Talk:Comparison of analog and digital recording/Archive 1

Archive 1

Archive 2

Early discussion

I read some of this page and I mostly didn't know what it all means. Most of the time, whenever I come across a term that I'm not familiar with, it's highlighted in blue so I can click on it and then learn so I understand the sentence and context in which it's used. This doesn't happen in far too many places. I guess what I'm saying is that the problem with this article is that it assumed far too much knowledge on the reader's part to be effective. -bleak

After reading about the claims about Analog and digital recording, I would be interested to know about if simlar claims have been made about analogue and digital video recording.

Probably yes, but to much lesser extent, especially since analog consumer-grade video equipment has always been quite low-spec in comparison with consumer audio e.g. VHS holds the lowest rank in fidelity among analog video formats and has inferior audio even to an audio cassette], while the modernminiDV format practically beats all consumer-oriented analog formats by leaps and bounds, and none ever claimed that "Video8 film or Hi8 tapes look more natural".

All right, I'll say it: Hi8 tapes look more natural. 71.242.7.208 (talk) 05:07, 4 June 2009 (UTC)

A comparison, if any, will (or is) being done with Laserdisc versus DVD versus film, as some people don't particularly like the occasional artifacts introduced by compression in DVD movies. Also, strangely, most criticism towards digital sound is directed versus the standard audio CD, which, if properly mastered and reproduced can deliver almost 96 dB Signal to noise ratio, which is far beyond the capabilities of most audio systems (and most people's ears).

At the same time, most DVD movies have got MPEG compressed soundtracks, even multichannel ones, but that doesn't seem to bother audio purists. Also, regarding video, whil audio can be quantized at 16 or even 20 and 21 bits with some success, video signals use special Flash DACs that are 10 or 12 bits at best. EpiVictor 22:45, 15 October 2005 (UTC)

"A comparison, if any, will (or is) being done with Laserdisc versus DVD versus film" Certainly, the analog vs. digital audio debate is mosty vinyl vs. CD, not DAT vs. compact audio cassette.

Of increasing importance is also the digital vs. traditional photograp, with the later one fastly becoming what analog audio already is: a niche product. --FaZ72 21:48, 1 November 2005 (UTC)

If I may comment: When my Cable company acquired digital signals, the quality of the televised image--and I am referring to classic films broadcast on stations like Turner Classic Movies (so that I can compare image quality from the same source)--improved so dramatically that if I make a VHS recording off the broadcast, the result is a noticeably better image than what I obtain by playing a brand new commercially-purchased VHS cassette of the same film--and digitally-remastered as well. While I assume that one of the reasons for the inferior quality of the retailed cassette is the mass-production techniques, the fact is that my televised image itself improved dramatically. I can only assume that the digital image is far superior to an analog image. 66.108.4.183 22:25, 1 June 2006 (UTC) Allen Roth

That is where you are wrong, my friend. While it is reasonable from your experience for you to "only assume" a digital image is generally "far superior to an analog image" until you get more information, your single experience does not logically support your conclusion. (I could conclude from the fact that my cousin's Ford sedan has been in the shop a lot while my friend's Toyota SUV has been very reliable that SUVs are much more reliable than sedans, but I'd be wrong.) Back when I lived in New Jersey, Monmouth Cablevision upgraded their entire distibution system in my neighborhood to fiber optic. We still had analog cable (it was the mid-1990s), but it was the best cable TV picture I've seen since. A relative of mine has Dish Network, a digital DBS system, and it doesn't hold a candle to that fiber optic analog--I can see artifacts all over it, along with quantization banding in dark scenes. It all depends on the system--some digital systems are better than some analog ones, and some analog systems are better than some digital ones. Have you heard some digital cell phones? Makes you long for an old analog landline. (Seems Sprint dropped the whole clear-signal "hear a pin drop" idea.) I love Compact Disc, but I chose Hi8 (an older model, not one of these new stripped entry-level things) over DVD when I wanted a new camcorder. I don't like the artifacts either, plus DVD is such a corporate-pirate kind of system--all for us and none for you, the consumer. I'll chose what plays when in my house, when I fast-forward, when I pause, and all that, thank you. 71.242.7.208 (talk) 05:28, 4 June 2009 (UTC)

---

The language in this article and its structure and exposition could be improved, in a way that makes it clearer what the respective arguments of analog and digital are, while honoring Wikipedia's neutral POV policy. Fazalmajid 10:45, 11 February 2006 (UTC)

---

I have followed the evolution of this article over the last year or so. I would be sorry to see it deleted, because it does offer a very good overview. Some of the more extreme assertions have been removed (a good idea, in my opinion); what remains seems to present a reasonably balanced view of a rather difficult subject. Xtrapnel 11:45, 7 February 2007 (UTC)

---

The article makes the assertion that any signal below the recording noise floor generally cannot be reproduced, and compares the theoretical digital SNR (given a bit depth) against the SNRs of ideal analog recording technologies. This is somewhat misleading, because the SNR at some frequencies is often higher than the broadband SNR for either digital or analog technologies. For instance, noise shaping for 16-bit PCM can improve the noise floor to well over 100db at some frequencies, and the SNR figured of vinyl are often dominated by low-frequency rumble that is generally inaudible.

I agree that using the lack of a quantization level as an argument for analog is specious, and that even 16-bit recording with noise shaping is probably going to beat the SNR of almost all analog technologies, at any frequency. But it's kind of a cop-out to simply line up and compare the broadband SNRs.

--Rtollert 1:12, 22 March 2006 (UTC)

From my experience I notice that records sound much "warmer" than compact discs. While compact discs sometimes provide more clarity, the mid range frequencies are lost in many cases. This is especially true of older recordings made prior to the digital format. A vinyl record in good condition provides the mid range frequencies which is more pleasing to the human ear, and also a warmer "richer" sound than a compact disc. Compact discs are simply easier to take care of and you are able to store more audio on a compact disc, obviously. —The preceding unsigned comment was added by 65.151.133.223 (talk • contribs) .

If LPs seem to be providing "more midrange frequencies pleasing to the human ear", is that not simply because CDs in comparison provide more of the lower and higher frequencies (and in greater fidelity) that are inherently troublesome for the vinyl medium as we know it? With a decent stereo it shouldn't be too difficult to adjust the frequency dynamics to be closer to those of an old turntable and platter, though emulating the hiss, rumble and scratches may be more difficult. (Hell, my 79.95 digital audio player loaded up with an uncompressed WAV copy of a CD of your choice would do the trick without needing to drop a lot of money on a seperates Hi-Fi or a suitable all-in-one that hasn't gone over to the dark side of having "Flat" "Rock" and "Jazz" buttons - it has a fully adjustable 5-band equaliser plus seperate general Bass and Treble controls). This insight brought to you courtesy of noticing that a cheap stereo - or my long since dead portable turntable - with "bass" and "treble" controls can be made to accentuate or de-emphasize the midrange by turning both controls simultaneously in the opposite direction... --tahrey 21/4/07

—The preceding unsigned comment was added by 82.46.180.56 (talk) 00:52, 22 April 2007 (UTC). Many LPs from the "golden age" of classical recording (1950-1969) sound "warmer" because record companies, especially Columbia, artificially warmed the sounds of the orchestras, in order to produce a "warmer" sound for the average listener. RCA and other companies doctored the sound much less. Consequently, orchestral music on Columbia was less "grating" to the ears. A real violin or trumpet has the characteristic high "top" that is harsh to the ear used to hearing reproduced recorded sound, and Columbia wanted a less "harsh" sound, at the expense of fidelity. As far as the issue of the reproduction of mid-range frequencies is concerned, I think you are referring to what conductors call the "texture" of the recording, i.e. the ability to discern in a recording of a group of different instrumentalists all the lines of the individual ensembles. In my opinion, one of the biggest problems with digital recording is this. When listening to an orchestral recording, the same recording on LP (analog-recorded) will display more clarity in the individual members of the orchestra, and their differing timbres and lines. In the CD, I sometimes perceive only the overrall sound, and find that the instruments carrying a counter-melody simply cannot be clearly separated from the sound of the complete recording, i.e. it sounds like a "recording," not like 80 members of an orchestra. 66.108.4.183 11:45, 26 May 2006 (UTC) Allen Roth

Undigitized sound available?

QUESTION: This article states that it is impossible to find recordings which have not been subject to digital techniques at some point. If I play one of my LPs that was manufactured before digital techniques existed, on my turntable, and use an amplifier also manufactured before c. 1980, am I not hearing analogue sound, with no digitization at any point? A reply is welcomed by anyone who has more technical expertise than I. Thank you. 66.108.4.183 11:45, 26 May 2006 (UTC) Allen Roth

Clearly, an LP produced during the analog age and reproduced on analog equipment will produce sound that has never been digitized. I suspect that the person who made the statement was referring to music produced today (or perhaps "within the last 15-20 years"?). Such music was almost certainly digitized when the original master recording was made, and was probably resampled a number of times before finally reaching you on whatever medium it did.

