. Instead the feedback makes the system output the same, making all effects of D irrelevant to amplifier operation, not by reinstating A but independent of what A becomes, so long as it remains large. It might also be noted that in error-controlled regulation the aim is to maintain system status so the output (e.g. cars spit out by the production line, or blood circulation by the heart) that depends upon maintenance of status is held fixed, while in the amplifier the aim is to pass a version of the set-point through to the output, nothing to do with some process activity that requires the system to remain unchanged, but regardless of the system status. Brews ohare (talk) 13:58, 22 June 2014 (UTC)

A fun analogy is the present US House of Representatives, which acts like an amplifier by passing no legislation but continuously dishing out its ideology (its set point), as compared to a working House that actually produces legislation by maintenance of a functional internal status. Brews ohare (talk) 14:16, 22 June 2014 (UTC)

The negative feedback amplifier compared to error-correcting regulation

In the case of a negative feedback amplifier, with resistors setting the voltage gain at 2, if the input is 1 V, then the output is 2 V. Then if the temperature changes and the gain of the op-amp drops, the feedback works to ensure the output stays at 2 V. If the op-amp supply voltage drops, again the feedback works to ensure the output stays at 1 V. This is a negative-feedback control system, just like a temperature-controlled furnace. Sure if the input rises to 1.5 V, then the output rises to 3 V. Exactly the same in a furnace where the temperature is set at 200 °C, the outside temperature changes, and the feedback works to maintain 200 °C. Then the gas pressure to the burners changes, and the feedback opens the gas valve to maintain 200 °C. If someone increases the set point, the temperature rises to 300 °C. They are not different systems. The disturbances come from all over the place, altering every part of the process, nearly always the feedback serves to reduce their effect - sometimes to make it negligible, other times it is still significant. But the disturbances ('perturbations') are not part of the feedback system. Including a disturbance in the diagram and discussion is to help explain a use case of a particular system, not to describe the system itself. --Nigelj (talk) 15:42, 22 June 2014 (UTC)

Nigelj: You are quite right that "the feedback works to ensure the output stays at 2V" in the case of a negative feedback amplifier with a β=1/2. However, it is not because this system works like a thermostatically controlled furnace. Such a temperature-controlled furnace is an example of error-controlled regulation, and maintains temperature by controlling the heat input to a house so temperature stays at the thermostat setting. Although the negative feedback amplifier maintains the gain at 1/β, that gain is not maintained like a thermostat by comparing the actual gain with a desired gain and adjusting the signal accordingly. The gain is maintained because the feedback loop results in a gain of A/(1+βA) ≈ 1/β regardless of A (so long as A is large) and without any cognizance of what the open loop amplifier behavior A happens to be in terms of the effects upon it of temperature variations, age degradation or whatever. The gain is set at 1/β by virtue of the circuit, period, and variations in the system do not play a role. If, of course, the feedback branch setting β were to vary, then the gain would vary too, and the feedback would have no mitigating effect upon the overall gain - it would no longer stay at 1/β. Brews ohare (talk) 16:41, 22 June 2014 (UTC)

To put it a bit differently, in the negative feedback amplifier there is no comparator such as a thermostat that computes the difference between the desired gain (desired temperature) and the actual gain (actual temperature) and informs an 'effector' (furnace) to take appropriate action (add heat) to achieve the desired gain. In fact, there is no change in A that would have any effect upon operation. Brews ohare (talk) 16:49, 22 June 2014 (UTC)

I think that to get away from this confusion you all have to move away from the situation where the forward gain A is so high that it can be ignored, because in that case there's no significant difference between an op-amp and a comparator (just some compensation in the op-amp to make it easier to keep the feedback system stable). The op-amp is comparing its plus and minus inputs (not shown as drawn, but imagine another zero input being compared, such that output divided by A is essentially 0, a "virtual ground"). But in the more general case, such as the vacuum-tube amplifiers that Black was working with, feedback was very useful to make a system's gain stable, and to reduce a system's nonlinearity, even when the forward gain was nothing great like an op-amp. Get away from the gain being set just by beta and look at the general case, otherwise it looks just like a comparator-based loop. Look at Küpfmüller's AGC-loop analysis for an early good example of a use of negative feedback in a loop that's not even linear (not exactly in modern form, of course). There's no comparator and no very high gain there. Even in the control framework, a high-gain comparator is generally a bad idea; using a lower gain, as opposed to a decision, helps to keep the loop acting more linear, which makes it easier to keep it stable. The typical thermostat is an example of a bang–bang controller; the cruise control in your car is an example of a more gentle feedback loop. Dicklyon (talk) 17:10, 22 June 2014 (UTC)

Dick: The thermostat and the cruise control both are used in an error-controlled regulation setting that is not a subject of controversy here. The automatic gain control employs feedback based on the amplitude of the output signal to modify the gain of the open loop amplifier. That is not what happens in the negative feedback amplifier, where no attempt is made to alter the open loop gain at all. So, although your examples are examples of negative feedback, they have no bearing upon the case of the negative feedback amplifier as envisioned by Black. This application is contingent upon the formula for the closed loop gain A(fb)=A/(1+β A) and in modern electronics such as integrated circuits, the applications assume the magnitude of βA >> 1. Brews ohare (talk) 19:36, 22 June 2014 (UTC)

Here are some typical discussions: Rao, Vorperian, Razavi Thompson Brews ohare (talk) 21:17, 22 June 2014 (UTC)

I don't know if you know anything about electronics (and this is not the place to teach you), but if you look at Op-amp#Internal_circuitry_of_741-type_op-amp you will see that the first stage of a typical high-gain op-amp is some variation on what is known as a long-tailed pair which is in very precise terms a voltage comparator. With the input set fixed at 1 V (just like a setpoint) the op-amp voltage amplifier with negative feedback works exactly as a control system: sensing the output voltage it alters the relative conductivity of the two output transistors so that there is better or worse conductance between the output terminal and the two supply rails, always striving (by amplification of the error voltage sensed by the long-tailed pair) to maintain the output voltage at exactly 2 V. It does this to combat of external disturbances, which typically might include supply voltage variation, temperature effects on the gain (as I mentioned above), or the load drawing more or less current through the output. I am worried that with your shaky grasp on the principles and practicalities, the current changes you are making to the article are not overall improvements. You have made quite a few, and I haven't had time to review them all yet. --Nigelj (talk) 22:07, 22 June 2014 (UTC)

You refer to the differential amplifier used in constructing an op amp. It amplifies the difference between two input signals; however, it cannot be mapped into the block diagram of the negative feedback amplifier, which does not do that. Brews ohare (talk) 23:25, 22 June 2014 (UTC)

(ec) Don't worry, Brews does know about electronics, being a professor emeritus of EE and Fellow of the IEEE, as he has pointed out before. The problem is more that he's not good at hearing the perspectives of others, and always insists on doing things his own way. I would disagree with your conclusion that the long-tailed pair is "in very precise terms a voltage comparator"; it is more precisely a transconductance amplifier; yes, both op-amps and comparators use the same input circuit for its high gain, but don't stretch the truth and call it precise. Good luck to you both. Dicklyon (talk) 23:26, 22 June 2014 (UTC)

And while it's true that many sources treat, or even focus on, the high-gain limit, which is an important limiting case, that's certainly not a usual or intuitive place to start. Black didn't start there, but it's a limiting case that his analysis of negative feedback led to. Dicklyon (talk) 23:47, 22 June 2014 (UTC)

Some strange assertions above. The differential amplifier in an op-amp can be mapped precisely to the little circle with a + and - just in front of the triangle representing the gain in our block diagram of the negative feedback amplifier. A transconductance amplifier is one that outputs a current proportional to its input voltage. This could be used to describe many transistor devices, but it does not describe a long-tailed pair if one includes the collector/drain loads, which convert the transistors' collector/drain currents into voltages for output purposes. There are hundreds of textbooks and tutorial websites that describe the way long-tailed pairs (and their derivatives) to a greater or lesser extent succeed in the goal of rejecting common-mode voltages and amplifying differential voltages applied to their input terminals. Here is a quick example from Google where the fact is taken as given in the opening words of a patent. I would rather be involved in less rambling discussions that have more basis in fact and more chance of improving some article text. --Nigelj (talk) 08:37, 23 June 2014 (UTC)

I don't see what's bothering you. Obviously Brews and I both understand about common mode rejection and why long-tailed pairs are used in differential amplifiers, and how load resistors can convert a transconductance stage to a voltage amplifier stage. My concern is that we not treat the operational amplifier or comparator as magic infinite-gain black boxes, but rather treat the finite-gain case, even not necessarily linear; from there, it is easy to see how feedback improves lots of things, as Black showed. Dicklyon (talk) 23:29, 23 June 2014 (UTC)

Nigelj: There is no issue here over the use of the differential amplifier as a stand-alone simple op amp, or as the input stage of a more complex op amp. But it is not an example of a negative feedback amplifier and so has no bearing whatsoever upon the basic issue here, which is simply that the negative feedback amplifier is different from an error-controlled regulator. Brews ohare (talk) 11:19, 23 June 2014 (UTC)

Dick: Whatever the historical facts about Black's initial ideas, the idealized negative feedback amplifier is described by the provided simple diagram of unilateral blocks. You believe that the case where the magnitude of βA is not large is somehow a form of error-controlled regulation. However, the low-gain case has a gain A/(1+βA), and its fundamental design concept still does not involve any attempt to modify the open-loop gain in response to disturbances. Instead, it uses feedback in a circuit trick that reduces the sensitivity of the closed loop gain to variations in A because A/(1+βA) is less sensitive to variations in A than is A itself, becoming completely independent of A altogether when A is large enough. The negative feedback amplifier in principle makes no attempt to modify A by monitoring and controlling its internal "essential" variables, as Ashby describes error-controlled regulation. Therefore, in the low gain case, as with the more typical high-gain case, the negative feedback amplifier is not an example of error-controlled regulation.

Your claim of what I believe is completely wrong. I in fact would tend to disconnect the concept of negative feedback from the concept of error-controlled regulation. And I'm not talking about modifying open-loop gain, in general; it just happened that in Kupfmuller's circuit that's how he was using negative feedback. Dicklyon (talk) 23:16, 23 June 2014 (UTC)

Neither of you all has to accept my arguments here, of course. You both have suggested example circuitry in attempts to contradict this position, but as I have tried to point out to you, these examples are not negative feedback amplifiers, and you do not represent the views of any sources provided, none of which make the claim that the negative feedback amplifier is an example of error-controlled regulation. In fact, the literature on the negative feedback amplifier and that on error-controlled regulation are very nearly disjoint, and the only connection I can find is Friedland who sets up the unity gain buffer where β ≡ 1. I have tried to generalize his diagram as shown here. This diagram makes clear that no monitoring and enforcement of some 'target' internal state of A is involved. Instead the disturbed system A′ is simply accepted and made use of. Brews ohare (talk) 11:19, 23 June 2014 (UTC)

A role for the op amp

Above, Nigelj has introduced the discussion of the op amp pointing out its utility as a comparator. Perhaps a subsection should be introduced to point out the part the op amp can play in some implementations of error-correcting regulation? Brews ohare (talk) 15:38, 23 June 2014 (UTC)

No, I didn't. I introduced the idea that the negative feedback voltage amplifier, which is often built around an op-amp, is an example of a negative feedback control loop, where the input voltage is the setpoint and the output voltage is the controlled variable. --Nigelj (talk) 15:51, 23 June 2014 (UTC)

Perhaps you can clarify this comment by suggesting a diagram (maybe this one, Figure 14.13?) that implements your vision. If your diagram employs an op amp, that op amp amplifies the difference between its two inputs. If that difference is driven toward zero by the feedback involved in the rest of the circuit, then we have an example of error-controlled regulation, which is not what is commonly described as a negative feedback amplifier à la Black, although of course, it may be an amplifier and it may involve negative feedback. Brews ohare (talk) 17:05, 23 June 2014 (UTC)

In Fig. A, a proportion of the output voltage is fed back to the inverting input. While the input voltage is constant, this feedback ensures that the output voltage is regulated - if extra current is drawn from the output the feedback tries to keep the output at the design voltage. This is identical in principle to taking a voltage from a temperature sensor and using it to try to ensure that an oven stays at a desired temperature. The main difference is in the systems' time response to 'fast' changes: in most cases the negative feedback amplifier will respond quickly, within the limitations of stray capacitance, slew rates etc. The oven could take many minutes to settle, due to thermal lag and limited heat energy available compared to the system's heat capacity. For this reason, people have spent a lot of effort speeding up the oven's response (using PID control etc) whereas most negative feedback amplifiers use nothing other than a little high-frequency filtering to avoid instability.

