Red Book Riddle Solved
default -- Fri, 07/10/2009 - 03:56
The advent of the Meridian 808.2 CD player and Rober Harley's comments on it prompted me to write this piece on standard resolution ("Red Book") digital audio. To be sure, advancements in playback have been very real over the past two decades. But with the advent of minimum-phase / "apodising" filters and the massive improvement they brought to CD, a lot of people will be wondering if 16 bits / 44.1kHz sampling was enough on disc, for playback.
The answer, after decades of debate and endless argument amongst audiophiles and audio writers, seems to be yes, to this big question. Beyond mike technique, venue, etc. an encoding format has three main areas of concern: capturing (sample rate, etc.) production and playback. Many audiophiles think there are only *two* main areas concerning a format - recording and playback. This is wrong - there's one more in the middle !
As it turns out, the areas affecting CD's sound were the last two - production and playback. The production problem was solved years ago and was an instant fix, as I show below. But first an overview: 16-bit digital encodes 95db of dynamic range - roughly 25db more than a large symphony orchestra produces. And orchestras produce far more "range" than a rock band or string quartet, both of which have dynamic ranges in the 50's db. This fact links up with important findings in recent years by audio researchers. One of them is Bob Katz. - Bob says for "original sources", 16 bits IS ENOUGH for all kinds of music. By "original source", he means the encoded music before it was processed. This observation was derived from controlled listening experiments done over the years, that either Katz or others conducted. The problem, as (all) recording engineers know, were the intermediate calculations done on the signal after recording - the "DSP processing" done in production. With DSP, there were losses in a 16 bit recording...losses that eroded into the music - esp. symphonic-style and grand piano.
But in the early 1990's a technique came along that cured the problem - 20 BIT RECORDING. This would absorb all the production losses that hurt 16 bit recordings. Right after this, a technique called noise-shaping came out - another improvement - but this time for playback (on disc). As for the CD-standard, I quote Malcolm Hawksford from the March 1996 Stereophile article "Bits is Bits ?": "When correctly dithered using triangular PDF dither, a 16-bit digital audio signal possesses a dynamic range of 93.3db with zero distortion and zero noise modulation. The 16-bit format holds the possibility of even *higher* subjective dynamic range - up to 18db more - with minimally audible noise-shaping employed during CD mastering".
Knowing this, why would *anyone* desire a new audio format ?!! Zero distortion, zero noise and dynamic range of 110db - that's "not enough" ?? One of the arguments for more bits was that "it's not the number of bits available...but the number used at one time". Yes, but that was a *production* issue - not a capturing or playback-disc issue !! The same goes for the sample rate - *if* more than 44.1kHz was needed, we got it - in capture - starting 16 years ago. The problem was that audio writers never explained the (real) virtue of 20-bit recording and that 16 bits might be enough, on disc
All this wasn't enough, however, as advances in playback were needed as well. But the issues in playback seem to concern the laws of electrical current and audio-chasis design much more than the "band-aids" needed to improve CD sound. Yes, CD was "upsampled", etc...but let's not forget that some CD playback systems *didn't* upsample and produced great sound - like Zanden. In other words, most of the parameters concerning digital playback had to be improved anyway - no matter how many bits were on disc.
That said, it's not a surprise that CD sounds as good as it does today. Some audio writers are struggling to hear a difference between the Meridian player mentioned above and true Hi-Rez audio. Robert Harley, strangely, did not compare the 808.2 CD player with his Hi-Rez files. That's probably because CD sounds a lot closer to Hi-Rez than he thought it could (or should). But it's right in the specs.......
These (new) observations are not a surprise to everyone. Below are four articles I dug up, starting with J. Gordon Holt from the mid-1980's and continuing forward. A little perspective and some solid answers is what we need. Enjoy......
http://www.stereophile.com/asweseeit/899/
http://www.soundstage.com/gettingtechnical/gettingtechnical200310.htm
"Conclusion" part
http://www.stereophile.com/news/110104aeshirez/
http://www.head-fi.org/forums/f133/24bit-vs-16bit-myth-exploded-415361/
Opening comment
Finally, an explanation as to why DVD-A sounded better than CD. I mean, if CD's specs contain all we need for sound, then why was DVD-A (which itself needs to be filtered, noise-controlled, etc.) preferred ? (at least through Meridian players).
http://www.novaphysicsgroup.com/page7.html
Towards bottom - "micrograph" pictures
All content, design, and layout are Copyright © 1999 - 2011 NextScreen. All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
That's my piece above. For the Nova Physics link, I forgot to inc. "WWW".