Atlant 12:11, 26 May 2006 (UTC)

If I may comment: When my Cable company acquired digital signals, the quality of the televised image--and I am referring to classic films broadcast on stations like Turner Classic Movies (so that I can compare image quality from the same source)--improved so dramatically that if I make a VHS recording off the broadcast, the result is a noticeably better image than what I obtain by playing a brand new commercially-purchased VHS cassette of the same film--and digitally-remastered as well. While I assume that one of the reasons for the inferior quality of the retailed cassette is the mass-production techniques, the fact is that my televised image itself improved dramatically. I can only assume that the digital image is far superior to an analog image. 66.108.4.183 22:24, 1 June 2006 (UTC) Allen Roth

The video and audio quality of a VHS recording cannot exceed the limitations of the medium itself, in terms of bandwidth and frequency response (which for analog video means line and color resolution), so even the best VHS recording cannot compare to an average quality analog TV broadcasting, at least in PAL.

What can happen however, is that some newer TVs (and VCR's) have some picture-enhancing technology built in e.g. "Super Picture" or something along these lines, while many new TV's have a "definition" adjustment, which generate an apparently crisper or more softened image (e.g. I have a 6-head Sharp VCR with a "Super Picture" key, which apparently increases horizontal resolution somehow and makes higher frequencies (and thus more pictures details) more visible, and many of my older tapes look much better on this VCR.

Also, using modern equipment combos (namely, recently built VCRs combined with recently built set-top boxes and recent TVs) will yield better results than using older VCRs, sometimes even compared with factory-made tapes. E.g. it's not improbable at all that the digital receiver set-top box has a better output quality (even over a normal composite output) than what was available even at factory level during e.g. the 80s or the mid 90s, so that even home-recorded VHS tapes may look better.

This is also the case for audio cassettes recorded from CD's or MP3, on most modern (even low-cost) mini or micro-hifi systems: even without Dolby enchancements, they will generally sound as good or even better than some older factory made tapes, partly because the source material will be better in most cases.

In fact, I have so many original tapes that suffer from an obviously bad mass production/dubbing job, and recording them again from a broadcast would with no doubt improve them....it's all very relative. You should try those "enchanced" VHS tapes with other TV/VCR combos, to be sure that they work OK, and try using different VCRs for recording the "digital" signal. :-) EpiVictor 23:31, 1 June 2006 (UTC)

It may depend merely on the particular circumstances of the recordings in question (...which probably summarises all the other comments! :) - I have had experiences both ways up on this. I have many what I consider high-fidelity recordings from broadcast television made in LP mode on both a mid-90s 6-head VCR and a ~2000 4-head that are better than certain pre-recorded tapes I own... and some very good commercial cassettes into the bargain, dating from the mid 80s to the commercial death of VHS a few years ago. Similarly various experiments in recording digital video to tape in LP and SP, using a cheap composite converter card installed in my PC, which was surprisingly sensitive to the quality of the original files (less in terms of resolution, more in terms of general brightness/colour/compression/sound). I'd take almost any of them over probably half (if not more) of the digital signals currently available on my TV via both free-to-air DVB and our local fibreoptic cable service... the resolution is good, sure, and it's fairly resistant to interference - but the basic quality of this 'uncorrupted' signal is poor enough that I wouldn't mind trading a little bit of noise to get rid of the awful blocking, crinkling and smearing resulting from greedy over-(digital)compression. This holds true even for some DVDs I have - whether it's the mastering, or the content (somewhat predictably it's more noticable on animated material), or trying too hard to stuff too much simultaneously-playable content on one disc (various sound tracks, multiple angles, etc), there are several points where I end up wondering if the core single video track and own-language stereo soundtrack might have looked and sounded "better" on simple videotape, with straight uncompressed recording and no extra features to steal bandwidth. Shame really - digital switching can be done right with a bit of care. There's no way I'd re-exchange vinyl for CD, or cassette for mp3 (having recently unearthed caches of both in a tidy-up and revisited a world of truly terrible (comparitively) recording quality). To bring the two of these together, I also have a small number of commercially recorded cassettes, and they generally match anything I've managed to create on my own. Only one or two are of questionable quality (and can themselves be slightly improved by tweaking the playback head tracking position, suggesting it's a minor manufacturing fault), by which I mean having similar dynamic range to a 112 or 128k MP3 instead of 192k, rather than any massive decrease to 96k equivalency or below (as apparent on some homemade tapes gained through swaps with friends). Tape seemed already to be as good as it could sensibly get by the time CD became popular... -- tahrey 21/4/07

—The preceding unsigned comment was added by 82.46.180.56 (talk) 00:07, 22 April 2007 (UTC).

Let's start cleaning up...

OK, where to start from? Oh here we go, we can start from the "arguments in favour" section:

Arguments made in favor of analog sound

* Shape of the waveforms: sound reconstructed from digital signals is claimed to be harsher and    unnatural compared to analog signals

OK now this one is ridicolous, as it's not specified for what class of devices this holds true or false, and from a signal processing point view it has no sense. So that means that a 10$ tape recorder reconstructs "less harsh" waveforms than digital systems? The only occurence of "harsh waveforms" would be when operating near the maximum samplable frequency (not sampling) for digital systems, and even an analog system would sound "harsh" if operated near or beyond its limit.

* Lower distortion for low signal levels

This contrasts with the "lower noise floor" for digital systems. Now, either a syetem has "X" dB of SNR or it doesnt. Whether a weaker signal would mean "sinking into the noise soup" or below "bit 1" doesn't matter, it wouldn't be recorded correctly in either case.

* Absence of quantization noise

This actually may have some sense, but is closely related to the SNR ratio and the quantization bits of the equivalent digital system. It's clear: 6 dB per bits, and 96 dB is already way beyond Hi-Fi specs ( 80 or so dB). So unless you're comparing reference vinyl with 8-bit Soundblaster samples or otherwise flawed digital recordings, and if it's true that the human ear cannot even discern between 14 and 16-bit quantization, then this is yet anoter myth to debunk. Sure, analog systems don't work in discrete steps, but they don't have infinite precision either.

* Absence of aliasing

Actually, recording an unfiltered signal that contains frequencies beyond the acceptable range with an analog or digital system results in something being recording, the fact that it's analog doesn't however magically filter out the offending frequencies, and even an analog system wouldn't yield a correct recording (try recording e.g. a high-pitched whistle with a microcassette recorder: it won't just "low pass filter" the very high frequencies beyond the tape's response.

* Not subject to jitter

That may have some validity, but requires to specify and explain which kind of "jitter" we're talking about, in which digital systems it occurs and how it can affect audio. Analog systems have their own problems (wow & flutter, repeated analog amplifying and filtering stages etc.)

* Euphonic characteristics

This one definitively needs expanding.

Arguments made in favor of digital sound
* Lower noise floor
* Dynamic range
* Signal to noise ratio

These ones are all directly related to the quantization level (and also to the recording equipment and pre/post processing of course).

* Absence of generation loss

This one needs explanation.

* Resistance to media deterioration

That's a big pro of most digital systems, altough there are analog systems which can be as "tough" (e.g. LaserDisc, FM-encoded audio on Hi-Fi VHS tapes etc.)

* Immunity to wow and flutter
* Ability to apply redundancy like error-correcting codes, to prevent data loss

There should also be a mention on how cheap and relativaly easy it is to build literally millions of identical, similarly-performing digital systems , either they are Hi-End SACD players or portable MP3 players, and how easy it is to grant to even the lowliest CD/Radio combo an enviable SNR ratio comparable to e.g. a Cassette/Radio combo. EpiVictor 10:12, 5 June 2006 (UTC)

Very poor article

Apart from the quantization sections and the included math, most of the article has ridicolous or easily rebuttable arguments. Strangelu, non replied to my previous list of obesrvations, so I will write a new one about the current "pros and cons" of both systems.

Advantages of analog sound

• Analog recording is a linear representation of a linear waveform, and therefore more accurate.

First of all, the "linearity" is a function of frequency, and only perfectly tuned recording instruments and media achieve this supposed "linearity". Actually that's something very hard to achieve AT RECORDING time, and even harder to preserve through analog processing. ANd, BTW, a digital system is actually able to store waveforms an analog system cannot, e.g. perfectly square waveforms. Reproducing them is another matter, though.

• Shape of the waveforms: analog sound appears "warmer", "smoother" more "three dimensional"

As long as something has been sampled correctly,includes all necessary information, and is played through the right equipment, even digital sound can sound "warm" and "three dimensional".

• Lower distortion for low signal levels

Nonsense, unless you can guarantee a sensibility of less than -96dB through recording, processing and mastering phases for either analog or digital media.

• Absence of quantization noise

...and presence of a lot of other kinds of noises. Plus, there IS a lower sensibility bound even for analog systems. Sure, it's not "step like" but it's equally unusable for recording.

• Absence of aliasing
• Not subject to jitter

These are the only ones that make sense.

• Euphonic characteristics

You've got to be kidding.

Disadvantages of analog sound

• Linear access

So what about vinyl records or laserdiscs? Are those linear access too? There are even remote-controlled servo assisted turntables, and even by hand you can skip to any track you want!

• Subject to electrical and mechanical hiss and noise
• Subject to wow and flutter

Those are true...

• Tape is expensive to buy and maintain

What kind of tape? Metal cassettes? Open reels? Who is using the latter anymore?

• Regenerations are inferior quality

True, but factory produced records or tape usually come from a digital or at least a high quality analog master and are first-gen copies. Assuming we're not talking about bootlegging, audio editing or home dubbing, this shouldn't even be an issue.

Advantages of digital sound

• Non-linear access

DATs and DCCs use TAPE, aren't those linear (non-random) access???

• Lower noise floor

This contrasts openly with the "lower distortion at low signal levels" analog claim.