Figure B shows a very similar negative feedback design, except here the input voltage is fixed by a zener diode, and the output current of the op-amp is boosted by an extra discreet transistor. This design is actually called a voltage regulator and the explicit purpose of it is to control the output voltage while other factors (load current, input voltage, temperature effects etc) act as disturbances. There's nothing new or controversial about any of this.

I offer these here simply because I don't want to see negative feedback voltage and current amplifiers treated as a different case to negative feedback control systems, or even as different in principle to homoeostasis systems found in biology, or negative feedback systems inherent in the planet's climate. The most general diagram at the top of this article should encompass all of these in one figure. I can't find what I consider the ideal diagram on commons, but I linked what I consider a very good starting point above: http://www.electronics-tutorials.ws/systems/closed-loop-system.html - the first one where G and H are used to label the blocks on that page, under their heading 'Typical Closed-loop System Representation'. --Nigelj (talk) 21:58, 23 June 2014 (UTC)

But the most general class, and where you'd want to start, I think, does not have a very-high-gain comparator or op-amp as part of it. Dicklyon (talk) 23:12, 23 June 2014 (UTC)

Is it correct to assume you two both agree that the first diagram in the linked article is an example of error-controlled regulation, the same as Ashby and all discussions of homeostasis? Brews ohare (talk) 00:14, 24 June 2014 (UTC)

That diagram is malformed, with "error", "input", and "controller" in mutually strange positions. But yes it's trying to be an error-controlled regulation. I would not in general subscribe to the idea this is a good model for all homeostasis, but yes, it's what a lot of discussions about that say. I agree with our Homeostasis article that says "Homeostasis requires a sensor to detect changes in the condition to be regulated, an effector mechanism that can vary that condition; and a negative feedback connection between the two." But I wouldn't go so far as to say that Homeostasis necessarily has a target input and/or an error signal. Dicklyon (talk) 04:03, 24 June 2014 (UTC)

It might be informative to try to fit the op amp circuit into a block diagram like Ashby's. One difficulty in doing that is that the op amp circuit contains no explicit source of disturbance, so it appears not to be a regulator at all, but just a way to set the output voltage. That omission also is the case with the block diagrams in the linked article, which have no indication of any disturbance that cries out for regulation. Brews ohare (talk) 13:54, 24 June 2014 (UTC)

The omission of disturbance in these circuits is not because it is not there, but because it is subsumed in the op amp design in its noise rejection. In short, the 'ideal' op amp idealizes all disturbances out of the system by assuming infinite gain and perfect noise rejection. Brews ohare (talk) 13:59, 24 June 2014 (UTC)

I think the omission of disturbances from the diagrams is OK because the potential sources of disturbance are, in general, manifold and unquantified. In the case of the op-amp (and there is nothing in the diagram that says that the triangle represents an 'ideal', infinite gain device), disturbances can come in the form of an increase in load current (load not shown), a change in supply voltage (supplies to the op-amp not shown), a change in the open-loop gain due to temperature change, and other factors. In the classic case of a temperature controlled oven or furnace, the fuel supply pressure can fluctuate, someone may open the oven door, a sealed container in the oven may burst releasing water into the chamber, the heat losses around the oven may alter due to a change in ambient temperature; hundreds of things may disturb it. In a biological loop like blood dilution, the subject may drink a glass of water, or alcohol, or eat salt or sugar, or go for a run. Describing the negative feedback loop itself is one thing. Describing how it responds to a particular disturbance is another (a specific use case). P.S. in that linked tutorial, I have always referred to, and recommended, the 'G, H diagram', not the first diagram. (I would leave out the theta's, and change some of the other labelling though - e.g. using setpoint and controlled variable instead of input and output) --Nigelj (talk) 14:31, 24 June 2014 (UTC)

Nigelj: You have identified some of the disturbances that arise as inessential to the concept of regulation, and that is the case for most of them. However, for example, in controlling house temperature (the favorite analogy for discussions of homeostasis), the whole problem to be solved is to keep the house temperature constant while the outside weather disturbs the heat input to the house. In the voltage amplifier circuits as presented, there is no such problem of compensating for unpredictable variations - the issue apparently is just to set the output voltage at a certain proportionality to the input voltage. The regulatory aspect of regulation doesn't come up. Brews ohare (talk) 14:48, 24 June 2014 (UTC)

I identified three factors that can bring up the 'regulatory aspect' of the feedback amplifier, and they do, in practice, every day. I don't understand why this discussion is such hard work. --Nigelj (talk) 15:04, 24 June 2014 (UTC)

The issue might be that the op amp circuitry is only a portion of a regulator, namely the comparator and controller blocks that then command an effector. (See Figure 4.8 here and §2.2.3 Comparator and controller, here). In effect, the op amp is a realization of a thermostat. Brews ohare (talk) 15:06, 24 June 2014 (UTC)

You may notice that in discussions of thermostatic systems and homeostasis in general, the problem at the forefront is regulation, not how the comparator or controller are constructed. Brews ohare (talk) 15:08, 24 June 2014 (UTC)

No. I think the problem here is WP:IDIDNTHEARTHAT, so I'm out of here for now. When you have finished here, I'll review your edits, and consider WP:BRD. --Nigelj (talk) 15:29, 24 June 2014 (UTC)

No, Nigelj, you are simply refusing to consider that a regulator regulates, while a comparator simply compares. Read the linked sources. Brews ohare (talk) 15:32, 24 J une 2014 (UTC)

No, this just demonstrates once again that Brews is not able to collaborate with anyone. Please, Brews, do not insist on doing this article your way if you can't find even a single other editor willing to work with you on it. And why is this not covered by your ban from editing physics-related articles, broadly construed? Dicklyon (talk) 15:38, 24 June 2014 (UTC)

Well, Dick, you can devolve into silly personalities if that is your wont. However, my reading here is that Nigelj has identified that feedback is useful in constructing comparators, and that this feedback application is separate from applications to regulation, although sometimes useful as a component of a regulatory system. The implication is that we now have three different uses for negative feedback: regulation, comparators & controllers, and negative feedback amplifiers. There is no point (beyond your personal desire for entertainment) in making this division of topics into a personality issue. Brews ohare (talk) 15:44, 24 June 2014 (UTC)

I was not commenting on the division into different uses, but rather on the fact that another editor has given up dealing with you due to his frustration at how you don't pay attention to what he is trying to say. It's all too painfully familiar. Dicklyon (talk) 23:52, 24 June 2014 (UTC)

I object to your characterization. I've stuck to sources and invited you to do the same. Brews ohare (talk) 03:36, 25 June 2014 (UTC)

Outline, again

We can put forward a draft, sure. I favour a reverse-historical approach myself. Something like:

A broad overview of feedback as A⇔B^{[R 1]}^{[R 2]}^{[R 3]} but adding that "negative" is where change in one is negated by the other.^{[R 4]} Example: A=freezer compartment, B=cooling unit. [compartment warms]→[unit switched on]→[compartment cools]→[unit switched off].
Ashby's 1956 diagram, adding in an essential variable E, and a reference value η. Enhance the above example, where T=freezer compartment, R=cooling unit, E=internal temperature, η=desired temperature. Make the point that some sort of comparison is involved.^{[R 5]}
Maybe* Black's 1934 diagram as an explicit example, where E = output amplitude, R = (β + summation), and η = raw input. The exact mechanism for control (cancellation of some of the input) should not be gone into in detail - it is a complex point to make, and best left to the other article.

*I'd rather not add this, but Black did coin the phrase "negative feedback", so it is historically important. Also widely cited.

Trevithj (talk) 08:56, 20 June 2014 (UTC)

References

^ W. Ross Ashby (1957). An introduction to cybernetics (PDF). Chapman & Hall Ltd. Internet (1999). p. 53. 'feedback' exists between two parts when each affects the other...
^ K.Johan Astrom, Richard Murray (2010). "§1.1 What is feedback?". Feedback Systems: An introduction for scientists and engineers. Princeton University Press. p. 1. ISBN 9781400828739. The term feedback refers to a situation in which two (or more) dynamical systems are connected together such that each system influences the other and their dynamics are thus strongly coupled.
^ Peter M. Senge (1990). The Fifth Discipline: The Art and Practice of the Learning Organization. Doubleday. pp. 75–76. ISBN 0-385-260-946. [Feedback] means any reciprocal flow of influence. In systems thinking it is an axiom that every influence is both cause and effect.
^ Norbert Wiener (1965). Cybernetics, or control and communication in the animal and the machine. MIT press. p. 97. Notice that the feedback tends to oppose what the system is already doing, and is thus negative.
^ Charles H.Wilts (1960). Principles of Feedback Control. Addison-Wesley Pub. Co. p. 1. In a simple feedback system a specific physical quantity is being controlled, and control is brought about by making an actual comparison of this quantity with its desired value and utilizing the difference to reduce the error observed. Such a system is self-correcting in the sense that any deviations from the desired performance are used to produce corrective action.

Trevithj (talk) 08:56, 20 June 2014 (UTC)

Comments

I'd rather not add this [Black's 1934 diagram], but Black did coin the phrase "negative feedback", so it is historically important. Also widely cited.

Trevithj (talk) 08:56, 20 June 2014 (UTC)

Trevithj: In view of Ashby's work you have pointed out, I think error-correcting regulation should be presented in a more general manner than at present. The definition of error-correcting regulation as one of monitoring the "essential" variables {E} of a process or life-form and attempting to enforce certain target values {η} for them based upon viability should be made very clear. Also what should be made very clear is that the negative feedback amplifier of Black does not follow this model at all, making absolutely no attempt to control the internal {E}′s of an open-loop amplifier. Brews ohare (talk) 16:07, 23 June 2014 (UTC)

Consequently, I am largely in support of your outline. Brews ohare (talk) 16:59, 23 June 2014 (UTC)

Following the discussion with Nigelj above, I am persuaded that the negative feedback article should identify three applications of negative feedback, and not attempt to combine them as one. The first is regulation, as described by Ashby, and error-controlled regulation in particular. This is applicable to homeostasis. The second is its use in control devices like thermostats and controllers & comparators involving op amps. The third is the negative feedback amplifier. Brews ohare (talk) 16:57, 24 June 2014 (UTC)

Well, I have read the above discussion with interest. Like I said, strong feelings around this subject are not uncommon!

