Why DOES CD sound harsh? Seems we may have a REAL answer...

[Griffinator]

Excellent white paper composed by a digital theory genius - well worth the read.

http://www.tllabs.com/files/Digital%20distortion%20white%20paper.pdf

[Mallette]

Horse hockey.

Dave

[Griffinator]

Care to elaborate on why you dispute this paper's thesis?

[WMcD]

I don't think this tells us everything that is wrong.

My read it that he's saying the output of DA system and reconstruction system is higher than predicted, so it is a problem with the analog system. He says the recording engineer's analog system may be able to handle the level but the system used by the home owner may not. That is a big assumption.

Perhaps it is a matter that the line amps following the DA converter go into clipping.

So, it is not that there are illegal digital levels quite so much that they create illegal (depending on the analog equipment) analog levels.

I somewhat wonder about any system where things are being thrashed to the limits of their performance. That goes for the clipping level and noise floor. The dynamic range of systems is well appreciated and one would think people know better than to push things. Yet critics seem to trot out a new theory every month saying a given system is flawed when operated at the edge.

That is not big news to me; what system is not flawed that way. OTOH it is also not news that people do push the limits.

Smile,

Gil

[jdm56]

Wow, you mean some people think CD sound is flawed? Geez, you could knock me over with a feather! They sound pretty darned good to me. Not that I don't have a number of sonic slugs in my CD cabinets, but doesn't just one really good sounding CD more or less prove that the medium itself is not flawed; at least not in any meaningful way? And I've got disc after disc whose sound quality is, IMO, only limited by the fact there's only two channels.

[Mallette]

Like Gil said, the issues with CD are a bit more complex than this paper suggests. I've participated at length in discussions of them in this Forum and touched on them in the "ATTN: DFW Hornheads" thread and others recently.

I am not disputing the paper directly, just the premise as being the main problem of CD sonic accuracy.

The main problem, both IMHO and as indicated by my own experiments in live, high resolution recording, lies with the limitations of the bandwith. The proof here is that, even though my hearing falls off rapidly after 12khz, I can easily detect the difference between a 24/192 recording and a 16/44.1 rendition of the same material from the next room. That's about all the proof I need.

Also, I've recently toyed with the idea that imaging may also be adversely effected by the loss of ultrasonic out-of-phase info, but that's another story. Hmmm. Just noticed JDM above uses a Dynaco QD-1, so just a little more on this...

I had noted in the past that I never get as much ambience recovery from CD's as from LP's with my DynaQuad circuits. Recent experiments recording LP's at very high digital resolutions and then at CD res seem to indicate that ultrasonic information is important to ambience recovery as well as sonics.

Dave

[hwatkins]

The alogrithyms used to convert A/D then D/A in something as trivial as a voice telephone call (say from 500hz to 3000hz) continue to be a bear in the accuracy department - even though our threshold for tolerance is very high. It is not surprising to see the findings in the article. Having misspent some of my youth working on these codecs, I must say that the accurracy we do see is amazing (even with the associated problems).

Of course higher/faster sampling rates will require less assumptive coding/decoding, but I believe we are becoming accustomed to the reproduction and that the better mixing studios are not falling wholesale into the trap mentioned in the article.

[fini]

Henry,

Could you post a larger image of (or a link to) your avatar? I want to see if what I am seeing is actually there, or if I am halucinating (not uncommon). Thanks!

fini

[sunnysal]

an interesting article but the author does seem to throw all his eggs into one basket. also I would remind you guys the article was written for and published by a company,TLlabs, that markets an "improved" digital metering system to avoid the problems mentioned in the article, generally I prefer articles not written to support commercial ends. I see Dan Banquer quoted there, I will ask him over at AA about his thougts on the article. regards, tony

[Deang]

I'm pretty much with JD here. However, I do think the digital medium is more suited to amplifed types of music like Rock. If I was into Jazz, Classical, and anything laden with unamplified instruments, I would more than likely opt for vinyl.

If you pluck an acoustic guitar string, and then the same string on an electric guitar -- there is a huge difference. The note played on the electric is made up for the most part of the fundemental -- there are no harmonic overtones as there are with the acoustic.

When CD first came out in the early 80's, it didn't take me long to notice what was "missing". Listening to cuts like Classical Gas, or Venus by the Shocking Blue -- it just seemed like the acoustic guitar was being played with 'dead' strings. CD has come a long way, but still falls a little short in this area. OTOH, material primarily made up of electrified hash and percussion sounds incredible on CD. Then there is SACD, which is much, much better.

[Mallette]

Correctamundo, DeanG. Within it's sonic limitations, CD has come a long way. However, you can't fix what isn't there.

I've ceased listening to my primarily acoustic CD's except in the car or in other non-critical listening environments a long time ago. However, electric music still sounds just fine on them.