• Regenerations are exact clones

..unless DRM schemes are used. An BTW, it's possible to get non-perfect digital copies or at least make them sound imperfect on real reproduction devices.

• Resistance to media deterioration

That's true for optical media, but there's also magnetic digital media, that suffers from deterioration as much as analog.

• Ability to apply redundancy like error-correcting codes, to prevent data loss
• Data channels allow digitally encoded information about the owner, track titles, and other information

True...

Anyway, all in all, this article needs serious attention. EpiVictor 11:39, 4 September 2006 (UTC)

Revamping the article

I started a timid attempt at revamping this article, by massively rewriting certain sections. However, we still need some expert's help to expand it and clean it up. EpiVictor 23:16, 4 September 2006 (UTC)

The problem is that many people who actually understand the technical issues consider many of the claims made by the analog sound proponents to simply be pseudoscience, not worth the effort to rebutt.

Atlant 23:43, 4 September 2006 (UTC)

Cmon, it can't be that bad...at least the purely technical claims could be left in and expanded. However I'd personally avoid referring to "perceived warmness" or "natural sound", and just use signal theory verifiable facts. EpiVictor 11:33, 5 September 2006 (UTC)

"Jitter" versus "wow and flutter"

Someone just removed the following "advantage" of analog sound:

• Not subject to jitter

saying"

Wrong, analog motors are not perfect either

I understand the point that this editor was making, but I think we should distinguish between jitter and wow and flutter.

Jitter, as the term is commonly understood, is purely a disease of digital systems and means, essentially, uncertainty in the exact timing of individual samples compared to a theoretically perfect sampling clock. Wow and flutter, on the other hand, are (pretty much) purely diseases of analog systems and mean long- and short-term variations of the sampling rate, but not really extremely short-term sample-to-sample timing uncertainties.

Analog systems don't jitter because 1) the data are not discretely sampled and 2) mechanical inertia makes it unlikely that you'll see extremely short-term variations in timing. Digital systems don't wow or flutter because the timing is almost always crystal-controlled; it may be wrong, but it's steady.

We might want to tweak the article text to make all of this more clear.

Atlant 19:34, 28 September 2006 (UTC)

That would still miss the fact that an analog (continous amplitude) system can have a non-continuous, discrete time base and viceversa, a digital system (discrete amplitudes) can have a continuous time base. Agreed, there probably aren't many Hi-Fi PWM audio systems around, but exactly like it's possible for such an analog system to suffer from aliasing, it can also suffer from jitter. Jitter is something affecting timed pulses, and unlike "common sense", sommething using pulses isn't necessarily "digital". EpiVictor 13:57, 30 September 2006 (UTC)

Isn't jitter, when you get down to it, just very high-speed flutter? (which is itself high-speed wow)... it's noted as being undetectable below a certain time constant, probably because like with clicks becoming a tone, and flickering pictures becoming a moving image, the speed of it and the increasingly minor effect it has eventually blurs into the background of the actual content. Just like with analogue wow, flutter, and background noise, so long as it's minor enough to be below the level of general perception, what does it matter? Really, if a CD transport and circuitry is good enough at delivering the bits from disc surface to output device (speaker or microchips) with a time accuracy good enough that there isn't a detectable bit out of place amongst 1.41 *million* of them being delivered each *second* (or 1.20 million for data at single speed - and i'm basing my theory on a CDROM drive able to happily read at 48x that speed without error) then I can't really see how my ears are going to tell the difference. The largest possible difference will be the MSB for one channel being inverted for one sample, thus producing a 22khz pop... barely perceptable and sounding similar to a minor click on an LP... and that would already require an inconsistency within a few percent of being bad enough to corrupt CDROM computer data. Yes, this was actually an issue with early CD ROMs, particularly when asked to rip audio data (hence now-redundant "jitter correction" options), but I'd be rather disappointed to find a modern-day transport suffering from it, at least one where I was paying with more than one banknote that didn't make the cashier quietly press the silent button to call his manager over to double-check. Don't think I've ever owned a drive that could do audio rips that was inaccurate enough to need it. Even if consumer players and portable devices be using really awful quality drives that might suffer jitter effects in live playback, the majority now come with anti-skip buffers; a Discman without at least a 10-second anti shock is a sign of a real bargain basement item (even my mother's 10-year-old device that boasts a 3-hour battery life has a 3-second memory..)... to complain about jitter in these would be to insinuate the actual solid-state digital playback hardware has significant timebase errors, and are you really so sure of your analogue superiority defence to make a claim such as that? --tahrey 21/4/07

The "jitter correction" in CDDA ripping software is generally not to correct bit timing jitter (that doesn't matter, because the data isn't being D/A converted in real time) but to correct "framing jitter", the random shifting of the data block frame relative to the actual frame boundaries due to the imprecision of the CD-ROM drive's ability to seek for a random indexed access. Because most older CD-ROM drives used CD player chips which separated the subcode and audio data into two separate, unsynchronized streams, it was impossible for those drives to do sample-accurate seeks within the audio data. So if you asked for the same block of n CD audio frames twice, you'd get the right number of samples, but the first time the block mishg start four samples too early and the next time it would start five samples--or 50, or 250--samples too late. The solution, called "jitter correction" was for the software to read overlapping blocks and then piece them together, kind of like the way DNA segments were assembled fir the human genome project. This has nothing to do with clock oscillator stability, or power supply ripple, or thermal noise, or anything like that--it's about trying to do something that requires precise synchronization of two separate data pipelines that were never designed to support that kind of operation, and compensating for that design limitation. And there were, and still are, a lot of those drives in service, so even if you never have one, or have the luxury to just go out and buy a better drive whenever you feel like it, there are people for whom those drives are all they have, I'm sure still. So [frame] jitter correction was and is not a useless or redundant feature of CD audio extraction software. 71.242.7.208 (talk) 05:58, 4 June 2009 (UTC)

Cleaning up

I did a fair amount of cleanup/correction on this article. It's shaping up nicely, although I'm not too thrilled that it ended up on the Answers.com website before it has been fully worked over. I have forwarded this page along to my mentor at the University of Miami, Ken Pohlmann, who wrote "Principles of Digital Audio" and by far knows the most about audio amongst people I know. I will say this article needed (and still needs) a lot of grammatical correction to be brought up to encyclopedia standards. I will likely work on it some more when I have some more time.

Interesting edits...however signing your comments would be nice :-) (use 4 tildes ~ in a row to do that), and if you really have access to such a qualified individual's consulence, that will surely benefit the article. As an electronic engineer, I tried to tidy it up myself several times in the past, but the argument itself is hard to chow down for some reason...and filled with ambiguities. From my point of view, it would be a non-problem, since digital systems are used EVERYWHERE by now, and nobody said e.g. that an analog industrial control system is somehow better or more precise than a digital one, or that VHS tape has "truer" image than DV tape....it's just something that applies to certain kinds of audio systems (usually, by analog audiophiles mean only vinyl records and open-reel tape, and by "digital" they only compare with standard, preferably early or erroneously mastered CDs) EpiVictor 14:15, 5 October 2006 (UTC)

Sorry, I am new to this whole Wikipedia editing thing. I would agree that this argument is pretty much moot nowadays. I remember when I first got into the business (not even that long ago) there were plenty of valid arguments for analog, but digital systems have come a long way since then. In any case, if this article is going to be up, it needs to be correct.pianomanum 01:14, 7 October 2006 (UTC)(AM)

One note, albiet not small, about vinyl vs. CD

Ok, this is something i'm going to touch VERY briefly on becuase i'm pretty sure it's never been mentioned. Yes, vinyl has a wamer sound...and there have been people who have said it has more dynamic range...which is entirely false. Compact discs at 16 bit do have a wider dynamic range than a vinyl LP...vinyl LP's do have a fair bit of naturally occuring surface noise which decreases the amount of range you can have.

So why claim they sound better? A lot of it has to do with the production/mastering done with CD's. In the early digital era, there was no real post-processing in the digital domain, analog mastered content was just about transferred directly from tape to disc, giving them a dynamic range closer to the original; leaving the disc overall quieter to the ear, not to mention early A/D converters weren't as efficient as what you have now.

Back in the vinyl days, music had to be mastered a specific way. you had limitations you had to follow in order to make sure the groove was laid out in such a way that it was trackable, if you had too much dynamic compression, the needle won't track properly. I can't remember the exact reasoning for it, but it relates to physics. Most music today is made on a "louder is king" ideal and in comes the dynamic compression. It's this over compression that makes some people say CD's have an inferior sound to vinyl, when in reality, they're just about the same, it just depends on the mastering. However, this doesn't mean you don't benefit from using high sampling rates and resolution in transferring vinyl LP's into the computer, afterall, you're dealing with an analog source and every bit of detail you can catpure will help. Some CD's sound quite smashing; Steve Hoffman DCC-label GoldDiscs and most anything in the Mobile Fidelity catalog was made to take full advantage of the CD format. Guys like Hoffman are the only ones doing thier mastering jobs like they're supposed to...making sure the audio sounds top-notch on the media and not what some record exec wants to sell. The industry is the one killing the music. --DewDude 21 Nov 2006

A lot of it also has to do with people who don't really care about music anyway, and most people don't, despite what they claim. The delivery system makes absolutely zero difference. Music is in your head, not in your ears. Worrying about whether it's on vinyl or CD is like worrying about what color your car is. ---cneron

Analog warmth

I added this small bit because it's something that is mentioned quite often - analog warmth. It's a very short entry, but so far it's the only source I've found. Enescot 20:53, 10 January 2007 (UTC)

Maybe we should try and find an answer to this question: having higher/absolute fidelity equates to better sound, for audiophiles? Or is some form of distortion desirable (perceived "warmth")?