Regards your suggestion about three applications. My feeling is that since all cases of negative feedback that I know of involve some sort of comparison, then the main difference between the applications is terminology. Two values are being compared. Too great a difference causes a response that reduces the difference.

I've tried to follow your argument about why this doesn't apply to negative feedback amplifiers. I agree that there is no explicit setpoint, but even Black described the setup as gain control. And what is gain, but the ratio of input to output? So the system is comparing the input to a sample of the output, and responding to the degree of difference. Ashby called that difference "error". Why is a finer distinction useful in the lede? Trevithj (talk) 09:24, 25 June 2014 (UTC)

Trevithj: Thank you for your interest. The words "gain control" are extremely vague, and do not adequately capture the situation. In the case of homeostasis and the standard analogy made with temperature control of a house heating system, the basic operation as described by Ashby is to select certain 'essential' variables (he labels E) as adequately indicative of the status of a system or process, measure those variables, and compare them with desirable values (he labels η). The difference is called an 'error' and the error is then driven toward zero in what Ashby calls 'error-controlled regulation'. So I'll ask you (somewhat rhetorically perhaps) whether you see this description as applicable to the negative feedback amplifier? In answering this question, I'd suggest these difficulties in equating these two applications of feedback. First, there is no attempt to identify 'essential variables' characterizing the state of the open-loop amplifier (the 'temperature' of the 'house'). Second, there is no mechanism provided to tinker with the properties of the open-loop amplifier to bring it into compliance (no analogy to adjusting the 'temperature' of a house using a 'thermostat' and a 'furnace').

The only response to these allegations that I can see is to claim that the overall gain of the closed-loop amplifier is indicative of the 'internal state' of the open-loop amplifier A (despite the fact that the feedback loop determining β is entirely independent of the open-loop amplifier, and not a property of it) and to claim in addition that the 'desired value' of the system gain is 'enforced' (despite the fact that there is no comparison made between actual gain and the desired value 1/β, so 'enforcement' does not consist in changing the open-loop amplifier operation in any way, no modification of its 'temperature').

The gain of the negative feedback amplifier is A/(1+βA) ≃ 1/β and the desired value of the gain is 1/β. Neither of these is a property of the open-loop amplifier; both are determined entirely by the attenuation factor β of the feedback loop by itself without reference to the open-loop gain; and the gain 1/β is not achieved by driving a measured 'error' in the gain toward zero.

Dicklyon has suggested that perhaps all this eliminates the negative feedback amplifier where βA is large as an example of error-controlled regulation, but possibly the low-gain case is not an exception. While I cannot agree that there is any difference in principle between the two cases (the value of A is not monitored or adjusted in either case), I'd argue that regardless of any failure to agree about this matter, if the most commonly used case of large A is an exception, we still have an exception. Brews ohare (talk) 14:26, 25 June 2014 (UTC)

The op amp voltage amplifier brought up by Nigelj is another case. The gain of the amplifier is determined by a resistor ratio that is entirely independent of the op amp properties and is not achieved by monitoring or altering the op amp in any way, nor is any comparison between the desired gain and the actual gain used to drive the difference between these values to zero. However, it is true in this case that the op amp drives the difference between its inputs to zero, forcing the voltage across R₁ to the value V_in. I don't think that the difference between the lower op amp input and V_in can be considered analogous to the E−η of Ashby as neither of these variables has anything to do with the internal state of the op amp. Contrast with Wilts' description Brews ohare (talk) 15:22, 25 June 2014 (UTC)

Well, as Ramaprasad pointed out, "The focus of feedback may be any system parameter: input, process, or output." That means that there are potentially nine "applications", if I understand your use of the term correctly.

E measures input	E measures process	E measures output
(1) R influences input	(2) R influences input	(3) R influences input
(4) R influences process	(5) R influences process	(6) R influences process
(7) R influences output	(8) R influences output	(9) R influences output

The negative feedback amplifier would count as an example of case (3). The process is neither measured nor directly influenced by the regulator. BTW I think that Ashby's T block refers to the entire system, and not specifically to the process - possibly a source of confusion.

We shouldn't attempt to cover all nine cases in the lede. It is unnecessary detail there. What should be covered is the commonality - something is measured, something is changed as a result of the measure. When the change returns the measure to an earlier/desired value, that is negative feedback.

I think the edited lede is much improved in brevity - thank you. However, it is still not generic IMO. In the first sentence: if "result" is the same as "output", then it describes case (6), which is still overly specific. Trevithj (talk) 00:14, 26 June 2014 (UTC)

I made some changes in wording that might be acceptable to you. I cannot access Ramaprasad on line, but he is quoted as saying:

"Feedback is information about the gap between the actual level and the reference level of a system parameter which is used to alter the gap in some way."

which is very close to Wilts':

"In a simple feedback system a specific physical quantity is being controlled, and control is brought about by making an actual comparison of this quantity with its desired value and utilizing the difference to reduce the error observed. Such a system is self-correcting in the sense that any deviations from the desired performance are used to produce corrective action."

These two comments apply to error-correcting regulation. Ashby has a more detailed view; see for example, his discussion of control viz à viz regulation pp. 202 ff and the application of the law of requisite variety pp. 206 ff. I'm inclined to think a general discussion like this belongs in a more general article, and it is sufficient here that the article is not limited to error-controlled regulation, but does look at some other examples. What is your recommendation? Brews ohare (talk) 05:06, 26 June 2014 (UTC)

Hi Brews ohare. My recommendation is as before - treat the differences as a matter of terminology. We should be looking for the similarity between op-amps, negative-feedback amplifiers and error-correcting regulation. Looking at the differences only leads to infinite hair-splitting. Trevithj (talk) 02:28, 15 July 2014 (UTC)

Operational amplifier circuits: In thinking over the various Op amp amplifier circuits, I now doubt that these are feedback circuits at all. First, they do not feedback anything - they set the input independent of any output and are feedforward circuits in this sense. Moreover, they are not designed to minimize 'disturbance' of the system in any way. So they do not satisfy the definition of the intro as an example of the "monitoring of a process used to influence the operation of the process itself, minimizing undesirable variations in its operation".

The subsection on op amp applications should be deleted. Brews ohare (talk) 13:12, 26 June 2014 (UTC)

The only opening for op amp circuits in this article is in the case that the op amp gain is taken as non-ideal, in which case the op amp gain appears in the equations for the circuit operation, and the circuit is analyzed from the same standpoint as the negative feedback amplifier, that is, that the circuit serves to insulate its operation from the variations in op amp gain under the assumption that this gain is very large. Such an error analysis is found here (more can be seen in Amazon's "look-inside" feature). The role of gain is expressed here Brews ohare (talk) 14:00, 26 June 2014 (UTC)

This point of view that the introduction of feedback is analogous to its use in the negative feedback amplifier is contrary to that of this source which claims feedback is present in (for example) the ideal infinite gain noninverting voltage amplifier circuit. I find that to be an odd use of the term 'feedback' as the very operation of the ideal op amp insures that its two inputs are equal regardless of how the circuitry connecting these two inputs is arranged. So the "feedback" is not feedback at all, but an imposition of the equality of two node voltages that is entirely independent of anything going on in the external circuit and independent in particular of its output. The term 'feedback' as used by this source seems only to mean that there is a wire connecting the output to the input, regardless of the fact that this connection carries no information back to the input, but to the contrary, imposes a condition upon the system that is independent of its operation. Brews ohare (talk) 14:07, 26 June 2014 (UTC)

Perhaps an exception is the differential amplifier circuit shown here, because there actually is feedback from the output to the input despite the imposed equality of the two op amp input voltages. Perhaps the accurate stance is that some op amp circuits do involve feedback, even if the op amps are ideal and have infinite gain, but not all of them. There still is the issue as to whether the differential amplifier circuit uses feedback to counteract system disturbances. A possible suppressed disturbance is achieved by common mode rejection in the two inputs. Brews ohare (talk) 14:49, 26 June 2014 (UTC)

I have modified the discussion of the op-amp section to reflect these remarks. Brews ohare (talk) 15:53, 26 June 2014 (UTC)

Unacceptable

This: "Negative feedback occurs when the monitoring of a process is used to oppose undesirable changes..." is not acceptable as an opening definition of negative feedback. To take just one of many examples, how does this relate to the negative feedbacks that operate in the earth's climate system? I don't know what you've been up to here since early June, and these walls of text are too long to read right now, but this is no longer a top-level overview article on the general topic of negative feedback. This will have to be fixed. --Nigelj (talk) 19:41, 13 July 2014 (UTC)

I agree; "undesirable" has no part in the proper definition of negative feedback, and probably we also don't need "monitoring of a process" to define it. Dicklyon (talk) 04:02, 15 July 2014 (UTC)

Thirded. It simply portrays a fundamental lack of understanding of the topic. The rest of the paragraph is little better and the whole lead far worse than e.g. that of a month and a half ago, which actually made sense as a lead section.--JohnBlackburne^words_deeds 04:34, 15 July 2014 (UTC)

Fourthed, I removed it. It's a subset of negative feedback; the definition is over-narrow.GliderMaven (talk) 14:55, 15 July 2014 (UTC)

I've no problem with these remarks. I see no reason, though, for deleting sources and deleting the remarks separating feed forward from feedback, nor for deleting any explanation of how negative feedback is to be separated from feedback in general. If these points seem misplaced here, they should be explained somewhere else. Brews ohare (talk) 22:03, 16 July 2014 (UTC)

I've just reverted an addition with essentially the same problem. Including "The purpose of ..." in the definition makes no sense. Negative feedback occurs in many situations where it has no 'purpose', it just is. Similarly there is nothing necessarily "undesirable" about the changes. This again portrays a fundamental misunderstanding; Brews ohare, you should not edit an article on a topic if you do not even understand its definition.--JohnBlackburne^words_deeds 20:40, 23 July 2014 (UTC)

Reintroduction of negative feedback amplifier diagram in the lede

In this edit User:GliderMaven reintroduced a diagram of an ideal negative feedback amplifier as an example of negative feedback. This reintroduction was made without explanatory comment in the text (needed to point out the oddity of this example) or on this talk page. User:GliderMaven may be unaware that, technically speaking, this circuit does not satisfy even the vague definition of the lead sentence defining feedback, which is:

[Something that] "occurs when the result of a process influences the operation of the process itself in such a way as to reduce changes".

The problem here is that the negative feedback amplifier does not use the result of a process (amplification in this case) to reduce changes in behavior of this system. Instead it directly reconfigures the the system so changes in open-loop gain are irrelevant, and no effort is made to reduce these variations. What negative feedback does do in this case, is not to reduce changes such as fluctuations or variations in the open-loop amplifier itself, but instead it directly alters the circuit itself so its operation has no dependence upon the open-loop gain at all, so long as it is large. The closed loop gain becomes:

{\frac {O}{I}}={\frac {A_{OL}}{1+\beta A_{OL}}}\approx {\frac {1}{\beta }}

and is entirely independent of the open loop gain so long as this gain is large enough. So, the introduction of negative feedback using this example is not an example of the customary usage of this term in a wide variety of situations from management techniques to homeostasis (the most common usages, which are much closer to error-controlled regulation), but is something else altogether, not an example of the definition provided, despite its vagueness. Brews ohare (talk) 18:54, 22 July 2014 (UTC)

Um... yeah, except no. Your OR is completely faulty.

The circuit is usually used as an amplifier, and it is certainly used because it effectively linearises the amplifier, but it certainly works using negative feedback.

The part of the circuit after the + sign is the quantity subject to negative feedback; and it is negatively fed back from the output of the circuit; that particular point is maintained by negative feedback at close to zero (volts). I believe that exactly meets the definition Wikipedia uses.