I edited the URL so folks can click on it.
Since so-called higher resolution playback (i.e. SACD) is mostly a dead issue for the time being, I take that your point is that technically the possiblity that Red Book CD could sound great has been understood for some time? In the long run, one supposes that theory triumphs, though in the short term empiricism tends to have a field day. That explains some of the different views on this over time. For example, I think the reason for all the excitement about certain D/A converters isn't based strictly on an assumption the Red Book couldn't work theoretically (most commentators don't know and those who are technically knowledgeable know enough to know they don't know everything), but that practically it seemed somewhere between hard and impossible to build a converter that realized the potential.
As a consumer, I would be frustrated when told that Red Book is in some way undebatably perfect (see the Head-Fi link above) only to find that much CD material sounds compromised in reality. Of course, the undebated perfection is much like Newton's physics: it ignores a lot of practical or uncertain matters (e.g. noise, filters, jitter, how we hear, what we can hear) to get at a basic understanding. Newton was thrilled with this, and so were rocket scientists, but none of them tried to put a man into space with Newton's assumptions alone. They had to go farther to make things work in practice.
If your point is that consumers should feel confident in Red Book as "good enough", well that may depend on circumstances and/or patience. The idea of higher resolution formats was, I think, partly that it was presumed to be much easier (you can read that as either cheaper or possible) in practice to generate high dynamic range, wide bandwidth signals. In hindsight, I don't know if knowledgeable people would still think about it that way? In any event, right now, using the Meridian 808 as an example, it seems very expensive to get the best out of Red Book recordings.
Moore's law being what it is, I'm optimistic that the approach Meridian takes in the 808 will eventually show up in $300 D/A converters. I don't know how much processing power is involved, so it would be interesting to get a view on how far off that day is.
A lot of the discussion you present is about bit depth and dynamic range. Since many of the observed Red Book problems don't involve dynamic range, one has to ask if a similar analysis works for filter distortions, audibility of very high frequencies, jitter etc.?
AVGuide webmaster and general drudge
Gadgetman - thanks for your comments. Yes, CD's sound was irritating enough that it prompted folks to ask for whole new audio formats. But there are two problems with this. First are CD's specs: I can't see how these are sub-standard or "low-resolution" by any stretch of the means. Esp. when processing the full 16 bits (of resolution) at home - which was finally possible with the advent of 20-bit recording. Noise-shaping, which is within CD's spec, was the final nail in the coffin.
Second, were the playback gains CD has always experienced. It *never* stopped improving - from the first Meridian CD player in the mid-80's to the Linn Sondek CD12 player in the late 90's to the current crop of players from Meridian, Boulder and Soulution.
Knowing these two things - CD's (Hi-Rez) specs and rapid improvement in sound, why did *anyone* ask for a whole new system ? It's a question I would love answered.
And I agree that formats (like SACD) were easier to decode. But that doesn't make them better. I use the Charlie Hansen analogy of baking: boxed cakes are easier to make than ones made from scratch. But the one made from scratch tastes a lot better. In other words, music waveforms need precise and detailed decoding, not cheaper and "easier".
Finally, yes, it still cost a hefty sum to hear great CD. But even at lower prices, the sound is making big gains. Case in point are the TAS-reviewed Bryston CD player (and now DAC) and Audio Research's DAC 7. Lower prices than this will have to wait for now - but not for long. Actually, digital is quite a bit better than LP - to get the best sound out of its respective format. For the very best LP, you'll have to drop 150K on a Clearaudio Master Reference system. With analog, the more expensive the system, the better the sound. Digital gives *all of us* a chance !!