A digital system IS able to perfectly recreate an analog system, if it's specified to handle it (frequency, dynamic range etc). But is that really desirable by audiophiles? E.g. *I* have an old Grundig transistor radio/cassette recorder that has a very distinctive sound and *I'd like* to listen to everything through it, but does that make it objectively "better" than a CD? EpiVictor 21:13, 10 January 2007 (UTC)

Stereophile I remember has had some musings on this, and I remember someone once saying how in a proper test it has been shown that people like a treble roll-off as you have with many vinyl records. Enescot 20:51, 16 January 2007 (UTC)

I've added some more stuff on this using two articles from Omni magazine. I'd prefer using a source like Stereophile because it's still in existence, but the Omni articles have some good quotes.Enescot 12:04, 8 March 2007 (UTC)

Is this whole article/argument moot?

Correct me if I'm wrong, but most "audiophiles" seem to claim that VINYL (or analog tape) sounds better to them, not that they are technically superior, provide better frequency response or superior SNR ratio (which, compared even to Audio CDs, they actually don't, something which is scientifically provable beyond doubt).

E.g. when recording a precisely generated signal with specific frequencies and amplitudes, the point in the whole "analog vs digital" dispute doesn't seem to be "which medium records it more faithfully", but "which medium subjectively sounds better to one's ears", which means the article should either be deleted or written in a way that emphasizes that what is being discussed is essentially a subjective perception, and that there's no doubt, scientifically wise, that both an analog and a digital system can reproduce frequencies/amplitudes they were designed for.

If, on the other hand, CURRENT digital or analogue media have inherent technical limitations or suffer from gross errors during production, that's another point. If the "vinyl masters" of the "golden days" mastered vinyl records using tape reel speeds of 3 m/s during recording/editing, or if some noncurant CD "masterer" produces his CDs starting from MP3 compressed audio, that's another story altogether, not a property of the media themselves. EpiVictor 11:57, 7 February 2007 (UTC)

I see what you're saying, but the theoretical performance of digital systems is practically limited by the quality of the converters. Therefore, a low-quality digital system can be worse objectively and subjectively than a high-quality analog system, e.g. a portable CD player versus a good quality record player (and of course vice-versa). With my edits, I've tried to concentrate on presenting information that explains how this is possible.

I see how this article does depart in style from other articles in Wikipedia on hi-fi, in that they often tend to focus on the subjective differences. The difficulty here, though, is that it is difficult to find high-quality source material on which to base these comparisons.Enescot 16:23, 9 February 2007 (UTC)

Actually, MANY audiophiles claim that vinyl is better, technically. I have heard and read countless claims about digital's alleged inability to accurately record and reproduce a waveform. Very frequently, such arguments are from a position of ignorance, and have little or no technical merit. As long as audiophiles claim that they like vinyl better than CD, it's an opinion. Not much point in a dispute about opinion. When they start to make specious technical claims, that's when facts can be used to counter the ridiculous claims about imaginary defects in the digital world. Of course, there are REAL defects in digital recording, but there are many audiophile claims that cite imaginary ones 139.68.134.1 (talk) 19:27, 21 August 2008 (UTC)

Exact reproduction/error correction

Moved from User talk:Omegatron — Omegatron 02:39, 28 February 2007 (UTC)

---

Hello

I had a look at your revision of analog sound vs. digital sound as of 25 Feb 07, and think that the removed sentence was a sensible introduction to the topic of digital sound reproduction. I based this specific sentence on a part of a book referred to in the article - Driscoll, R. (1980). Practical Hi-Fi Sound, Hamlyn:

'...The signal begins and ends in analogue form, as it must for our ears to recognise it, but in the tape recorder it is converted into digital form (analogue to digital conversion), in fact into a binary code of discrete numbers or digits. Hence the signal which is recorded onto tape is not continuous, as an analogue signal, but takes the form of particular, discrete values. In this form, the signal is far less susceptible to noise, distortion and speed variations occuring in the tape recorder. In fact a digital recording system is more immune from these problems because it does not attempt to preserve the exact, continuous form of an audio signal.

It is necessary, of course, for the digital system to carry sufficient signal information so that after it is converted back into analogue form (digital to analogue conversion) the result is indistinguishable from the original sound, but the limitations lie only in the electronics of the system, not in the tape medium or in the mechanics of the tape recorder.'

- page 63

A different description of this is provided in Chris Dunn's and Mark Sandler's paper, 'A Comparison of Dithered and Chaotic Sigma-Delta Modulators', page 1:

'A perfect analogue-to-digital (AD) or digital-to-analogue (DA) conversion process would convert a signal from one domain to the other without introducing either linear errors (that is, magnitude and phase-response abberations within a specified passband) or nonlinear errors (noise and distortion). While linear errors in modern converters can be made vanishingly small, quantisation theory tells us that any conversion process must contribute some degree of nonlinear error, hence a fundamental issue which must be addressed when assessing the quality of a conversion process is deciding what represents an acceptable degree of nonlinearity'

- http://www.scalatech.co.uk/papers/jaes496.pdf

The revision you made to the part about error correction, I am slightly unclear about. In 'An Introduction to Digital Audio', Focal Press, John Watkinson describes a channel in quite precise terms. An example of a channel would be the S/P-DIF interface. Error correction is, however, used when reading data from Compact Disc's, and takes the form of, I believe, a decoding from Reed-Solomon error correction coding. I don't think that this reading of data is usually thought of as a process requiring the use of a channel. This makes me feel that the original sentence - 'Error correction coding, essential in digital audio systems, helps to eliminate bit errors', is more general than the edited one - 'Error correction coding, essential when transferring digital audio over noisy channels, helps to eliminate bit errors.'

When thinking specifically of the common consumer S/P-DIF interface, it is true that error detection coding is used for the channel status information, which tells the decoder whether or not the audio data has been corrupted. The audio data however, is not coded with Reed-Solomon error correction, and corrupted audio data can only be muted.Enescot 21:25, 27 February 2007 (UTC)

because it does not attempt to preserve the exact, continuous form of an audio signal

Kind of. But when people say "digital is an approximation of analog" or whatever, it's usually because they don't really understand how digital works. We shouldn't allow the article to encourage them in this.

• Sampling - People often misunderstand and think that, because the ADC is only measuring the continuous wave at discrete points in time and missing everything in between, digital is just an "approximation" of analog. Of course the Nyquist–Shannon sampling theorem proves that all of the content below half the sampling frequency is reproduced exactly (with ideal filters). With real filters, as long as your flat region is significantly above (and below) the limit of audibility (beyond which you can't call it "sound" anymore), you've reproduced all of the sound exactly.
• Quantization noise - All audio already has a noise floor, due to the limitations of thermal noise in analog electronics or mic elements (or the air itself if the electronics were perfect?) As long as your quantization noise is well below this noise floor, you've reproduced the signal exactly.

If I digitize an old record with a -60 dB noise floor and bandwidth of 16 kHz by putting it through my 24-bit/96 kHz sound card (realistically around 20 kHz bandwidth and 110 dB dynamic range), I'd have reproduced the signal exactly. With no loss of information at all. In fact, at 24/96 I'd have wasted a significant amount of hard drive space by storing more information than is actually present on the record.

error correction

I dunno. I'd consider a CD to be a "noisy channel", but if you want to change the wording, go ahead. Error correction codes aren't necessary and aren't used if you know there's not going to be any signal degradation, though. — Omegatron 02:58, 28 February 2007 (UTC)

With real filters, as long as your flat region is significantly above (and below) the limit of audibility (beyond which you can't call it "sound" anymore), you've reproduced all of the sound exactly.

As long as your quantization noise is well below this noise floor, you've reproduced the signal exactly.

If I digitize an old record with a -60 dB noise floor and bandwidth of 16 kHz by putting it through my 24-bit/96 kHz sound card (realistically around 20 kHz bandwidth and 110 dB dynamic range), I'd have reproduced the signal exactly. With no loss of information at all.

I see that from a practical standpoint, you're right, but as the Dunn and Sandler paper points out, you are still adding something to the original analog signal/information, so it is therefore correct to say that digitization is an approximation. I'm afraid that your perfectly logical position about reproducing a signal exactly is likely to be true from the point of view of audibility, but audibility is a contentious area. The key point, I feel, is that digital equipment can deviate a long way from the ideal, which can produce audible differences between equipment, e.g. a high-quality separate CD player and a portable CD player.

Also, the specification of digital audio equipment is, in practice, fairly complex from an absolute standpoint. Measures such as dynamic range, bandwidth etc. do not fully define the objective performance of the system, i.e. 'Nonlinear dynamics of bandpass sigma-delta modulation', Feely and Fitzgerald:

'...Despite the popularity of these converters, understanding of the operation of Σ∆ systems is far from complete. This is due to the nonlinear nature of the modulation process. The response of most engineers confronted with a nonlinear system is to approximate it by a linear system to which standard linear analysis techniques can be applied. This “linearise, then analyse” approach pervades much of the scientific literature on Σ∆ modulation. Many of the results thus obtained are incorrect not only quantitatively, but also qualitatively.'