Indeed, if it did not, Wikipedia's definition would require modification, not the diagram being removed, this is literally a classic textbook example of negative feedback. How is it you could be editing this article and not be aware of that?GliderMaven (talk) 22:08, 22 July 2014 (UTC)

There is no call to get personal about this with the "How could you" stuff and the specter of WP:OR, which is not supported here. It has been pointed out that the example of the negative feedback amplifier does not comply with the definition of the lead sentence. So an adjustment is necessary if this amplifier is to become an example. Your description of some voltage being maintained near zero in this circuit by the mechanism of negative feedback is incorrect, as the math leading to the closed loop gain proves beyond any controversy.

Please propose a more general definition that can include the negative feedback amplifier. Brews ohare (talk) 03:59, 23 July 2014 (UTC)

Unfortunately this is nonsense; the existence of feedback within a circuit or system cannot, in general, be easily established by looking at the system inputs and outputs; the variables being controlled are often internal values, as here really, where the internal value is being maintained quite accurately at zero via negative feedback. Your maths on the input/output relationship proves nothing at all.

Unless you have a reference to your completely bizarre claim that a negative feedback inverting amplifier does not in fact use negative feedback I'm certainly not going to be editing the article to remove the diagram, and I will revert any such very clearly OR edits you may make. This is electronics 101 stuff; negative feedback is used incredibly widely in all kinds of circuits, and very often follow the form given.GliderMaven (talk) 10:48, 23 July 2014 (UTC)

GliderMaven: I do not doubt your sincerity in these remarks, but you might recognize that they consist only of your assertions on the matter. In particular, your claim that the negative feedback amplifier circuit is based upon control of internal variables does not identify what these variables might be, or how the circuit "controls" them. In fact, all analyses of the effect of this circuit upon variations in the open-loop gain A_OL are based upon introducing these variations in the closed loop gain expression. Thus, supposing the variation in this gain to be δA, this expression provides the change in closed loop gain due to these variations as (see, for example, Sanitram Kal):

{\frac {O}{I}}={\frac {A_{OL}+\delta A}{1+\beta (A_{OL}+\delta A)}}\approx {\frac {1}{\beta }}+{\frac {\delta A}{A_{OL}}}{\frac {1}{\beta }}(1-{\frac {1}{\beta }})

and, because A_OL is large, the δA/A_OL contribution of the gain fluctuations is small. This reduced influence of gain fluctuations is not due to any attempt to modify these fluctuations in A_OL (which are left entirely unaddressed), but is a consequence of circuit topology making them irrelevant. Thus, the emphasis in the lead sentence upon direct control of variations is not illustrated by this example, which "controls" the effect of fluctuations, not by their mitigation, but by reducing their presence to irrelevance. Brews ohare (talk) 15:36, 23 July 2014 (UTC)

A way to incorporate the negative feedback amplifier as an example of negative feedback would be to revise the definition of negative feedback. It is evident that the negative feedback amplifier involves feedback from the output to the input. It is called negative feedback because this feedback is subtracted from the input. However, it also is clear that this feedback has no effect upon the operation of the open-loop amplifier or its fluctuations. It does have the effect of making these fluctuations unimportant to the closed loop amplifier operation. So a general enough definition of feedback has to avoid the impression that it necessarily controls fluctuations, and has to emphasize that it limits their effect, either directly (as in error-controlled regulation) or indirectly by rendering them ineffective (as in the negative feedback amplifier and various op-amp circuits). However, it might be noticed that feedback in the negative feedback amplifier has many advantages completely unrelated to the control of open-loop gain variations, and the design of such amplifiers is largely predicated upon these other aspects, and does not emphasize the role of sidelining open-loop gain variations. From that respect, the negative feedback amplifier is not a poster-child for negative feedback. In particular, the approach of designing independence from variations in place of their mitigation is completely different from the use of negative feedback as commonly understood in homeostasis and management systems, which are based upon error-controlled correction. Brews ohare (talk) 19:30, 23 July 2014 (UTC)

For unidentified reasons, Blackburne has elected to revert the sourced, generalized definition of negative feedback that is capable of including the negative feedback amplifier. He has not engaged in the above discussion explaining the issues. He has amply illustrated no intent of using sources or argument to support his actions. Brews ohare (talk) 00:21, 24 July 2014 (UTC)

The issues were discussed in the previous section, #Unacceptable, and I've given my reasoning there, as have other editors already.--JohnBlackburne^words_deeds 00:44, 24 July 2014 (UTC)

The issues described in the thread #Unacceptable you refer to are aimed at an earlier form for the lede, and do not apply to the paragraph you have reverted. IN addition, you have not addressed the inapplicability of the version you favor to the negative feedback amplifier, as documented immediately above. Frankly, I see no evidence that you have any source that supports your activities here. Brews ohare (talk) 00:56, 24 July 2014 (UTC)

I completely support John Blackburne's reversion. Your weird lead fails to differentiate between what negative feedback is, what it can do and what it is used for or how it is sometimes implemented. These are almost completely different things. The consensus, stable definition identifies what negative feedback is. Your attempts to base it on what it is used for (such as control or error correction etc.) or how it is implemented, are deeply flawed.GliderMaven (talk) 01:11, 24 July 2014 (UTC)

Well, GliderMaven, I am sorry that you have decided that sources and reasoning play no role here, and that you refuse to come to grips with the explanation for insensitivity to gain fluctuations provided above by Sanitram Kal, typical of virtually every source that discusses the negative feedback amplifier, and again typically, where negative feedback is defined in the context of the negative feedback amplifier as simply a subtraction from the input of a modification of the output. The present definition in the lede is incompatible with its meaning for the negative feedback amplifier, and does not agree with sources discussing this example.

Of course, there are alternative ways to define the subject for a wider context, and the one reverted could be improved upon. You can choose to duck that issue here in favor of some vague and unsourced arguments, but the inapplicability of the lede you two favor to the negative feedback amplifier remains solid and sourced and universal in circuit literature, and is not dealt with by you or by Blackburne. Brews ohare (talk) 01:42, 24 July 2014 (UTC)

Brews ohare your definition makes no sense as a general definition. Two editors have explained this two you today, two more in the section immediately above. If you really cannot understand our reasoning then you frankly don't understand this topic well enough to be editing this article. I would add that with four editors against your changes consensus is clearly against you, and was a week ago. Repeatedly trying to change the lead to your preferred definition against this consensus is disruptive editing, and you should consider yourself warned of this.--JohnBlackburne^words_deeds 05:21, 24 July 2014 (UTC)

Blackburne: You have yet to address the issue raised: namely, the present lead sentence says:

"Negative feedback occurs when the result of a process influences the operation of the process itself in such a way as to reduce changes."

As pointed out in exhaustive detail and supported by Sanitram Kal and countless other sources, this definition does not describe negative feedback as it applies to the negative feedback amplifier. Feedback does not reduce changes in the open-loop gain. I cannot understand your reasoning on this score, because (i) you have provided none, and (ii) no source extant supports this version of negative feedback as applicable to the negative feedback amplifier. Can you focus upon this problem? It requires a different definition. Go ahead and provide one general enough to include the negative feedback amplifier. Brews ohare (talk) 05:31, 24 July 2014 (UTC)

There are lots of issues, but the first sentence is a good place to start:

(a) Negative feedback occurs when the result of a process influences the operation of the process itself in such a way as to reduce changes.

(b) Negative feedback occurs when a signal related to a system's operation is subtracted from its input.

Sorry to be blunt, but previous experience shows that time spent debating the issue is time wasted. Let's go with (a) which is good, and forget about (b) which is some mystical attempt at who-knows-what. Johnuniq (talk) 08:15, 24 July 2014 (UTC)

... the negative feedback amplifier does not use the result of a process (amplification in this case) to reduce changes in behavior of this system.. Correct! But irrelevant! Nobody is saying that every possible result/consequence of a process is fed back! That would be crazy. That doesn't mean you get to pick a result and note that it isn't being fed back, and claim that there's no negative feedback! The gain is not being fed back. What is being fed back in this case is the output voltage. The gain is not being measured, nor fed back. The gain is really a meta result of the process, it's not a direct result of the amplification process; the output voltage is. You have to take at least two measurements and do a calculation to even measure the gain of a real circuit.

I mean, you could measure and feedback the gain... in a completely different circuit. But that's not what's happening here.GliderMaven (talk) 11:32, 24 July 2014 (UTC)

Johnuniq: Your point (a) is compatible with negative feedback in the form of error-controlled regulation. However, the basis of negative feedback in the negative feedback amplifier is not of this form, as has been supported above by reference to the typical sensitivity analysis of closed-loop gain by Sanitram Kal. On the other hand, (b) is applicable to the negative feedback amplifier, as is supported by Kal, who uses a subtraction as an indicator of negative feedback in this context, and addition as indication of positive feedback. This simplified sign-based classification is used by all circuit-analysis texts in this context. Point (b) is compatible with error-controlled regulation as well, but it does not require the detection of an error that is the basis for error-controlled regulation. This is the issue at stake here.

If you have the courage of your convictions, you would find sources that suggest the negative feedback amplifier is an example of error-controlled regulation. You would find that cannot be done, and you would then perhaps become persuaded to change your opinion to agree with sources. In any event, conviction born of me-too's is not how accuracy is to be established. Brews ohare (talk) 12:11, 24 July 2014 (UTC)

GliderMaven: You are correct that a fraction β of the output voltage is fed back, and is subtracted from the signal input. The subtraction makes this feedback negative according to Kal, while addition would make it positive feedback. However, you are at variance with sources when you suggest that there is some form of control of some internal variables at work, as is the case in error-controlled regulation. Nothing is measured in the negative feedback amplifier circuit, and no errors are used to control the operation. You realize that there is no attempt in this circuit to monitor this gain and 'fix' these fluctuations - they simply do not matter because of the circuit topology. Fluctuations in open-loop gain are brought up in most treatments of the negative feedback amplifier, and that is why they were mentioned here. This circuit has some connection to error-controlled regulation in that both approaches make fluctuations less important, although by entirely different means.

So, the form of the lead sentence is incorrect in this instance in saying this circuit is based upon "the result of a process influenc[ing] the operation of the process itself in such a way as to reduce changes". That does not happen here. No 'changes' are 'reduced'. That is why this definition has to change, or the diagram has to be explained as not illustrating the definition. Brews ohare (talk) 12:11, 24 July 2014 (UTC)

You already stated your irrelevant and critically wrong opinion several times; and everyone else has understood that you are wrong. The quantity that is being actively controlled is the voltage after the adder; and it is changes to that (caused by noise or variations of the input) that are reduced via the feedback loop. It is trivial to show that when that voltage is well controlled at zero, that the system's overall gain is almost completely independent of the amplifiers gain.GliderMaven (talk) 13:39, 24 July 2014 (UTC)

You discuss noise variations at the input, which are unaffected by feedback. However, noise introduced by the amplifier itself is reduced, as already described. If it is trivial to show what you mean by noise suppression with some or another 'voltage well-controlled at zero', please elaborate or supply a source. I suspect you are referring to the op-amp voltage amplifier and not to the negative feedback amplifier, but maybe you can clarify your position?. Brews ohare (talk) 15:30, 24 July 2014 (UTC)

For the f***ing nth time, look at this diagram:

The point on this diagram to the immediate right of the '+' sign, that is also connected to the input to the amplifier A; the voltage of that wire is the only bit that is subject to negative feedback. It is actively maintained by the circuit at zero volts, independent of noise in the circuit and input variations, by negative feedback. The output.. is not directly maintained. The input... is not maintained by negative feedback. The overall gain of the circuit, is not specifically maintained by negative feedback in any way.