I think one reason people thought another format was desirable has to do with high frequencies. With a 44.1k system, 22khz is the theoretical maximum frequency that can be reproduced. But in reality, building filters to do this job well seems to have been difficult. My limited understanding is that real filters create problems well within the audio band. Certainly, high frequencies are the empirical bete noire of most CD players and D/A converters. And all this talk about filters assumes that 20khz or so is sufficient. I also understand that there is some evidence that people can hear well above 20khz, and as far as I know, no one is suggesting that a 44.1khz sampling frequency can reproduce signals above 22khz.
The reason that the Meridan 808 (and the Spectral and Berkeley DACs) are so interesting is that they suggest that these empirical problems are significantly filter distortions (filter pre-ringing) that have engineering solutions. Even more interesting is that these companies build D/A converters that address the problem. We can then listen and see if removal of filter pre-ringing addresses the issue.
But, since you ask about why we'd want a "higher-resolution" format, I think the answer in this context is simple: higher sampling rates move pre-ringing out of the audio band without requiring fancy filters. Thus, SACD and minimum phase filters are two alternative approaches to the same problem. Of course those higher sampling rates might bring other benefits, too.
Now this is speculative on my part, but if minimum phase filters are just now appearing (2009) I presume that is because the DSP power required to implement them is large. Then take your mind back to the mid-90s when SACD and DVD-A were conceived. 2009-level DSP power was way over the horizon, so at that time the brute force of a higher sampling frequency probably looked like the better answer. Of course royalties and other commercial factors were part of the picture too.
Personally, if there are two engineering solutions that each acheive the same result, then I think the cheaper one is better. Similarly, the one that arrives earlier is better. Cheaper and earlier is better because it is better for the consumer and gets more people in touch with high quality music. Whether SACD and DVD-A were actually cheaper and earlier is debatable.
Strangely, it is interesting to apply your observation that CD playback (and recording and processing) have improved steadily for the 30 years of CDs existence in a different way. I think some people apply that same faith in technological progress to an imagined world full of high sampling frequency, high bit rate recordings and wonder what might have been. In other words, would the slope of progress have been steeper with a different format?
In seems to me that such a world could come into being once we've stripped recordings from the shackles of a physical distribution medium. In a post-internet environment (which we've been in for some time), recordings could all be distributed electronically in the format you desire. $0.99 for MP3, $1.29 for Red Book, $1.49 for HD of some sort. Such an approach seems to benefit everyone.
AVGuide webmaster and general drudge
We've been recording beyond 44.1kHz since 1993. *If* there are benefits to doing this, we got it (in capture) many years ago. I say "if" because not all recording engineers are convinced that sampling beyond 44.1 is better, like Peter McGrath. But they all did anyway - 88, 96, 192 - you name it, it was used.
The ultimate proof is in the listening. If we trust the ears of multiple reviewers, then the two-piece Metronome CD system (reviewed in TAS), the Soulution CD player (blogged online by JV) and now the Meridian 808.2 show that *no-one* is complaining - not even in the least - of deficiencies in the high frequencies. These are just a few reviews...but it's just enough, esp. when these guys were apparently quite familiar with great extension in the highs. This, from either top-flight LP or high-sample Hi-Rez digital - on disc.
With high sample rate *capture*, it works. Again, no reason for a new consumer format !!!!
And it's a sobering moment indeed when we see (the best) CD surpass "high resolution" SACD in sound quality. It does - through cutting-edge playback gear. SACD's "easier" system claim is slowly caving in. Let the truth be known !!!!
The head-fi.org article by gregorio makes a fundamental error: yes 24-bits only improve dynamic range over 16-bit digital, thus one would not expect to hear a difference between them, but frankly no one is talking about 24 bit at 44.1kHz sampling! So when one talks about 24 bit digital one usually implies 24-bit at higher sampling rates, like 176.4 and 88.2 or 196kHz, because there is NOBODY that records with 24 bits at 44.1kHz. So when people claim that 24-bit digital sounds better than 16-, it's not because of the 24-bits per se, but the combination of it and higher sampling rates.
It may not matter.....
In my thread opener, I forgot to include sampling rates as far as original-source sound is concerned. Apparently, 44.1kHz is enough before the intermediate calculations are done on the signal. But calculate we must...and phase errors, etc., were the result. That's a big reason why they switched to 88, 176kHz, etc., in the early 90's - to alleviate this problem, IF it was a problem.