- http://www.eurasip.org/content/Eusipco/1996/paper/fi_4.pdf

If you then approach the audibility issue subjectively, and look to the general hi-fi press for information, you're mostly into a realm dominated by listening tests done without adequate experimental controls, and that is an debate I wish to avoid getting involved in!

Noisy channels

Sorry, it's my ignorance of Information Theory that lead to this criticism. It's just that the word 'channel' is used for so many different things in audio, so I was a bit confused by your edit here. Perhaps it would help to put in an internal Wikilink on the words 'noisy channels' which leads to the article on 'Noisy channel coding theorem'? Maybe a brief example would help as well i.e.

'Error correction coding, essential when transferring digital audio over noisy channels, e.g. reading data off a Compact disc, helps to eliminate bit errors.' Enescot 18:34, 3 March 2007 (UTC)

I'm afraid that your perfectly logical position about reproducing a signal exactly is likely to be true from the point of view of audibility, but audibility is a contentious area.

No. It's not about audibility; it's about reproducing the signal exactly.

The key point, I feel, is that digital equipment can deviate a long way from the ideal, which can produce audible differences between equipment, e.g. a high-quality separate CD player and a portable CD player.

Of course. Ultimately, analog vs digital vs tubes vs transistors vs whatever is completely irrelevant; what matters is the quality of the design. Any system can be just as good as any other, regardless of what it happens to be made out of.

If you then approach the audibility issue subjectively, and look to the general hi-fi press for information, you're mostly into a realm dominated by listening tests done without adequate experimental controls, and that is an debate I wish to avoid getting involved in!

Good.  :-)

You can replace the word "channel" if you want. Something about storage media? — Omegatron 18:54, 3 March 2007 (UTC)

More thread arguing nonsense

Thought i'd chuck in yet more rambling thoughts connected with what other people have hashed out above. Just a few things struck me as I was reading through. Haven't figured out how to properly "quote" yet, so bear with it if the structure is a bit unreadable.

Shape of the waveforms: analog sound appears "warmer", "smoother" more "three dimensional"

What of surround sound recordings on e.g. DVD-A or SACD? Aren't they "3 dimensional" enough? :-) :-) (or Dolby Pro-Logic II encoded CDs, for that matter)

Linear access

So what about vinyl records or laserdiscs? Are those linear access too? There are even remote-controlled servo assisted turntables, and even by hand you can skip to any track you want!

Good luck switching tracks with any degree of accuracy or automation, however, unless your servo-controlled thing is VERY good (how does it know where to drop the needle?). Doing it successfully by hand, if you're not a professional DJ, requires sniper-grade muscle control and probably some diazepam... i've never been able to do it very well, anyhow. Laserdisc is a stickier point which I'll decline to comment on as I don't know much about it, beyond it being an odd analogue/digital hybrid precursor to CD. (The advantage of optical/digital pickups in this regard is that you don't risk media or needle damage if you muck it up, and it is automatically muted until properly cued without you needing to be a dab hand with the volume control/mixing desk and a pair of monitoring headphones)

Tape is expensive to buy and maintain

What kind of tape? Metal cassettes? Open reels? Who is using the latter anymore?

A few dedicated souls... but even bog standard compact cassette isn't too cheap any more. Some years ago when I was making vast quantities of home recorded cassettes for my Walkman, to play in the car, recording interesting radio programmes, etc, and just experimenting with my first CDR drive, the comparitive costs were about £0.20 to £0.25 per hour for standard Fe cassettes (TDK D for preference), £0.75 to £1.20 for CDR (awful generic toss to reliable major brands, and significantly more for re-usable CDRW)... and even high quality discs had low maximum write speeds and poor durability once written, assuming your burner didn't suffer a buffer underrun or other fatal error. Copying a tape or backing up one that was damaged to a fresh blank could be commonly done at 2x speed on a home hi-fi, and it was fairly bomb proof so long as you could get it past the knackered parts somehow; CD duplication was generally done direct between two drives at the lowest speed you could stand (for safety's sake, in case any unneccessary stress interfered with the process), and you were lucky to have enough spare hard disc space to do a "rescue" copy of a damaged disc that demanded read speeds below what the writer could handle, as even a minor defect could lead to it endlessly re-reading a patch.

Nowadays it's drastically reversed - CDR is about £0.15 per hour, CDRW not much more, and the few tapes to be found on retail shelves are in the region of £0.66 per hour (for the slightly lower fidelity TDK FE), with the gap slowly but steadily increasing. Even advanced technology outstrips it in terms of value - an £80, 20gb MP3 player offers £0.22ph (assuming you record at the roughly FE-equivalent 128kbit rate), and is only slightly more expensive than yesteryear cassettes with higher, more transparent VBR rates. CDRs are much more reliable (benefitting from the advances that have given us DVD-R) and the discs and drives are a lot faster, more fault tolerant, and big hard drives give us all the scratch space necessary to fix up a damaged one and write a fresh copy. Recently I've had to make backups of a badly scratched Windows original CD, and try to repair a degraded cassette that wore through and snapped. The tape was far more of a nightmare and doesn't actually work 100% - most of it now plays, but the first few seconds of side B is inaccessible as I had no splicing materials available, only regular adhesive tape, which is thin enough to go around the spool but not to pass through the capstan. Though it cannot be done-over now I've kludged it, I don't expect the necessary kit to do it professionally comes cheap. In contrast, despite a superficially similar amount of catastrophic damage, the CD backup went basically without a hitch, once I changed the read settings to "only" 8x speed and let it chew over the multiple unreliable sectors for an hour or so... Plus, of course, doubled tape decks have fallen in popularity on all but the most exlusive (or the most tacky) home hi-fis, and you'll be lucky to see 2x dub any more, except on the few that offer it as a quicker way to get stuff from CD to tape.

Non-linear access

DATs and DCCs use TAPE, aren't those linear (non-random) access???

Actually the only properly non-linear storage used in audio recording is RAM (including flash memory), hard disc, and maybe minidisc (I don't know the specifics of it's surface layout, except that it's magnetic-optical, which suggests some extra jiggery pokery beyond it being a miniaturised CD). Everything else is linear, it just depends on the access method and speed. CDs use a single spiralling track analogous to Vinyl's groove, it just happens their "needle" is able to scan across and through this groove a lot quicker and more accurately than is possible with other media owing to the non-audio information stored at the start of the disc and throughout the track itself. It isn't actually able to automatically home in on a point and play straight away, as a hard disc or minidisc head may quickly skip to a known track position, or a memory chip instantly engage a different logical address... it has to take the position specified by the TOC, and read the addresses live from each segment as it passes over to keep track of roughly how far it's gone, stopping & optionally backtracking once in the right area (theoretically within 1/4 second if it lands on the right lateral bit but displaced angularly) - the inaccuracy of this in early players is part of the reason for a 2 second silence (preferably, a complete lack of data) between tracks being standard, so the kludginess of the emulated needle-jump is hidden from the end user. Nowadays, properly made ones will step back slightly after finding themselves in the track-switch zone, and wait for the disc to reach the correct point in it's spin to start play from the exact 1/75th second point defined in the TOC.

Digital cassettes have similar tracking info, but as their read mechanism is, if you like, permanently engaged (unlike CD and vinyl that can 'pick up' the read head and scoot it quickly back and forth throughout the media), their seek speed is governed by how fast the tape can be spooled through the system. There's nothing except the laws of physics and the skill of the mechanical engineers involved to prevent this being effectively as fast as CD... Of course, it's as accurate intriniscally, and similar to the index-skip or music-search function of conventional VCRs and modern tape players, as it instantly detects when the correct point is reached (and can pre-empt) and rapidly stop, wind back, and play at a perfectly queued point.

Resistance to media deterioration

That's true for optical media, but there's also magnetic digital media, that suffers from deterioration as much as analog.

Optical media isn't perfect either - it can be damaged by overexposure to sunlight, or simply corrode of it's own accord, and this holds for both factory pressed and home burnt discs. I have several pristine-looking CDRs that, however, are very difficult to read now, and show massive errors when run through a detailed test program, and have had to be carefully backed up... whilst tapes of the same era are still going strong, save for a little print-through from being neglected unplayed in their cases for several years. Not to mention the commercial discs suffering "laser rot" (with the silvered material starting to flake away at the edges, and moving inwards to the data area)... though this isn't as common now as it used to be (cf. "The Adam and Joe show", s1e1, 1996, as recently re-run on local TV, containing a complaint against such..)

Absence of aliasing

Actually, recording an unfiltered signal that contains frequencies beyond the acceptable range with an analog or digital system results in something being recording, the fact that it's analog doesn't however magically filter out the offending frequencies, and even an analog system wouldn't yield a correct recording (try recording e.g. a high-pitched whistle with a microcassette recorder: it won't just "low pass filter" the very high frequencies beyond the tape's response.

I will have to try this - I can't say I've ever noticed any problems with aliasing using an analogue tape setup. But I can see where you're coming from (the magnetic domains in the tape material effectively being a miniature and chaotic semi-digital recording mechanism, divided up by the bias signal)... It's the basis for the odd whistling noises that interfere with AM transmissions, rising and falling in note as you move through the dial, and needing the 'beat cut' recording switch on your stereo, right? However, isn't the bias on most decks set to 20khz or more anyway? In which case I'm not sure where to find suitable whilstling equipment - or a suitable microphone with which to record it, as not many non-studio class models are rated for much beyond 18khz. Maybe an old computer monitor set to VGA mode (35-ish khz, so should alias to a very audible 5khz if the theory holds water) with a couple turns of audio cable wrapped round it. Though given the more chaotic nature of the sampling, I can't see it producing a steady, pure and highly noticable tone - isn't this the sort of thing the Beatles tried to copy-protect their albums with, to absolutely no effect? Besides, a simple roll-off filter at 6dB/octave above a certain frequency is a piece of cake and unthinkably cheap to produce (having made one in a physics lab lesson almost without thinking about it, and half asleep, using a basic op-amp chip and a couple resistors and capacitors) so I don't see why it shouldn't be standard kit on a cassette recorder's line-in/mic port, just as it may appear on any other piece of recording or output equipment as part of the basic circuitry.