That single point in the circuit is subject to negative feedback. That's why it's a negative feedback amplifier, because it uses negative feedback to work.

So where you said above: "There is no monitoring of an internal state and adjustment of that state", that is flat wrong. That point in the circuit IS monitored and IS adjusted, and when it is adjusted correctly, then the output voltage and input voltage can be shown to be in a predictable ratio (i.e. stable, very linear gain).

And yes, you can show that the resulting overall gain is very stable and that the overall gain certainly has a secondary effect on the output. But the gain is not directly controlled by negative feedback in any way at all. It's a side-effect of the circuit, almost a fortuitous thing, that using the feedback in that way nearly cancels out the variations and non linearities in the amplifier. That's why the maths you keep spouting showing input/output relationships are completely and utterly irrelevant.

And before you say something stupid like 'that's an internal point, it doesn't count', you must also consider that a point that is not within a feedback loop can never be directly stabilised by negative feedback, because the loop doesn't close around it. Significantly neither the input or the output are internal to the loop; but the output is obviously only stable because (if) the controlled point is.

The diagram is actually very general and common but there are certainly more complex negative feedback diagrams with multiple loops and so on.GliderMaven (talk) 17:03, 24 July 2014 (UTC)

GliderMaven: If you examine Kal's analysis linked above, you will find that the node at the input to the open-loop amplifier to the right of the summer is not at zero volts. If this voltage is labeled V_i, the output voltage of the circuit is A_OLV_i, and the voltage fed back to the summer is βA_OLV_i. If the input signal is V_s, then the input voltage at the input node following the summer satisfies V_i=V_s–βA_OLV_i. Solving for V_i, one finds:

V_{i}={\frac {V_{s}}{1+\beta A_{OL}}}\ .

This node voltage is zero volts only if the signal is zero. The output voltage V_o satisfies:

V_{o}=A_{OL}V_{i}={\frac {A_{OL}}{1+\beta A_{OL}}}V_{s}

which provides the closed-loop gain as:

A={\frac {V_{o}}{V_{s}}}={\frac {A_{OL}}{1+\beta A_{OL}}}\approx {\frac {1}{\beta }}

which is the well-known standard expression for the closed-loop gain. So whatever one can make of your subsequent remarks, at least this portion is mistaken. IMO, this erroneous conception is possible only because you have not taken the time to examine the careful discussion of the negative feedback amplifier by Kal, which does not have anything in common with your description. Brews ohare (talk) 18:42, 24 July 2014 (UTC)

Although you seem not to recognize this point, we do agree that the open-loop gain is not modified by negative feedback. However, I do not understand why you think the math I "keep spouting showing input/output relationships are completely and utterly irrelevant" when they serve to establish the point you find attributable to 'monitoring', namely that "the output voltage and input voltage can be shown to be in a predictable ratio", that ratio being, of course, 1/β. Brews ohare (talk) 19:04, 24 July 2014 (UTC)

Well, yeah, it's not exactly at zero, feedback in this case cannot stick it precisely at zero all the time, it will vary a small amount depending on the gain of the amplifier (which is usually very large) and the input voltage and the non linearities, but the feedback enormously reduces the size of the variations, and this behavior meets the definition used in the lead of the article for negative feedback.GliderMaven (talk) 19:19, 24 July 2014 (UTC)

No, GliderMaven, this is not an issue of 'exactitude', nor of the amplifier somehow striving to make this voltage exactly zero. The node voltage V_i is at a value V_s/ (1+βA_OL), and if the signal varies by a factor of 10, the node voltage V_i will vary by a factor of ten as well, which is why the output voltage also will vary by a factor of 10/β, and as β < 1, this will be an amplification of the signal. In short, the entire operation of the amplifier qua amplifier is due to the variation in node voltage V_i. The negative feedback amplifier is not operating based upon the "feedback enormously reduc[ing] the size of the variations", and consequently it does not meet the definition provided, even when written in bold face. Please read Kal or some other source. Brews ohare (talk) 20:37, 24 July 2014 (UTC)

So if the input is at 0.1 volts, output is at 5 volts, β is 50, A_OL is 1000, then this 'non zero point' is at 0.005 V. Oh I'm sorry, that's not 0 volts at all. And if the input voltage doubles, this point goes up to 0.01 V. Oh wow, that's such horrible stability, the input has varied by 0.1 volts and the fed-back point has gone up by 0.005 V. I can certainly see why you are upset about the bit in the lead that refers to negative feedback acting 'in such a way as to reduce changes'.

I can only apologise on behalf of myself, all the textbooks, all the books on cybernetics, all the other people that have written on the talk page for our completely inadequate understanding of this very basic circuit. I'm just glad that Brews Ohare has understood it in a way nobody else could; he truly is The One.GliderMaven (talk) 00:51, 25 July 2014 (UTC)

@GliderMaven: Thanks for continuing this, but some people are able to argue indefinitely about anything, and it's fine that they have the last word. No source contradicts (a), and a source showing something about gain stability is showing something about gain stability—not the meaning of negative feedback. The current version of the article (11:18, 24 July 2014 GliderMaven) is good. Johnuniq (talk) 00:06, 25 July 2014 (UTC)

I agree, I pulled one of the sources that Brews referred to, and there was nothing that obviously disagreed with it.GliderMaven (talk) 00:51, 25 July 2014 (UTC)

Continuation of thread discussing the lead sentence

Well, the source's failure to mention GliderMaven's ideas cannot be taken as support for them. Rather, this highlighted link identifies positive feedback as simply feedback that is added to the input, while this link identifies negative feedback as simply feedback that is subtracted at the input. This link makes the same point. Nowhere, as GliderMaven points out, does this source suggest reduction of unidentified "changes" are involved, only reduction of the input signal because of subtraction.

So, at this point this source is clear on what feedback means in the negative feedback amplifier. This definition does not involve corrective measures, which are the point of negative feedback in error-controlled regulation. Force-fitting these two approaches into the present lead sentence crafted for the error-controlled case, and then incorrectly arguing on this Talk page that the mechanism of error-controlled regulation is operative in the negative feedback amplifier is unhelpful in crafting a superior, more general definition, that can include both mechanisms. The idea that error-controlled regulation underlies the operation of the negative feedback amplifier has not been sourced (and cannot be, as it is not the case). That idea is WP:OR. A broader definition is necessary (and possible). Brews ohare (talk) 14:03, 25 July 2014 (UTC)

That's how they do work in practice, and that can be sourced. You need to go away now. Everyone at this point have presumably worked out that you don't really have much clue how these circuits really work. It's one thing to quote a textbook on the equilibrium state, it's another to run a circuit simulation or put probes on a real circuit and watch the changes ripple through; or even just to work out in your head what happens when input voltages perturb the circuit by doing a step change; and you clearly haven't done any of these things.GliderMaven (talk) 15:37, 25 July 2014 (UTC)

The assertion that error controlled regulation applies to the negative feedback amplifier can be sourced is shown as valid only if the sourcing is done. So far, that has not happened. So, GliderMaven your unsupported views of how the negative feedback amplifier operates to date are unsourced misconceptions that do not agree with the specific links to specific sentences in Santirim Kal's book provided above. Brews ohare (talk) 16:18, 25 July 2014 (UTC)

Genuinely, I can't be bothered. Plus, examination of one source you've quoted indicates you haven't understood it, and so it's pointless anyway; you would probably just think and claim it meant something else.GliderMaven (talk) 20:19, 25 July 2014 (UTC)

GliderMaven: Not very accommodating of you, and insulting besides. You have yet to address the very clear request that some source be presented that agrees with you that error-controlled regulation is at work in the negative feedback amplifier, nor have you addressed the quite different picture laid out by Santiram Kal, who presents the standard approach found in circuit textbooks. In effect, you have set yourself up as an irrefutable authority upon your own say-so. As a person who has taught and written technical articles in this area, my own expert opinion is that you have only a partial understanding of this circuit. However, these assessments are not part of WP, which simply reports what reputable sources say. The lead sentence in this article is not sourced and is overly narrow, and should be changed to agree with sources describing not just error-controlled regulation, but also the negative feedback amplifier. Brews ohare (talk) 21:30, 25 July 2014 (UTC)

A simple change that might work is provided by Runkel & Runkel: "if increases in the output cause the feedback to decrease the output, the feedback is called negative feedback". This definition avoids reference to reducing changes and applies equally to cybernetics and to the negative feedback amplifier. It also is sourced. Brews ohare (talk) 22:29, 25 July 2014 (UTC)

Righttttt. So you're barefacedly claiming to be the 'expert', an expert that, not incidentally, has failed to come up with a source that a negative feedback amplifier is not an example of an error controlled regulator, and we're supposed to make the changes that you have indicated and tried to force through multiple times, because you're an 'expert', but somehow you're also cynically claiming that I'm the one to have supposedly somehow set myself "up as an irrefutable authority upon your own say-so."

You do know that claims of expertise such as you have made are irrelevant in Wikipedia, and such claims have a long history of being proven fraudulent?

And the referenced quote is not good "if increases in the output cause the feedback to decrease the output, the feedback is called negative feedback", it sounds like the output must go both up and down. While some feedback loops will do that, some won't, especially if they are critically damped, but they're still negative feedback loops.GliderMaven (talk) 01:32, 26 July 2014 (UTC)

Well, not quite. What was said is that expertise doesn't matter, it's sources that matter. You have supplied neither.

You have misunderstood the quoted sentence: "if an increase in the output causes the feedback to decrease the input, the feedback is called negative feedback". The conditional if means "in circumstances such that". There is no supposition about the circumstances that might cause an increase in output, or even whether this hypothetical circumstance can arise in practice. But if we imagine it occurring, the feedback can be identified if this hypothetical increase in output causes a corresponding hypothetical decrease in the input. The sentence seems clear to Runkel & Runkel, and it is very nearly a paraphrase of that by Kal. So I don't think the sentence is at all obscure, neither is it an uncommon way to identify negative feedback. Brews ohare (talk) 02:47, 26 July 2014 (UTC)

Me-too consensus in place of reporting sourced material

In this edit Johnuniq removed the following material and diagram that provides a sourced definition of negative feedback.

Negative feedback in a system or process returns (feeds back) features of the system output to the system input so that, should an increase in output occur, that increase causes a decrease in the input.^1,2,3

¹ Philip Julian Runkel, Margaret Runkel (1984). A Guide to Usage for Writers and Students in the Social Sciences. Rowman & Littlefield. p. 107. ISBN 9780822603825.

²Santirim Kal (2004). Basic electronics: Devices, circuits and IT fundamentals (Paperback ed.). Prentice-Hall of India Pvt.Ltd. p. 191. ISBN 9788120319523.

³Rahul Sarpeshkar (2010). Ultra Low Power Bioelectronics: Fundamentals, Biomedical Applications, and Bio-Inspired Systems. Cambridge University Press. p. 32. ISBN 9781139485234.

The links to the three sources highlight sentences that are close paraphrases of this definition, which is found in both of them despite their coming from widely differing subject areas.

The reason provided for removing this sourced definition is: original is good, per consensus on talk. Consensus as to personal opinions is hardly a substitute for reporting what sources say. And an incapacity to come to grips with sourced material and to prefer personal opinions in its place only accidentally, if ever, leads to accurate reporting of sourced material.