The argument for higher sampling rates comes (mainly) from Keith Johnson and a few others. Their claim is that though we can't hear beyond 20kHz, higher rates help the "timing of the transients", etc. A big disciple of this theory is - and has been - Robert Harley.
The problem is that there isn't much evidence for higher rates - in actual listening. Take some of the top (CD) players reviewed. Writers who are *very* familiar with great high-frequency reproduction did not complain, even in the least, of deficient highs in the units they reviewed. One player is the Meridian 808.2. Another is the Soulution - with review by JV pending.
Here's an updated version of my (AVguide) thread opener:
http://www.audioasylum.com/cgi/vt.mpl?f=digital&m=146743
I don't think you have all the information necessary to support your claims. For example, it's been scientifically proven that harmonics go out to >102kHz - see http://www.ocf.berkeley.edu/~ashon/audio/Ultrasonics.htm which makes the eye-opening claims below - or ven http://www.cco.caltech.edu/~boyk/spectra/spectra.htm (the paper that started it all), and I would say that samplig rates even at 196kHz are not enough. The case of cymbals below with the astonishing 40% power above 20kHz makes for a very compelling case of more dynamics and very high sampling rates.
Instruments With Harmonics
Instrument
SPL (dB)
Harmonics Visible To What Freq.?
Percentage
of Power Above 20 kHz
1. Trumpet (Harmon mute)
96
>50 kHz
0.5%
2. Trumpet (Harmon mute)
76
>80 kHz
2%
3. Trumpet (straight mute)
83
>85 kHz
0.7%
4. French horn (bell up)
113
>90 kHz
0.03%
5. French horn (mute)
99
>65 kHz
0.05%
6. French horn
105
>55 kHz
0.1%
7. Violin (double-stop)
87
>50 kHz
0.04%
8. Violin (sul ponticello)
77
>35 kHz
0.02%
9. Oboe
84
>40 kHz
0.01%
Instruments Without Harmonics
Instrument
SPL (dB)
10 dB Above
Bkgnd. to What Freq.?
Percentage of Power
Above 20 kHz
10. Speech Sibilant
72
>40 kHz
1.7%
11. Claves
104
>102 kHz
3.8%
12. Rimshot
73
>90 kHz
6%
13. Crash Cymbal
108
>102 kHz
40%
14. Triangle
96
>90 kHz
1%
15. Keys jangling
71
>60 kHz
68%
16. Piano
111
>70 kHz
0.02%
Also, your audio asylum article mentions that:
> The 16 bit format holds the possibility of even *higher* subjective dynamic range - up to 18db more - with minimally audible noise-shaping employed during CD mastering".
That's another reason people want more bits - why would one care about "minimally audible" noise shaping to get higher dynamic range, when you can have 24-bit audio (at higher samplig rates, as discussed above)
If someone screams RED IS BLACK! long enough and hard enough others may start to believe them.
It's a shame that JR-1's evangelistic fervor isn't directed in more useful ways, such as saving whales or advocating universal spaying for those with IQ's under 60. :)
Personally I've done enough A/B tests over the years comparing my own recordings made at both 44.1/16 and higher bit rates to be be convinced that we do need higher bit rate options. I'm also happy to see that many others, including Keith Johnson and Peter McGrath agree with me.
Some listeners, such as JR-1 appear to be happy listening to 44.1. Fine, listen at 44.1, or 320 BPS MP3, or even with lower. If you listen through a dixie cup and a string they will all sound the same. :)
Fortunately enough of us have determined that higher bit rates offer definite sonic advantages. Also fortunately more and more sources of higher bit rate recordings are available.
I know of several audiophiles who claim that LPs never reached the level of sonic quality achieved by 78 RPM mono recordings. JR-1 would very likely feel right at home in their company.
Steven Stone
Contributor to The Absolute Sound, EnjoytheMusic.com, Vintage Guitar Magazine, and other fine publications
There can't be much information above 20kHz if SACD (with its ultra-high sampling rate) doesn't sound better than CD. I don't know what players you people listen to...but I've had the opportunity to hear some great (CD) machines. And these simply outperform SACD - even in the highs. Audio reviewers agree.