--tahrey 21/4/07

Compact Audio Cassette capabilities

I can't definitively argue against it, but I would call into question the stats being quoted for CAC in the article and ask for clarification on where the information is coming from or what standards it is being measured against. Specifically, the frequency response seems too low, and the potential signal-to-noise ratio too high.

First of all, the claim of 14khz frequency response being "good" for a high-end Metal tape - what's the dynamic cutoff level here? I've had enough vanilla iron oxide cassettes that are good for about 16 to 18khz (that is, still making an audible difference on playback vs. a test recording with these ranges filtered out) before the sound starts to sink into the background noise, and it's still discernable on a frequency scope (though far less audible) up to about 20khz in a lot of cases. There have been a few - mainly from friends with ancient decks - that have sounded rather flat in comparison, and on testing have revealed a response as poor as 12-13khz, but these are notable by their rarity... certainly if all my Ferrics came out like that, I'd have chopped in my simple FeO-only (not even dolby-equipped) hardware for higher-end multi-type friendly stuff long ago. Is that "14khz" figure corresponding to some artificially demanding measure of response, e.g. is attenuated less than 3dB from source (barely noticable dulling) or somesuch?

Secondly the dynamic range - 60dB, and higher? That's got to be some very high end kit, with a finely controlled recording level, overdriving the record coils to the very brink of clipping and distortion. I've only seen one or two - expensive - tape decks advertised as being in the 58 and 60dB region, and not come across anything in practice that delivers a true SNR (i.e. "minus" level of background noise vs almost-clipped loudest sound at -0.01dB) much beyond 45dB. True, I'm not using Dolby, but even so I haven't heard it claimed or seen it in practice to remove over 15dB of noise... certainly not 25dB (if this is true, point me towards the dolby-C equipped decks with my bank card in one hand and credit card in the other ... need an excuse for a new stereo anyway!). Even careful and concerted digital noise-filtering efforts on commercial cassettes I own can't push the SNR much past 66dB without inducing a whole load of less desirable warbly side effects, much like overcompressing an MP3.

Shoot me down, correct my ignorance, whatever, it's all good, as i'm quite intrigued by these ideas displayed here. --tahrey 21/4/07

Hello. I made the additions on the compact audio cassette based on this info from the Britannica:

'At the high-frequency end of the spectrum the weakest link is still the cassette. While many decks can claim a frequency response (+/- 3 dB) to 18 000 Hz and a few go to 20 000 Hz or slightly beyond, the response measurement in this case is made at a low (-20 dB) signal level. At a 0-dB recording level even premium-quality ferric and chromium dioxide-type cassettes begin to reach tape saturation at about 7 000 Hz; with metal-particle cassettes this high-frequency saturation point is extended nearly an octave higher (14 000 Hz).'

'Studio master tapes can be made with the professional Dolby-A noise-reduction system to approach 80 dB [dynamic range], and the consumer Dolby-C system can raise the measured signal-to-noise figures for cassettes to about 70 or 72 dB. (A different noise-reduction system, dbx, can achieve a dynamic range of nearly 100 dB, but its incompatibility with the widely available Dolby-B and Dolby-C systems has tended to limit its availablity.)'

Stark, C. (1989). Encyclopedia Britannica, 15th edition, Volume 27, Macropaedia article 'Sound', section: 'High-fidelity concepts and systems', page 625.

Enescot 07:07, 2 May 2007 (UTC)

So many unknowing people

This article is as many other audiotechnology related articles pretty much complete nonsens. Take SACD for an instance. SACD is not, has never been and never will be digital. It is a 1-bit signal, which means it is analogue. Besides that it is a PDM-signal and PDM (and PWM) are analogue signals. Only PCM is digital. And digital signals are not necessarily binary signals in the physical sense. Only in the conceptual sense. SACD is analogous to the analogue signal and as such analogue. Calling it digital is a deliberate (marketing-)misnomer - or perhaps a result of confusion between "digital" and "binary" (two VERY different concepts!). Kristian Poul Herkild. 80.167.218.195 15:58, 23 June 2007 (UTC)

Sort of. Yet the modulation is called Direct Stream Digital. — Omegatron 17:43, 23 June 2007 (UTC)

Wait a second. PWM is analog, but how is PDM not digital? Digital means that the signal is quantized and carried by discrete symbols, so that the signal can travel through a noisy channel, and as long as the symbol is received, no noise is added to the signal. In PCM, the symbols are strings of binary digits. In PDM, the symbols are single binary digits. The PDM signal can be passed through multiple noisy channels without any degradation, as long as the noise isn't large enough to cause bit errors. The values are either 1 or 0 and the clock is carried along with the signal (I would imagine a good receiver uses a PLL to sync to the bit stream and re-generate nicely clocked pulses before sending them to the demodulator, so the exact original signal is reproduced.) — Omegatron 23:41, 27 June 2007 (UTC)

Of course there is no "pure digital" signal anywhere in the world, unless you are referring to elementary quantum physics states. For all other practical purposes and applications, even digital signals are conveyed by "analogue" means. What makes them "digital" is not the pulse-shape, but their having an INTENDED restricted alphabet of symbols, and being quantized in both the time and amplitude domains (there are hybrid systems too: e.g. PWM can carry analog non-quantized signal width information, but it will be necessarily quantized in time. But SACD uses a signal with a precisely quantized width (1 or 0, no matter what logic voltage levels are used to represent those) and a precise time quantization (some 1.1 MHz sampling rate), so there's nothing at all to classify it as analog. This whole discussion reminds me of the trick question "how can something as analog as a wire carry something as digital as data". EpiVictor 18:02, 28 June 2007 (UTC)

One early supporter of digital audio was the classical conductor Herbert von Karajan, who said that digital recording was "definitely superior to any other form of recording we know".

Sorry but this is neither a technical reference nor a musical reference ! von Karajan was payed and much sponsored by Philips during the development of the Compact Disc. If you read the news of the 80ies you'll discover all this partial situation.

~~Mario C. 22h04 May 4th 2008~~ —Preceding unsigned comment added by 83.186.18.106 (talk) 20:05, 4 May 2008 (UTC)

I think we're getting a bit confused between the meanings of digital, analogue, and signal. The vibration of something will cause a signal to be propagated through air (water, buildings, etc, etc) as continually-changing pressure. We hear this as sound because this signal causes our ear drum vibrate, which we detect and interpret as sound. If we want to capture this signal of continually-changing pressure, store it and/or transmit it, and then reproduce it, we have various options. A typical microphone will convert this signal of continually-changing pressure into a signal of continually-changing electrical current. A pre-amp will convert this to a signal of continually-changing electrical potential difference (voltage). The electrical signal is considered to be analogous to the pressure signal - hence being called an analogue signal or analogue audio. We can transmit this signal and receive it. One simple way is down a length of cable with no real need for any conversion or modulation. With proper termination, many miles can be achieved. The signal is received the other end. This signal could also be transmitted through the air (radio) by using the electrical signal to modulate the amplitude of a suitable and fixed carrier signal (amplitude modulation) or it can be used to modulate the frequency offset of a suitable carrier signal (frequency modulation). Another type of modulation is pulse code modulation where the signal goes through a periodic sample-and-hold process during which the signal is compared to a number of fixed and equally-spaced quantized steps--each of which can be numbered 0 through whatever. It is this numbering aspect that gives us the term digital. In France, they call it Numerique (which I believe is more correct). We say digital because you count with your fingers (digits). Digital means numerical. The original signal is now converted to a sequence of numbers. The more numbers (samples) there are per second, the higher the frequency of signal that can be converted to numbers. The more quantised steps there are, the more accurate is the conversion (lower distortion and lower noise). That sequence of numbers can be used to reconstruct an electrical signal. Simply generate a potential difference level equivalent to each number and the sample-and-hold signal is recovered. Pass this through a low pass filter and the original signal is reconstructed with the addition of some noise and distortion (given that the original signal had no significant components above 1/2 the sampling frequency and did not over-modulate the system). The magic is that whilst we have sound represented as a series of numbers, we can do many things with these numbers. We could write them down on a length of toilet paper in decimal and that would be a digital audio recording. We could represent the numbers as a base 2 (binary) sequence and use computers to store the numbers on hard disks. We could convert the numbers to a base 3 (tertiary) sequence and broadcast them with a PAL television pictures (as does NICAM). We can even do maths on the numbers and change level, add compression, filter, error correction, edit, etc, etc. We could even do perceptual coding to the numbers and reduce the amount of numbers we need to represent the original suitably-enough. It's still digital (numerique) audio, but it doesn't have to be PCM. When transmitting / recording the sequence of numbers, the signal used must be chosen for the medium that the transition will pass through in such a way as the sequence of numbers can be received without error. Send a signal of high and low voltage levels down a cable and the received signal will be rounded off. So long as it is clear when a 1 is intended and when a 0 is intended, then the number sequence can be preserved. Pulse Width Modulation is not quite analogous to the original pressure signal because it needs to go through a low-pass filter to reconstruct the signal. Just because it is oscillating between two states that could be represented by a 0 and a 1, does not make it digital audio because the signal is not a series of numbers. It is simply a square wave signal that has had its mark-space ratio modulated by the signal we want. A PWM signal can also be generated directly from the PCM. The feature that PWM is a 2-state signal is attractive because this signal can be re-quantised without degradation, because it's pretty clear when it's supposed to be high and when its supposed to be low, and stored given a medium with high enough record frequency. I don't think we should get too hung up on when audio is digital or not. Digital is great for storage because it becomes very simple to maintain the quality by simply storing the number sequence without losses. The care needs to be in the choice of digital modulation system. PCM is very inefficient, because the quantisation is linear (our hearing is logarithmic), but accurate and predictable in its performance. A 16 bit system will give you a noise floor around 96 dB lower the peak, but, for many the distortion as the signal gets lower in level towards the lower bit-depths is rather noticeable and not at all as musical as the "warm" harmonic distortion you get from nice things like valve amplifiers and vinyl records. A 24 bit system is preferable because the noise floor and distortion of lower level signals is not really an issue. 48 kHz sampling rate is considered enough in most circles. The absolute maximum signal this could produce is 24 kHz, but to give room for anti-aliasing filters and reconstruction filters, the band pass needs to be limited somewhat lower than this. Most system need to reach 20 kHz to be considered hi-fi, but 15 kHz is probably enough for most people above the age of 15. The increase to 96 kHz is probably not worth the 100 % increase in data, but the use of 24 bit is certainly worth the extra 50 % of data from 16 bit. Analogue is great when you don't need to store it and the transmission channel is pretty clear. If you do want to store it as an anologue, then the degradation of the storage medium through time is very likely to degrade the quality of the record. Digital storage will degrade, but error-correction and systematic re-recording could preserve the record indefinitely in theory. If you do want to store as analogue, perhaps the very best you can do is actually encode it with Dolby SR and then store it on a 24 bit 48 kHz PCM recorder (together with the Dolby Noise!) for later replay through another Dolby SR decoder! That would sound very nice indeed... anon... —Preceding unsigned comment added by 79.79.92.206 (talk) 20:29, 11 July 2008 (UTC)