The main problem with this revert is that it substitutes an unsourced, vague definition that is inapplicable to the negative feedback amplifier, possibly the most common example of negative feedback. This inadequacy has been explained carefully in the above Talk page threads, and none of the authors behind this 'consensus' has dealt with the matter or considered any source. Rather than finding a way to provide a sourced definition applicable to both error-controlled regulation and to the negative feedback amplifier, we have a 'consensus' of self-congratulatory me-too's disregarding sources entirely. Brews ohare (talk) 04:16, 27 July 2014 (UTC)

Consensus as in the discussions above. Wikipedia operates by consensus, as you very well know, and the consensus is for the current version. Your repeatedly ignoring consensus in favour of your own interpretations is disruptive and has to stop.--JohnBlackburne^words_deeds 07:02, 27 July 2014 (UTC)

John, WP is supposed to operate on informed consensus regarding a balanced presentation of sourced opinion. The only sources present in this discussion are those I've introduced, notably Kal and Runkel & Runkel, whose reputably published opinion agrees entirely with the material reverted by Johnuniq. What remains in the main article is an unsourced majority view that is inapplicable to the negative feedback amplifier.

Of course, editors can ignore policy with impunity, even that about sourcing instead of editors' opinion, to indulge a penchant for polemic and sound bites. The cost is in quality. Brews ohare (talk) 12:49, 27 July 2014 (UTC)

No, you're one doing soundbites. The definition you tried to edit war into the article is incoherent and incorrect. A feedback loop cannot change the input, for most sensible definitions and situations that have an 'input'. In the negative feedback loop, the input is in no way, shape or form affected by the feedback loop. The input is the starting or boundary condition or dynamic external inputs and cannot change. Hence the definition is incorrect.

Or, if this is some specialised, jargon form of the term 'input' then I'd like to see where that is defined, but even then it shouldn't be in the first few sentences because that's supposed to be, as far as possible, the one part of the article that practically anyone can read and understand and as free of jargon as possible.GliderMaven (talk) 18:53, 27 July 2014 (UTC)

GliderMaven: The meaning of 'input' and 'output' for a simple case are labeled in the figure. The three linked statements in the three sources above 1. 2, 3 are unanimous in saying an increase in output results in negative feedback reducing the input. Change in the input due to feedback also is clear from the diagram, where the input is reduced from I to I–βO. So, are you contradicting the sources and confused by the diagram? Brews ohare (talk) 19:37, 27 July 2014 (UTC)

I'm not confused in any way, I'm simply trying to be very precise and clear. The way you are trying to present it is not helpful; the input is not being reduced; the input is (can be) the same at all times; it's the output from the adder (or equivalent point) that is reducing. I suppose in some systems the only 'input' is the initial boundary conditions and then the values of the system vary from there; in that situation the system in a real sense the input is varying. But that's a less general situation than a system with an input which can vary over time, and very many systems have that. Even the weather has an input in the form of solar forcing where the solar irradiance will vary over time in addition to the initial atmospheric state.GliderMaven (talk) 21:05, 27 July 2014 (UTC)

While in Wikipedia while we certainly can't do any OR, we can and are allowed to phrase things in equivalent ways. In fact for legal/copyright reasons we often cannot use exactly the same terminology.GliderMaven (talk) 21:05, 27 July 2014 (UTC)

GliderMaven: Let's agree that your quarrel over terminology about 'input' is not with me. It is used by 1. 2, 3. If you think these sources have been inaccurately reported upon, that is the issue to raise here. Perhaps you could comment upon (i) what the sources actually say, as you see it, and whether the proposed sentence portrays their views correctly, and (ii) whether you agree that both the negative feedback amplifier and systems employing error-controlled regulation can be included in this definition?

If I understand you correctly, you interpret this definition as excluding error--controlled regulation, such as thermostatic control of house temperature. That is not the intent of the sources or their wording, as I understand them. To elaborate upon this example, in error-controlled regulation the "feature of the system" fed back to the input is the departure of some status parameter from its set point, and the feedback is negative because any change in input that increases this error is reduced by the feedback. These systems are often more complex than that in the simple diagram because the feedback is not just a fraction of the output, but a more complex function of the system parameters that may include algorithmic interpretation of the error signal, selection and activation of some control mechanism, and execution of a response by some 'effector' that actually implements the reduction of the input disturbance. Another difference, of course, is that the simple diagram does not include any set-point or error determination, which is also true of the negative feedback amplifier. However, the wording applies to both types of system.

Please comment on the two issues raised at the beginning. Brews ohare (talk) 15:01, 28 July 2014 (UTC)

I have added a figure for error-controlled regulation using negative feedback. This figure uses 'input' to mean 'input to open-loop system'. That seems to fit your conception of 'input' better than the simple diagram where 'input' looks like it refers to 'signal'. In the simple case the 'input' as designated by the sources also refers to 'input to open-loop system', so the simple diagram, although common in electronics texts, may be confusing to the present discussion. I've modified this figure accordingly. Brews ohare (talk) 15:46, 28 July 2014 (UTC)

If the objective here is to find a diagram for the lede that is sufficiently generic to cover all uses of the term (and so matches the opening sentence) may I suggest the diagram from the feedback article, with suitable caption:

"When two elements influence each other in such a way that a change in one is opposed by the other, this is negative feedback."

With the understanding that "element" can refer to any physical, electronic, biochemical or human system or sub-system.

The negative feedback amplifier is important historically in this subject. It isn't as simple as it first appears however, and outside the electronics field is a long way from being a generic example. It is also thoroughly covered elsewhere. Trevithj (talk) 19:33, 29 July 2014 (UTC)

But that isn't correct, since if they both negatively feedback to each other, that makes it into positive feedback, so that definition fails also. It's the overall feedback loop sense that makes it negative feedback.GliderMaven (talk) 21:02, 29 July 2014 (UTC)

Hmm? Positive feedback would be when a change in one is reinforced by the other.Trevithj (talk) 03:25, 30 July 2014 (UTC)

That's right, it's positive feedback if they both oppose each other. The loop gain multiplies, in multiplication two negatives make a positive, and then the loop gain is positive, and you have positive feedback. For example if the first one moves up a tiny bit it pushes the second downwards, the second moves downwards, as it does so it pushed the other upwards; this change makes the first push the second even more downwards, which moves the second downwards and pushes the first up even harder. The second races off to negative infinity and the first goes to positive infinity; classic positive feedback behaviour.

That's when they're both wired negatively to the other. For it to be negative feedback, one of them should be negative wrt other, and the other should be positive. Let's say the first is wired negatively. Then if the first one moves up a bit it pushes the second downwards, which moves downwards but it pulls the first down with it; tending to cancel out the movement; classic negative feedback.GliderMaven (talk) 13:26, 30 July 2014 (UTC)

Understood. But the proposed definition doesn't say they oppose each other. I think we are delving too deeply into mechanism, which is what is fueling much of the debate. The two components don't have to be "wired", for example. Think steam governors, or Drebbel's thermostat - circa 1600. They both oppose changes in another element (steam engine and incubator, respectively). It doesn't follow that the other element opposes changes in them. Trevithj (talk) 09:57, 2 August 2014 (UTC)

GliderMaven: Can you respond upon (i) what the sources actually say, as you see it, and whether the proposed sentence portrays their views correctly, and (ii) whether you agree that both the negative feedback amplifier and systems employing error-controlled regulation can be included in this definition?

There is no point in arguing your own viewpoint - you must address sourced views, whether they are these sources 1. 2, 3, or others. Brews ohare (talk) 22:25, 29 July 2014 (UTC)

Yes, having checked the third of the references at least, I am happy that you were reverted perfectly fairly and squarely- you had clearly misquoted the references.

You defined it as: "Negative feedback in a system or process returns (feeds back) features of the system output to the system input so that, should an increase in output occur, that increase causes a decrease in the input."

Whereas the third reference stated: "In negative feedback loops, the output feeds back to decrease the effects of the input" (my emphasis) and it gives a very similar picture 2.2 to the negative feedback amplifier. That's a more reasonable definition, although replacing 'the output' with 'an output' would be even more accurate.

The second one seems to be incorrect, but if you read it in context it implies a rather different meaning, the plain words you quoted are wrong when taken out of that context. That's another sense of what I wrote above about quoting things blindly, it's not enough to quote things, you often have to correctly summarise multiple sources to get an encyclopedic self-contained definition of the topic. It's very easy to get a too-narrow or sometimes too wide definition.

The first reference didn't load correctly for some reason.GliderMaven (talk) 23:26, 29 July 2014 (UTC)

The Runkel & Runkel reference did not load correctly because Google books changes the pages that are accessible from time to time. It says "if increases in the output cause the feedback to decrease the output, the feedback is called negative feedback." You might try this link, or try the words without grouping them. In any case, I don't think any of these sources contradict each other. Brews ohare (talk) 01:04, 30 July 2014 (UTC)

What is this all about? Do you want the reverted edit to be reinstated? The one which misleadingly describes the topic of this article? How do these discussions at Wikipedia normally finish up? Johnuniq (talk) 04:02, 30 July 2014 (UTC)

What do I want? I want a sourced definition that distinguishes negative feedback from feedback in general, and one that does not apply only to error-controlled regulation like the thermostatic control of house temperature. In other words, a sourced definition that applies even where no error assessment and no set-point are used. An example of such a sourced description is: 1, 2, 3, 4, 5, 6, 7.

How do these discussion end? Ideally they end when the various sources are examined and properly summarized. Brews ohare (talk) 04:29, 30 July 2014 (UTC)

At least the current definition is correct. I'm surprised you can't recall any other ways that discussions have ended. Johnuniq (talk) 04:43, 30 July 2014 (UTC)

As it is unsourced, it is so far correct only in your opinion. In my opinion it is limited in scope to error-controlled regulation. It also is unclear about its terminology. Brews ohare (talk) 05:31, 30 July 2014 (UTC)

Glidermaven's comment that reference 2 is "wrong" and also "taken out of context" and read properly "has a very different meaning" is simply mystifying. First, Glidermaven has not said what, in his opinion, the 'true' meaning is, or what is the 'correct' context. To me it is clear that the context is the negative feedback amplifier, and when it is said that the feedback is 180 degrees out of phase with the signal that means the input is the signal minus the feedback, exactly as depicted and discussed using the diagram at the start of this thread. Brews ohare (talk) 14:54, 1 August 2014 (UTC)

Summarizing sources

May I make another attempt to do the summarizing? In response to GliderMaven's criticisms, I'm applying Brews' referenced examples to the proposed alternative diagram (above):

Ref 1, changes in the amplifier block are opposed by the feedback circuit block (from section 6.2)
Ref 2, changes in the basic amplifier (A) are opposed by the feedback network (B) (from figure 6.1)
Ref 3, changes in a(s) are opposed by f(s) (from figure 2.2)
Ref 4, changes in conveyor belt speed are opposed by the speed controller (from example in text)
Ref 5, changes in open loop gain A are opposed by the feedback factor B (fig 11.1)
Ref 6, changes in arterial blood pressure are opposed by the baroreceptors.
Ref 7, changes in the weight of the animal are opposed by the intake of food.