And those familiar with 24-bit/88kHz audio (on playback) are *stunned* at how close CD is to this as well. CD is closing the gap....
Must studies show that humans cannot *hear* musical waveforms beyond 20kHz, no matter what an analyzer shows concerning harmonics. This is why 44.1 sampling was chosen. The creators of RBCD, however, weren't aware of the potential phase errors occuring in production. Hence the move to higher sampling in capture....
We don't need noise-shaping, since 16/44 is at - or beyond - the maximum what music contains. Engineers used it because it was available...and they thought they needed it, at the time.
No-one doubts the advantage of higher-bit/ higher sampling IN CAPTURE. It gives us more headroom *plus* absorbs any lossses which we know were occuring before, with 16/44 recording.
LET THE TRUTH BE KNOWN !!!!!!!!
And McGrath is "hard pressed" to hear beyond 44.1, for music. Stated right in his TAS interview a year or so ago....
> And those familiar with 24-bit/88kHz audio (on playback) are *stunned* at how close CD is to this as well. CD is closing the gap....
I don't know where who "those familiar" are, and generalizations like that carry no weight. What I can tell you, though, is how stunned *I* am when I compare red book CD on my Berkeley Alpha vs 24/176.4 hi-rez from Reference Recordings with the same exact material of the exact same recordings. And by stunned I mean how impressively superior hi rez sounds over CD (HDCD in this case - a major forte of the Alpha's).
> This is why 44.1 sampling was chosen.
Nope. 44.1 was chosen as a compromise considering the then-current technology, as was the 16 bits, and *without* the research available in recent years.
Like a lot of things, the bottom line is in the listening. If you like CD just sit back, relax and enjoy your music. There are oters, like myself, Steve Stone, Keith Johnson and I am sure lots more who have heard hi rez and consider it far superior, and can easily demontrate so with the exact same material in different formats. To be quite honest with you, I have performed similar A/B tests against CD (through the Alpha) vs the exact same material available on LP (VPI Aries 3, Dynavector, Ayre P-5xe, Spectral electronics), by time-aligning the tracks and matching the volume (via the Alpha's digital volume control), and the results will stun you again: LP still easily outperforms CD through that DAC in every single test. A better DAC may well fair better, but until proven so (and considering the number of DACs I have auditioned except for dCS), CD has a long way yet to catch up, if that's even possible.
Examples of the musical material I used for direct, time-aligned A/B comparisons include: Brubeck Time Out (Columbia 180g reissue vs. HDCD); Holst Planets (Decca, Los Angeles Phil, Mehta, 180g LP reissue vs. XRCD 24); Saint Saens Symphony #3 (Munch/BSO, RCA 200g LP reissue vs XRCD2); Diana Krall Live in Paris (ORG 180g LP, Verve CD). Only the Krall transfer sounded as good as the LP, although that doesn't mean CD sucked everywhere else - it just means the LP was still superior in every aspect.
You will do a major service to yourself to go listen to good hi rez, and my advice is get into analog too, since there is so much more material available. Then report back how stunned you are at that point. Meantime, I am afraid your evangelical posts are the laughing stock of online posts.
If you read the Audio Ayslum thread, I'm anything *but* a laughing stock. Actually, viewers there *added* to the high-bit/high-sample absorbtion theory.
And listeners who have heard the Meridian 808.2 and Soulution CD players will put CD in a virtual tie with 24-bit/88kHz. I know what *I* hear through the Meridian...and the editors of Stereophile and TAS agree. (Actually, Harley of TAS never stated it's a "tie", but it sure sounds like it, if you read his review. To this day, Harley refuses to comment on the difference).
44.1 was *not* a compromise. Every study showed that 20kHz is the maximum we can hear. And it must be true if 88, 96, 176kHz, etc. does not sound better, through the latest (and costly) CD-only playback gear. This last point is important - a dedicated machine for the format is the best way to go. The player you cite is not of this variety, sorry.....
> 44.1 was *not* a compromise. Every study showed that 20kHz is the maximum we can hear.