Over-simplification

The arguments on this page are in my view highly problematic and the article's structure is unclear. Clearly digital and analogue sound storage and reproduction systems both have their own different faults. The extent to which these faults are audible depends on both the quality of equipment being used and the sensitivity of the listener to particular types of distortion. The fact remains that there is no such thing as a perfect system whether you're talking analogue or digital.

I appreciate the article's attempt to discuss the debate in quantifiable terms, but the fact is that it's very difficult to understand the relative subjective effects of different types of distortion. For example - who can say whether wow and flutter or jitter is a greater contributary element in the reduction of "accuracy" of a system. It's possible to measure both of these factors but harder to quantify what they both mean subjectively and in terms of absolute fidelity. Although it might be easier for the ear to detect wow and flutter than jitter, by the same token, it might be easier for the brain to "correct" or otherwise allow for the error.

Can I suggest that the article is over-simplistic in its discussion of both measured and subjective factors, particularly on the side of analogue sound. The discussion of analogue LP playback is at a particularly basic level. In high quality vinyl playback equipment problems such as rumble and wow and flutter have for a long time been reduced to a point whereby they are largely inaudible, and certainly reduced to a point where subjectively they have no influence. But even within the "mature" technology of LP playback there are still problems inherent in the electro-mechanical process which are not possible to measure, or at least are not yet fully understood. I know far less about digital technology but to suggest that we understand all its problems, or worse to suggest that it is without problems is arrogant at best.

Until we are able to measure what the ear is capable of detecting and the brain's absolute response to that (on a conscious and subconscious level) I would suggest that subjective responses and judgements are both a useful and neccessary tool and should be seriously discussed within this article. Statements such as "analogue sounds warmer" are simplistic to the point that they are insulting to anyone who critically listens to recorded music. Subjectively I hear differences between both systems which are far more complex than these kind of remarks. For example, to my ears a good digital system resolves "space" better than an equivalent analogue system. Yet to my ears digital cannot resolve "physical presence" with the same degree of accuracy as good analogue. The terms I use could fairly be described as vague or unscientific but ultimately at the heart of this debate is the fact that there are areas of audio engineering and psychoacoustics which may never be fully understood.Bluetomgold (talk) 04:00, 25 February 2009 (UTC)

Nicely put. Binksternet (talk) 04:39, 25 February 2009 (UTC)

Noise Gates/Loss of "nuances" and "cues" used by the brain in soundstage perceptions

I haven't seen this covered or even know if this is described correctly. I used to compete in IASCA car audio competitions and at the time (10 years ago) the discussion centered on digital recording equipment often having a noise gate employed with a threshold high enough that right from the studio, certain "musical cues" and "nuances", such as the reflected sound of fingers squeeking on the fretboard of a classical guitar, and the time it took them to echo off the walls of large rooms VS small, were things often lost in the absolute world of 1's and 0's of digital recording. I mean we're talking miniscule bits of sound that you could even call noise or a mistake, but they were arguing the brain processes them nonetheless and subconsciously uses them to establish the parameters of an imagined soundstage as one perceives when the "phantom channel" is realized. Any thoughts on this? Batvette (talk) 21:12, 9 June 2009 (UTC)

If it's a noise gate cutting in, then it has nothing to do with digital limitations! Oli Filth^{(talk|contribs)} 23:13, 9 June 2009 (UTC)

So (not trying to be argumentive, don't have a "side" in this issue) the digital medium does not act in itself as a "noise gate" inherently by having a minimum amount of information required to register a 1 instead of 0? I may have misstated the point with my wording, I implied an additional piece of equipment (noise gate) was used, my bad, I meant to say it was implied that the use of digital VS analog was inherently a process which had a noise gating effect in itself, and tiny bits of information only recognized by the subconscious interpretations of the mind were lost. (allowing that this may have been complete BS perpetuated by the pro-analog group at the time, just thought it beared mention)This is probably a philosophical argument which could be put to rest with the most rudimentary technical points. It is akin to arguing that the transformation of electrical signals which are curved waveforms to bits of 1's and 0's must inherently be changed (with possible lost information)even if the ear can't hear the difference. Such a philosophical argument would serve to give people a "hole" to argue from which leaves them free from the encumbrances of the absolute technical data to prove them wrong- but since we're talking about "music" here which has no "good" or "bad" definitions, some subjective posturing would be expected with emotions involved. Were they saying that to validate the "it doesn't sound natural" theory even though technical facts show that is wrong, or do technical facts follow the old "the difference is so small it's inconsequential" fallacy?Batvette (talk) 01:23, 11 June 2009 (UTC)

There is a well-established theory (the Nyquist–Shannon sampling theorem) behind analogue-to-digital and digital-to-analogue conversion. Good quality analogue-to-digital and digital-to-analogue converters have very good performance. This can be established objectively, e.g., by measuring linear and nonlinear distortion. These measurements can then be related to subjective performance by conducting properly controlled, double-blind tests.

Your description of emotional involvement confuses me a bit. We're talking about electronic systems which have measurable performance. Double-blind tests tell you whether or not the performance of these systems is adequate.

Kind regards. Enescot (talk) 03:46, 18 June 2009 (UTC)

In mentioning emotions I was suggesting that is a factor behind motivations of people arguing this issue, their love for music might get in the way of their judgement on technical issues. On double blind testing, I'd actually disagree as what I was suggesting was that barely audible cues that are used by the brain subconsciously in a live performance to place the parameters of the soundstage in the mind, were potentially lost if too subtle in the conversion. I am told it is difficult for studio recording technicians to properly capture some of these nuances, which can reproduce that "phantom channel" effect where one closes their eyes and the system becomes transparant and the perceived soundstage in the mind far exceeds the listening room's dimensions. We wouldn't be listening for them, probably couldn't identify them if we tried, but we'd sense something was lost in the music if they weren't there. All that said, (mostly to clarify what I was getting at) as the below user describes, and you yourself imply by mentioning a measurement of distortion, this is probably a faulty theory as nothing, no matter how subtle, is lost if the threshhold is so minimal.Batvette (talk) 13:35, 30 July 2009 (UTC)

If you had ever seen how many ones and zeros are involved in recording just 'dead air' in a quiet recording studio, you'd realize that the entry level for one bit is so low as to be inconsequential. Going from a quiet studio to the first tiny touch of a finger on a violin string is not like going from a row of zeros to a single "one". Binksternet (talk) 15:41, 18 June 2009 (UTC)

Question answered. Thanks. Batvette (talk) 13:35, 30 July 2009 (UTC)

I am sagar?