With respect to the discussion about inputs and outputs, I would note this example: that changes in a reservoir level (an internal state) are opposed by the floodgates (an output). The input to the reservoir is unaffected. I submit this is still negative feedback. I also note that Ref 4 says that any system with feedback has an error signal. That rather contradicts the idea that some negative feedback systems are not error-controlled regulation. However, proposed alternative definition would cover both in either case.Trevithj (talk) 06:38, 30 July 2014 (UTC)

Trevithj: Thanks for initiating a conversation about what sources (1, 2, 3, 4, 5, 6, 7) have to say instead of WP editors' opinions. As you point out, reference 4 and some of the others focus upon error-controlled regulation, which indicates, as you also point out, that the definition espoused by all applies generally to both error-controlled regulation, and to the negative feedback amplifier, and to op-amp circuitry. That is what we need here - a generally applicable and sourced definition of what distinguishes negative feedback from feedback in general, including not only engineering applications, but cybernetics, economics, management, education, and psychology. Brews ohare (talk) 14:15, 30 July 2014 (UTC)

You have identified the general view as something like "changes in something are opposed by feedback when the feedback is negative." How would you word the definition provided by these sources? Brews ohare (talk) 16:34, 30 July 2014 (UTC)

The 'something' could be as varied as a system input or output, an error or status assessment, a public policy, or a mental attitude. Would you agree? Brews ohare (talk) 16:46, 30 July 2014 (UTC)

Agreed, we are looking for as generic a definition as possible. In fairness to GliderMaven's point, we may need to paraphrase sources, rather than quote directly. In fact, given that many sources use specific definitions for specific purposes, we may have to in order to get a general definition.

Re the definitions in the sources: these can perhaps be better covered if we designate A as the regulated element and B as the regulating element. Then "changes in A are opposed by B" should be generic enough to cover most (if not all) cases, and uses familiar enough language for the lede. I believe we should leave the details of what the changes are and how they are opposed to the specific instances further down the article.

Trevithj (talk) 08:51, 31 July 2014 (UTC)

Of course, negative feedback is only one category of feedback in general.Smit & O'Byrne identify negative, positive and bipolar feedback. The point is that this article has mainly the objective of identifying and discussing the negative form of feedback. The general concept of regulation, on the other hand, includes not just negative feedback but all kinds of feedback and even feedforward. For more on types of regulation, see Rhoades & Bell. Concerning distinguishing negative feedback, Runkel & Runkel take the approach of explaining feedback as feeding back information about a system to "controls of the input" and then defines negative feedback as the special case where "an increase in the output cause[s] the feedback to decrease the output." If we want to avoid digressing into what constitutes feedback control and want to generalize beyond 'output' and 'input', we could simply say:

"Negative feedback is any form of feedback in which a change in system behavior causes the feedback to oppose that change."1, 2, 3, 4, 5, 6, 7

Possibly one could add:

"Among many examples, change may be detected as departure of one or more system parameters from their set-point values (as in error-controlled regulation), or as a change in system output, a fraction of which is fed back to adjust the system input (as in the negative feedback amplifier)."

Brews ohare (talk) 14:55, 31 July 2014 (UTC)

A very generic way of defining feedback is in terms of "circular causality". Runkel & Runkel talk in terms of its original sense of "reprocessing". Not as generic, so I feel it should be avoided. Your proposed definition is basically the same as mine (feedback that opposes change), but leaves the definition of feedback open. Which could work, I guess.

Do you have an issue with use of the terms "regulating/regulated"? The terms could be changed. I suggested the "two elements" idea to make the "circular causality" approach clearer, and to get away from any confusion between feedback path, feedback mechanism, and feedback loop. This is based on Ashby's approach:

"When there are only two parts joined so that each affects the other, the properties of the feedback give important and useful information about the properties of the whole."^[1]^: 54

I would avoid getting into bipolar feedback. This covers the case of more than two elements, and both types of feedback, interacting in the one system. Realistic but complex, and so unsuitable for the lede. Also, regards your suggested elaboration (set-point, etc), these are good points to make, but make them in the main article. Keep the lede simple.

^ W. Ross Ashby (1957). Introduction to cybernetics (PDF). Chapman & Hall Ltd.; Internet (1999). p. 54.

Trevithj (talk) 09:32, 2 August 2014 (UTC)

Trevithj: I agree that this proposal basically is yours. I'd say that leaving open the definition of feedback is prudent because there is an article about that and it is likely to cause controversy here that can be avoided. That course makes this article strictly about negative feedback, which seems appropriate to me. As an implementation of this narrow scope, it seems desirable to me to avoid generalities about 'regulation', a broad topic. Likewise, the idea of circular causality is better in the feedback article. My ramblings about regulation and bipolar feedback were simply background, not intended for inclusion in the introduction.

I don't know your reaction, but I'd like to see the single-sentence definition implemented without additional complications, which possibly could be raised later, in the body of the article. If you could support this action, the me-too'ers might agree as well. Brews ohare (talk) 13:30, 2 August 2014 (UTC)

Consensus on feedback

I feel we would be better off with a consensus of how we regard feedback. The issue with the lede at present seems to be a conflict between the verbal definition (feedback=circular causality) and the amplifier diagram (feedback=reprocessing). As I said above, I don't think the original "reprocessing" sense of feedback is generic enough here. Also, since the negative feedback amplifier is mentioned in the History section, in its own sub-section, and in a separate dedicated article, it seems unnecessary to repeat it in the lede. Especially since it is not as simple (or generic) an example as it first appears, and has obviously attracted strong debate in the past.

I suspect your above arguments were aimed at making this point - if we modify the verbal definition to better match the diagram, the lede stops being generic. Better to do it the other way around. The A ↔ B diagram is my suggested alternative, and could be accommodated with a fairly minor rewording. Trevithj (talk) 22:59, 2 August 2014 (UTC)

Trevithj: I don't care about the diagram in the lede. It doesn't matter, and it is illustrative only of a particular example. It could be left out. On the other hand the figure you like is not illustrative of negative feedback, and it's explanation is a digression into feedback that can be found in the article feedback, where one will find your diagram already.

So it would seem that I'd like to keep negative feedback on topic, and see no need to make the lede a rehash of feedback itself, while you'd like to engage the lede in matters already found in feedback. These matters, if they need be brought up again here, could be presented in the body of the article. If it seems essential to do this, a subsection could be written about the connection of negative feedback to various conceptions or applications of feedback. There is no need for this material in the lede, which can be just the definition:

"Negative feedback is any form of feedback in which a change in system behavior causes the feedback to oppose that change."1, 2, 3, 4, 5, 6, 7 Negative feedback can produce stability and reduce the effect of fluctuations. Negative feedback loops in which just the right amount of correction is applied in the most timely manner can be very stable, accurate, and responsive.

along with the present second paragraph about various contexts where the term 'negative feedback' shows up - management, politics, education, control theory ... Brews ohare (talk) 16:01, 3 August 2014 (UTC)

I like your proposed definition - it is about as generic as we can get, is concise, uses non-technical language, and incorporates much of the existing text. I still have some reservations: 1) about the chosen references, and 2) about the diagram. If these reservations can be addressed, I'd happily support the change.

The reservations center around unspoken assumptions, which in the past have lead to much debate on this subject. See, the new definition implies "circular causality" (which I support), whereas most of the references (and the diagram) imply "output-to-input" reprocessing. I don't believe that clarifying definitions is a digression. If the references and the diagram could be modified to support the more general definition, this would help the focus of the topic. Trevithj (talk) 22:25, 3 August 2014 (UTC)

I don't have any great alternatives to offer, but I don't like "change in system behavior", when it's really system output that's most relevant. And I don't like relying on our article on feedback to justify the circular definition. I haven't been following the recent discussion and attempts, so maybe this has already been hashed over. Dicklyon (talk) 00:53, 4 August 2014 (UTC)

I'll mull over Trevithj's comments a while before responding. However, Dicklyon's remark that "it's really system output that's most relevant" is incorrect in almost all cases. For example, in homeostasis the emphasis is on maintaining survivability. In the negative feedback amplifier the emphasis is on making open-loop amplifier variations unimportant by changing how the amplifier works altogether. Many of the sources cited above also use 'input' and 'output' in discussing negative feedback, but that is really not essential to the subject. In any event, using the definition that negative feedback is feedback that opposes changes (including changes in 'output') accurately describes every instance of negative feedback, regardless of one's particular myopia regarding the many forms of feedback. Brews ohare (talk) 01:17, 4 August 2014 (UTC)

Conceptually, the thing that is fed back is most often regarded as an "output", though it does not necessarily need to be. I don't get your point on the amplifier, where it is explicitly the output that is fed back, as your diagram on the article you linked shows, is it not? I'm not being myopic here, just looking for something with more meaning than "change in system behavior", which is not a concept I've encountered this way in feedback analysis. I realize you made it up to cover the range of concepts you have in mind, but it's idiosyncratic, not a normal way of talking about feedback. If you look at how "change in system behavior" is used in books in relation to feedback, it usually doesn't mean at all what you're using it for. Dicklyon (talk) 04:52, 4 August 2014 (UTC)

Hi Dick: You bring up a number of approaches here. Probably we do not have to reconcile our views on the negative feedback amplifier, which is just an instance. From a wider stance, your reservation is based not upon the suitability or eloquence of the definition:

"Negative feedback is any form of feedback in which a change in system behavior causes the feedback to oppose that change."

but upon the use of "change in behavior" as 'idiosyncratic' in this context (a WP slur often used to squelch discussion as merely crank digression). Now that seems to broach the issue of just what feedback is, so we can decide what the 'context' of feedback encompasses, and so decide whether "change in behavior" arises in this context. So we have, for example,

"Feedback is the use of a measurement of some aspect of system behavior to correct or adjust that behavior." Lucertini, p. 103

"Climate forcing is a change in direction... A positive feedback enhances a forcing; a negative feedback works against a forcing".Farmer & Cook, pp. 12, 13

"if an upward movement in X will directly or indirectly cause a change which in its turn will tend to cause a downward movement in X, we have a case of negative feedback" Meade, p. 49

and see Figure 1.1 here, that has no 'input' or 'output' labels.

So, away we go, examining various approaches to feedback, including Trevithj's "circular causality" and bipolar feedback. That is a long conversation, best made in the article feedback where sources can be discussed. This prolix digression can be avoided here by simply restricting attention to what identifies negative feedback, and accepting as everyday language "opposing change", and leaving the issues of what constitutes 'feedback' to the article on feedback and to specific examples found in subsections of negative feedback.

The applicability of this wording to various examples 1, 2, 3, 4, 5, 6, 7 could be examined to see whether it works in these cases. Many of these sources use the words 'input' and 'output' and, Dick, you have suggested the way to go is to make 'input' and 'output' philosophical terms of wide enough meaning to include the use of error-controlled regulation by making the monitoring and measurement of system error an 'output' of the system, and the regulatory actions upon internal parameters as an 'input'. I'd suggest instead that we don't need to make a specialized technical jargon out of 'input' and 'output' by complicated additions to their everyday meanings, but can use the everyday meaning of opposing change, said change being possibly a change in 'output' or a change in internal parameters, or a change in state, or a change in mental attitude, or a change in political policy.