No one disagrees that most people, on average, cannot hear above 20kHz. What you are missing is that with 44.1kHz sampling, you only get two samples per cycle at the highest frequency sampled (22.05kHz) and the picture gets marginally better at 10kHz (four samples per cycle). So while something vibrates at 22.050 times a second, CD sampling will get 44.100 samples in that time frame, which amounts to just two samples per cycle - for all intents and purposes, this basically means the peak and valley of the waveform, hardly representative of anything and the reason why interpolation is sometimes used. This is one of the myriad of reasons why red book digital simply fails to meet the needs and is nothing but a compromise.
Bad vibrations ? It still covers the range of human hearing - no matter how you choose to deny it .
And I don't hear *any* compromise with Meridian's 808.2 - compared to HDTrack's full-band downloads. I call 'em as I hear 'em....
I should say those higher sampling rates don't sound better *compared* to a cutting-edge CD-only player.....
Eclipse
I get so tired of hearing how digital does not sound as good as Analog. Either get better gear or a hearing aid. You are missing the lows and highs and dynamic range of good digital. If all you want to hear is a single guitar with a limited dynamic range then go ahead and listen to your vinyl rig.
> I get so tired of hearing how digital does not sound as good as Analog. Either get better gear or a hearing aid. You are missing the lows and highs and dynamic range of good digital. If all you want to hear is a single guitar with a limited dynamic range then go ahead and listen to your vinyl rig.
You must have missed the list of recording material I listed which clearly indicates large scale orchestral works, which digital has so far been impossible to match in my system. If you consider the Alpha DAC siply an inferior DAC, then fine, you have a point, but I find that DAC to be an excellent performer. My analog is not missing anything from what you describe; with respect to the mucic citations I gave: a) It can measurably reproduce the 16hz organ in Uranus (of Holst's Planets) with aplomb and gut wrenching power; b) the highs are very silky unlike digital's; and c) the dyamic range, for example in Saint Saen's 3rd Symphony, is much wider than digital's, especially in the 3rd movement where the organ makes its solo entry.
I feel you you are basically yet another one of those who just hasn't tried good analog, that's all, so I suggest you give it a try too.
I've compared Reference Recordings HRx 176.4/24-bit files with the same music HDCD encoded on CD and then transferred to my fan-less, drive-less music server. HCD is arguably the best method of getting the most out of 44.1/16. Nonetheless, there is no comparison; high-resolution is vastly superior to Red Book resolution. It's simply not an argument.
Does JR-1 know why 44.1kHz was chosen as CD's sampling rate? Because it fits mathematically with the line and frame rate of black-and-white NTSC video. The first CD mastering systems used 3/4" U-Matic tape for storage. Each horizontal line stored 196 bits, encoded as black or full-scale white video. Remember the Sony PCM-F1 format, with its 44.056kHz sampling frequency? That frequency was chosen to work with the 29.97 frame-per-second rate of color NTSC video rather than the 30 frame-per-second of black and white.
Moreover, 44.1kHz was a compromise (as was 16 bits) to make CD a viable format with a long enough playing time. Remember that CD's specs were established in 1978.
Okay, Robert - then let's hear what you have to say about Meridian's 808.2 compared to Hi-Rez audio. I have the Meridian...and compared it (extensively) to HDTracks, via DCS decoder. Result ? I am *in line* with what John Atkinson experienced in his audition - struggling to hear a difference !! A cutting-edge CD-only player is the way to go. It's a shame you don't know this (or do you?)
And if Mark Porzilli is correct, then removing error-correction bits from RBCD will help it even more. That's one advantage of H-Rez - it isn't so loaded up with these codes.
16 bit coding was *not* a compromise - when it codes 25db more dynamic range than a symphony orchestra produces. They raised the bit rate due to unforseen losses in production. But problem solved, years ago.
44.1 wasn't deficient either, as the mathematical reasons you cite were not the *only* reasons why it was chosen. It codes beyond the range of human hearing - and they knew it in 1978. But here again, they've recorded much higher anyway. So what's your problem ? You side with Johnson - I'll go with McGrath and Katz, who are not convinced that beyond 44.1 is needed for acoustic-style music.
It may take some time for everyone to understand what's really going on here - so I guess until then, we'll have lots of arguing.....