I'm sure I'm forgetting the right place to post this question, so I'll stick it here. Is there any clear "winner" in this discussion, or, as I suspect, is it all simply preference or (possibly imagined) advantages? If the former, I couldn't find it in my twenty minutes' reading of the article or in an hour's perusal of Audiophiles. If the latter, could someone create a clear list (succinct as possible, I know it's difficult)of pro/con's for each? Very interested in the topic, any help on issue would be greatly appreciated. Thanks "Are you suggesting coconuts migrate?" (talk) 06:40, 25 December 2009 (UTC) 32Lex22

Wikipedia's articles are meant to be neutral. Did I solved ur confoosion ^~^? Ivaneduardo747 (talk) 01:39, 23 March 2011 (UTC)

If you look at sales of recordings and recording equipment, digital is the clear winner. If you want to know which sounds better, it depends on who you're asking and what they're listening to. --Kvng (talk) 16:57, 24 March 2011 (UTC)

Suggested extension to the section on Analog Warmth

Philosophers who have addressed the issue of "analog warmth" point out that its source may result from unique imperfections that appear on only one copy of an analog LP (a token of a type), played on a specific system, and at a particular time wherein the entire environment works to create an irreproducible moment. This moment may not transmit the live performance "accurately" and is considered by the musicologist and philosopher Theodore Adorno a distinct esthetic experience (cf Current of Music published by Polity international and Stanford University Press, but I'll be damned if I do you convenience store clerks your work for you any more than I goddamn well have to). A digital component would reduce the uniqueness (what Adorno's friend Benjamin called the "aura") mathematically.

In riposte to claims that such nuance is "beyond the range of human hearing", Adorno's riposte, or at least the printable part, would be that "studies" of "human" "hearing" use empirical subjects either selected or self-selected out of a normalized population which would, under actual conditions of production, not include audiophiles and people with superior hearing.

I seriously think the above should be included because the rest of the article is mostly bullshit scientism, typical of wikipedia's borderline retarded output.

Edward G. Nilges Hong Kong 4 Jan 2012 — Preceding unsigned comment added by 203.218.45.1 (talk) 16:36, 4 January 2012 (UTC)

Stalking Edward from the sopa article, he seems to have correctly identified some problems, and my god this article sucks on approachable language, it's on my mental list of things to do. Anyhow Edward, I'm not so sure such flowery language is better than the audiophile and engineering jargon that's crammed the article full, but if I can find it, I'll include it. Maybe a quick comparison to 'old fashion lightbulbs' (incandescent ones), pointing out that whilst they don't render colors so perfectly, they indeed impart warmth especially to people and portraits, restaurants, and so forth, because like analog recording, or at least LP's, their dynamic range is so characteristic. Penyulap _talk 19:51, 24 January 2012 (UTC)

Recent experiences

I came unto tube amps quite accidentally five years ago; I found an early-60s Sherwood tube stereo (possibly the last one before the change to transistors). I dug out a SONY turntable, two Yamaha natural speakers (also found), pulled an MJQ record. (Even before the amp warmed, my cat became very interested in the sound coming from the needle; to this day, he only "smiles" for the MJQ.) When it warmed, it was as if the long gone quartet was in the room -- it was not like listening a good stereo (such as my AV 85 PY), but as if they were actually in the room. Over time, the driver tubes gave out and were replaced with guitar amp tubes, and then other problems including a ceramic resistor (mechanical) breakdown put the Sherwood into storage.

I think that the 60-70s hunger for music drove the conversion to transistors, as everybody needed music (or other sounds) to learn from to perform and not necessarily appreciate such as audiophiles of the 50s-60s had. I know that I was satisfied with second-rate recordings of African and Caribbean music from distant radio stations recorded on salvaged tapes with salvaged recorders, as I wanted the sounds to learn from.

Getting to the heart of the matter -- I attempted to record these fairly distorted but usable tapes to my laptop, and the sound going from the tape deck through the computer (audacity) and to the solid state amp was so much further distorted that it because useless. When piped directly to the amp, it is actually pretty good, which is what I am listening to as I write this (perhaps the only recordings of college station gigs to survive).

I am not seeing any material actually describing what I am experiencing; and, no doubt, there may be a rush of what I call "AudioFOS" to attempt to "protect" the existing material here promoting digitalization.

I attempt to abstract tube amplification (as ArtSci) so that I may someday engineer tube amps, as I want the MJQ back in my living room.

Another example of how digital/solid state weakens music is Jimmy Guffre's third-stream Serenade, which I believe is the finest piece of performed music (through the Sherwood tube amp from a record recorded with tubes), sounds labored and weak in the Amazon MP3 version.

It may be possible that I would benefit from a high-dollar digitizing device, though I cannot (yet) find material suggesting this (most material supports sound cards), but I really don't see spending money on such a thing when I will benefit most by sticking with what I have (including an also-found early-90s Technics cassette player).

Also, I believe "warmth" is a red-herring (or more audioFOS). --John Bessa (talk) 17:24, 17 February 2012 (UTC)

Misstating what "analog" means

While the explanation given (direct correspondance to a physical property) may have originally have been true (I will not presume to judge this), it is not in accordance with modern usage. The difference between analog and digital can, with some oversimplification, be stated as the difference between a continous dependency and one that considers (e.g.) a signal a series of discrete information units. (This statement is off the top of my head and may be sub-optimally formulated; the principle is still correct.) A physical dependency is only needed in as far as everything (including the digital signals in a computer) ultimately has some physical connection. 88.77.150.159 (talk) 02:09, 22 February 2010 (UTC)

There is some confusion regarding Nyquist sampling. This is applicable to cw sine waves that are time invariant, i.e. continuous with time. In this respect it is accurate as stated by the theory. However, this would make for very boring music. Luckily, most music involves dynamic transients that result in a time envelope around the fundamental frequencies of the instruments. The fundamental note may only be at a modest frequency of a few kHz. However, to reproduce the time attack and decay of the note accurately can require much higher sampling frequencies to reproduce the time variation of the waveforms. To reproduce these requires sampling the waveform at a time resolution of atleast half the finest transient time that is needed to be reproduced. The other information that is needed to provide an accurate stereo 3-D soundstage of the music is the correct phase of the frequencies. i.e. It is not enough to reproduce a 10 kHz signal, but its time alignment with respect to the other frequencies in the overall signal needs to be correct. Simple Nyquist sampling does not have enough information to reproduce this. This provides the spatial placement of the instruments that adds to the interest and accuracy of the music reproduction.

The specifications for typical casette decks are too pessimistic in the article. Typical midpriced casette deck available in the 1990's such as a TEAC V-1030 provide a usable SNR of about 70 dB with Dolby B and 80 dB with Dolby C. The frequency response of three head casette decks can extend from 15 to beyound 21,500 Hz at -10dB to -20 dB levels and to better than 20,000 Hz at 0 dB with suitable tape. Please note that this is +/-3 dB and thae actual usaable frequency response extends further into the frequency extremes. By adjusting the tape bias for a given tape formulation, very good reproduction of vinyl LPs or Bluray soundtracks can be provided with extended high-frequency response. —Preceding unsigned comment added by 207.35.47.66 (talk) 22:57, 3 May 2011 (UTC)

Your statement regarding Nyquist sampling is mostly correct. It applies to periodic, time invariant signals, which does NOT include audio recordings. For the time invariant signal you must sample at twice the highest frequency component in order to reproduce exactly. For signals that vary in time you must sample considerably higher, like 4-10 x the Nyquist rate. That the 44.1kHz sampling rate of a typical audio channel works as well as it does is only because the frequency components of audio above about 6kHz are quite small most of the time. Music having a synthesizer generated signal can contain significant signal above 6kHz, though. DSP's that handle comm signals need to sample much higher than the Nyquist frequency in order to properly recover the signal. You are also correct in stating that this (44.1kHz) undersampling produces phase and amplitude distortion in the reconstructed signal that affects the perceived spacial placement of the instruments, especially at high frequencies.71.214.218.217 (talk) 19:15, 9 August 2012 (UTC)

I have reverted uncited changes you've made to the article. Your assertions are not supported by Nyquist–Shannon sampling theorem or Sampling (signal processing). --Kvng (talk) 00:21, 12 August 2012 (UTC)

Katz research and conclusions about low-pass filtering

Added a paragraph relaying his results in his book that suggest variance in sample rates is due to pre-A/D conversion filter design. Feel free to discuss. Radiodef (talk) 18:16, 13 September 2012 (UTC)

Nice addition. I'm not convinced this directly relevant to the A vs. D topic of the article; It's more of a D vs. D discussion. The whole section might want to be merged into Sampling_(signal_processing)#Sampling_rate. --Kvng (talk) 21:30, 13 September 2012 (UTC)

The section on higher sample rates in this article could be better used, I think. Some say that analog is better than digital because its frequency response is wider and "more detailed". In the case of analog tape, for instance, 15IPS goes up to 30kHz, and exhibits a fall-off slope, like a low-pass filter. At higher frequencies with digital, the wavelengths are short enough that there isn't much space between the positive and negative ends of a cycle. Figures e-g of the second image here show what I mean, though the chapter in this book is about something different:

http://www.dspguide.com/ch8/4.htm

There are various reasons that this could mean that analog performs better, and I think that's what the higher sample rates section is trying to get at. Technically, from a mathematical perspective, no detail is really lost in the high frequencies of sampled audio, but it looks that way. Radiodef (talk) 16:12, 14 September 2012 (UTC)

The narrowness of the record and playback head gaps of the tape recorder combines with tape speed to set the high frequency response limit on tape. Bias signal has to be jacked up for narrower head gaps, but too much bias reduces the high frequency headroom. (It's not just tape speed.)

There's a practical human limit to all of the comparisons: most people cannot hear much above 15kHz, and even a Golden Ears person has a high frequency response that is quite a bit less sensitive than the range between 1k and 4k. Binksternet (talk) 16:34, 14 September 2012 (UTC)