It must be kept in mind that cybernetics, management theory, economics, educational philosophy and on and on also use the idea of "feedback" and "negative feedback" and this topic is not just a matter of control theory or computer programming. Brews ohare (talk) 12:06, 4 August 2014 (UTC)

Dick, your search turns up things like: "The relevant point is the cycle itself: system behavior → feedback → change in system behavior → feedback → change ..."1 and "It is called negative [feedback] because the feedback negates the change in system behavior."2. Is your point that this search turned up only a few supporting results and not thousands? Brews ohare (talk) 14:31, 4 August 2014 (UTC)

My point is that "change in system behavior" usually refers to qualitative behavior, such as changing from stable to unstable, when what feedback usually works on is system state (including typically input and output), not "behavior", which is too vague a term to be useful. Dicklyon (talk) 16:31, 4 August 2014 (UTC)

"Change in system behavior" can be qualitative, but is not so restricted. I suppose your objection really comes down to the implication that this terminology will mislead many readers. In fact, even in engineering applications, the term 'negative' feedback is entirely qualitative, indicating decrease rather than amplification of changes. That includes specifically 'input' and 'output' changes as is exceedingly well documented among the linked sources 1, 2, 3, 4, 5, 6. Brews ohare (talk) 16:46, 4 August 2014 (UTC)

Feedback "works" on a system state (where that is in fact the context) by using some information "about some aspect of system behavior to correct or modify that behavior". Apparently 'behavior' is clear to some reliable sources. Brews ohare (talk) 17:56, 4 August 2014 (UTC)

Dicklyon, it seems to me that it is the word "behavior" that is the sticking point, and that the idea of "opposing change" is basically sound. If so, and noting that we are more interested in a "generic" definition than a "common" one, do you have an alternative to "behavior" in mind? While the word is used - as per Brews' refs - it isn't necessarily central to the definition. Trevithj (talk) 07:55, 5 August 2014 (UTC)

The term 'behavior' is used to identify 'what changes' by Lucertini, p. 103, by Encyclopedia of Human Behavior, and by Lewis. Here is a bunch of sources using 'behavior' in connection with 'feedback'. On the other hand, 'a change in direction' is used by Farmer & Cook, pp. 12, 13, 'a change in a variable X' by Meade, p. 49, a change in 'a process' by Dilts, a change instigating 'corrective mechanisms' by Smit & O'Byrne, and simply unspecified 'changes' by Inglehart and by Heneman & Greenberger. Brews ohare (talk) 13:45, 5 August 2014 (UTC)

It seems Dick finds a change in 'behavior' to be vague, and wants some very definite identification of what exactly changes, and possibly exactly how 'it' changes. While that is feasible for each specific system and each specific feedback mechanism, there does not seem to be an abstract description that would apply generally and precisely to all types of system and all types of feedback. Brews ohare (talk) 16:00, 5 August 2014 (UTC)

Most uses of "behavior" in the "bunch of sources" you link are not aligned with the meaning of how you're trying to use the term; e.g. the first one. In striving for complete generality, you make the definition essentially meaningless by using this term. Many of the social-science uses of "feedback" talking about "behavior" can't really be related to the usual concept of "negative feedback"; like this one. In more engineering-oriented sources about feedback and behavior, like this one, behavior is not the thing in the feedback loop. It just doesn't work, according to most sources, as I pointed out before. Dicklyon (talk) 20:27, 5 August 2014 (UTC)

Perhaps this is better:

"Negative feedback is any form of feedback in which a change in a system's state or output causes the feedback to oppose that change."

Although I'm still not very happy with the degree of apparent circularity in "X feedback is ... feedback in which ... causes the feedback to ... Dicklyon (talk) 20:48, 5 August 2014 (UTC)

Yes, defining something in terms of itself is a bit dodgy. But if we focus on trying to define negative in the context of feedback, then it can be simpler still:

"Negative feedback is feedback that opposes change."

Contrast this to the often-used social-sciences sense...

"Negative feedback is feedback that is unpleasant to receive."

...which is definitely outside scope of the present article. Although that is not to say that we should disregard the social-sciences - they have plenty of cases where things "oppose change." Trevithj (talk) 23:54, 5 August 2014 (UTC)

The definition that:

"Negative feedback is feedback that opposes change."

sounds perfect to me. Sources are needed, though, and these may require introduction of examples, such as "in error-controlled regulation, for example, negative feedback is use to reduce changes identified as system errors"{ref} Other examples are discussed specifically in the body of the article. Brews ohare (talk) 00:22, 6 August 2014 (UTC)

That sort of definition seems to be supported by lots of sources, such as one that says "Negative feedback is a circular chain of effects that opposes change." and one with "Negative feedback occurs when the output of a system acts to oppose changes to the input of the system, with the result that the changes are attenuated." Dicklyon (talk) 04:18, 6 August 2014 (UTC)

To take perhaps a more scientific approach - are there sources that contradict this definition? I ask with some hesitation, since I have come across some careless definitions. But we want a generic definition, and the real test is its applicability. Trevithj (talk) 04:28, 6 August 2014 (UTC)

There are plenty of sources among those found by Dick that support this definition. That means it is widely used definition and should be reported here. If there are conflicting views, they can be added when found in keeping with NPOV. We can therefore proceed. Brews ohare (talk) 05:08, 6 August 2014 (UTC)

Made an attempt - I'm sure you will feel free to amend it. Brews ohare (talk) 17:22, 6 August 2014 (UTC)

Your attempt reads very well. I have no improvements to suggest. Trevithj (talk) 19:07, 6 August 2014 (UTC)

It's OK as far as it goes. I'd add at least one ref to a more technical or EE type source, and maybe expand a bit in the lead to help the unfamiliar reader understand what this short definition is getting at, by getting more specific (not more general). Or maybe it's best to just leave it at this point. Dicklyon (talk) 06:09, 7 August 2014 (UTC)

I've added such a reference to the figure caption. Brews ohare (talk) 14:37, 7 August 2014 (UTC)

In my opinion, this Talk page discussion beginning away back here is amazing in many ways. First, it is an example of how the avoidance of sources and its replacement with editors' opinion can lead to very prolix and stupidly long discussion. It might be noted, though, that a few useful additions resulted: 1, 2, 3, which were not brought to bear upon the main issue raised about the lede. But finally, this Talk page discussion shows that despite the tendency of Talk page discussion to become a vehicle for venting, preening, and me-tooism, attention to sources can lead to a good outcome, provided a few editors are willing to go in that direction and observe WP policy despite distractions. Thanks go to them. Brews ohare (talk) 16:16, 7 August 2014 (UTC)

Is it that people are not sticking to sources, or is it you that is basically trying to use one source, which has particular, non general phraseology, and even more particularly you that is picking out that phraseology in strange ways?

For example the diagram you edit warred into the lead even though you 'didn't care', is one you made, and has the label 'input' in the middle of the feedback loop. While your favourite source does use the term input for that point, because it's the input to the amplifier A, but it also uses 'input' to refer to other points in the loop. And in the source, it's used in context whereas you are using it out of context.

For this particular use it's potentially confusing and hence bad. The original much simpler diagram doesn't have this problem, and I'd like to know immediately why you have replaced it with a worse diagram; before I remove it with extreme prejudice. For this use, where we want a fairly general picture of feedback, it's a significantly worse diagram, and the fact that you did the edit is emblematic of how you are not doing good work here.

We're not trying to summarise your one favorite reference, we're trying to summarise dozens of sources.GliderMaven (talk) 16:59, 9 August 2014 (UTC)

GliderMaven: There is no 'favorite' source involved here. About a dozen have been discussed (but none by yourself). The term 'input' can be ambiguous unless one distinguishes between 'input signal' and 'input adjusted by feedback'. Context can make it clear.

If you'd like to summarize dozens of sources, perhaps you could begin by linking them and the pertinent passages? Brews ohare (talk) 18:01, 9 August 2014 (UTC)

I agree with GliderMaven, the diagram doesn't belong in the lead or anywhere as a badly drawn mess; it's already been discussed but has somehow been made even worse since. Even if it were correct and appropriate (it isn't) it should be redone in SVG to fix the drawing errors and so it can be easily modified and reused. In fact it was a few years ago, File:Block_Diagram_for_Feedback.svg, which was based on the earlier clearer version. No-one should be using the PNG version any more.--JohnBlackburne^words_deeds 20:52, 9 August 2014 (UTC)

The diagram is unimportant. It is only illustrative of a particular case. The idea that it is incorrect is ludicrous as it mirrors similar diagrams found in the cited source. But what matters is the text. Brews ohare (talk) 03:03, 10 August 2014 (UTC)

Assessment comment

The comment(s) below were originally left at Talk:Negative feedback/Comments, and are posted here for posterity. Following several discussions in past years, these subpages are now deprecated. The comments may be irrelevant or outdated; if so, please feel free to remove this section.

Rated "high" as high school/SAT biology content. This article needs biological/biochemical examples of feedback regulation, e.g. of enzyme activity or gene expression. It also needs references. - tameeria 00:17, 19 February 2007 (UTC)

Last edited at 00:17, 19 February 2007 (UTC). Substituted at 21:46, 3 May 2016 (UTC)

[ashby-1] William Ross Ashby, "Introduction to Cybernetics" 1956, pg53. "'feedback' exists between two parts when each affects the other..."

[astrom-2] K.Johan Astrom and Richard Murray, "Feedback Systems", 2008. Page 1. "The term feedback refers to a situation in which two (or more) dynamical systems are connected together such that each system influences the other..."

[senge-3] Peter M. Senge (1990). The Fifth Discipline: The Art and Practice of the Learning Organization. New York: Doubleday. p. 424. ISBN 0-385-260-946. page75 "[Feedback is] any reciprocal flow of influence ... every influence is both cause and effect."

[wiener-4] Norbert Wiener "Cybernetics, or control and communication in the animal and the machine". Cambridge, Massachusetts: The Technology Press; New York: John Wiley & Sons, Inc., 1948. page97 "Notice that the feedback ... tends to oppose what the system is already doing, and is thus negative."

[wilts-5] Charles H.Wilts, "Principles of Feedback Control", C.I.T 1960. pg1 "...actual comparison of [the essential variable] quantity with its desired value and utilizing the difference to reduce the error observed"

[mindel-6] David A. Mindell (2002). "Between human and machine: feedback, control, and computing before cybernetics." Baltimore: Johns Hopkins University Press. "In analogy to James Watt’s flyball governor on a steam engine when the engine speeds up, the spinning balls slow it down, and when the balls spin slower, they speed up the engine. Hence the feedback is negative."

[Ashby-7] W. Ross Ashby (1957). An introduction to cybernetics (PDF). Chapman & Hall Ltd. Internet (1999). p. 53. 'feedback' exists between two parts when each affects the other...

[Astrom-8] K.Johan Astrom, Richard Murray (2010). "§1.1 What is feedback?". Feedback Systems: An introduction for scientists and engineers. Princeton University Press. p. 1. ISBN 9781400828739. The term feedback refers to a situation in which two (or more) dynamical systems are connected together such that each system influences the other and their dynamics are thus strongly coupled.

[Senge-9] Peter M. Senge (1990). The Fifth Discipline: The Art and Practice of the Learning Organization. Doubleday. pp. 75–76. ISBN 0-385-260-946. [Feedback] means any reciprocal flow of influence. In systems thinking it is an axiom that every influence is both cause and effect.

[Wiener-10] Norbert Wiener (1965). Cybernetics, or control and communication in the animal and the machine. MIT press. p. 97. Notice that the feedback tends to oppose what the system is already doing, and is thus negative.

[Wilts-11] Charles H.Wilts (1960). Principles of Feedback Control. Addison-Wesley Pub. Co. p. 1. In a simple feedback system a specific physical quantity is being controlled, and control is brought about by making an actual comparison of this quantity with its desired value and utilizing the difference to reduce the error observed. Such a system is self-correcting in the sense that any deviations from the desired performance are used to produce corrective action.

[ashby-12] W. Ross Ashby (1957). Introduction to cybernetics (PDF). Chapman & Hall Ltd.; Internet (1999). p. 54.

[1]

[2]

[3]

[4]

[5]

[6]

[R 1]

[R 2]

[R 3]

[R 4]

[R 5]

[1]