While CD's will continue to account for the vast majority of releases for a long time to come, SACD and DVD will grow in popularity. In fact, I believe 24/192 will be added to the DVD-A standard once blue light laser technology becomes general available. 24/192 is the only digital standard I've heard that actually surpasses LP, especially in being able to do what LP can't...provide surround.

I've urged Hank Waring of FDS Labs to look into making his stereo to 5.1 process available via software. Now that would be cool. Take your LP's and extract the number of channels that floats your boat out of them with no sonic degradation. He says certain mono material, including 78's, also can be processed to multiple channels successfully. Don't confuse this with the old "Electronically ReProcessed for Stereo" method. That was horrible. His process is far more complex and uses all sorts of built-in cues to determine directionaly and image characteristics.

He didn't sound quite ready to get into the software business yet, but it will come.

Dave

[Deang]

I thought DVD-A was 24/192?

[Griffinator]

----------------

On 8/1/2003 7:39:59 AM Mallett wrote:

Like Gil said, the issues with CD are a bit more complex than this paper suggests. I've participated at length in discussions of them in this Forum and touched on them in the "ATTN: DFW Hornheads" thread and others recently.

I am not disputing the paper directly, just the premise as being the main problem of CD sonic accuracy.

The main problem, both IMHO and as indicated by my own experiments in live, high resolution recording, lies with the limitations of the bandwith. The proof here is that, even though my hearing falls off rapidly after 12khz, I can easily detect the difference between a 24/192 recording and a 16/44.1 rendition of the same material from the next room. That's about all the proof I need.

Also, I've recently toyed with the idea that imaging may also be adversely effected by the loss of ultrasonic out-of-phase info, but that's another story. Hmmm. Just noticed JDM above uses a Dynaco QD-1, so just a little more on this...

I had noted in the past that I never get as much ambience recovery from CD's as from LP's with my DynaQuad circuits. Recent experiments recording LP's at very high digital resolutions and then at CD res seem to indicate that ultrasonic information is important to ambience recovery as well as sonics.

Dave

----------------

Uh oh. Here comes the red herring - resolution.

Let me ask you a question:

How is "resolution" an attribute of audio?

Audio is made up of three fundamental components - amplitude, time, and frequency. "Resolution" does not fit into that list.

The article I linked started right in the beginning with an explanation of the already proven Nyquist theorem.

The only thing you gain with 24/96 over 16/44.1 is increased frequency response (potentially 48Khz) and increased dynamic range (approx. -144dB vs -96dB) You do not get any increased accuracy of rendering the audible frequency spectrum.

What you do gain between those formats that isn't immediately apparent or covered in theory books is the reduced audibility of the HP filters that remove the information above 22Khz. That's the only reason that a CD sounds better on a DVD player. 24/96 audio tends to sound better than 16/44.1 because of the increased dynamic range and the HP filter issue. If you do a spectrum analysis on the audio information from the vast majority of DVDs, DVD-As, and SACDs you'll find no HF information above 23Khz. Why? Because it has been proven over and over again in double-blind testing that the human ear cannot hear, nor can it perceive, nor is there any perceptible effect from HF information beyond 22.05Khz.

Nyquist's theorem, proved by Shannon in 1948, states that any frequency within the audible spectrum (<=20Khz) can be accurately reproduced by sampling it at twice the rate of the highest frequency. Therefore, with proper reconstruction filters (which every CD player contains), anything in the audible spectrum can be accurately reproduced using the redbook standard.

What these consumer-grade CDs cannot handle is information which exceeds 0dBfs. This is a reality, and one that the redbook standard tried to incorporate - no sample may exceed 0dBfs in a redbook standard CD. What they didn't consider was that for every sample which hits but does not exceed 0dBfs, there are incredibly good odds that the reconstructed signal will exceed 0dBfs - creating clipping, and a "harsh" sounding final product.

Sure, there are other issues in CD reproduction - poorly designed HP filters that create audible artifacts in the high frequency spectrum, unstable clocking contributing to jitter errors - this is why the high end CD players cost so much more - the high-grade filtering and discrete wordclocking costs money - it's much cheaper to use an integrated IC that does it all, no matter how poorly.

[Mallette]

I'm just not going there. If you can't tell the difference between CD and a live performance, you're a happy man.

In the early years of the last century, a double-blind test conclusively proved that a Columbia record was indistinguishable from a live performance.

Wish I had a penny for every "conclusively proven" theorem that tumbled in my lifetime.

I don't have a dog in this race. Just my own need for high quality music reproduction. Having done one sort of audio engineering or the other since 1967, I suppose I am just a bit crotchety and set in my ways. Please worship at the feet of the Nyquist theory if you wish, but don't confuse it with law.

(EDIT) I'm going to insert this so as to close a breach in my response. The Nyquist theory certainly appears to be fact, at least in this area of the universe. However, it implies absolutely NOTHING about sonic qualities as experienced by the human brain nor does it suggest in anyway that inaudible information does not impact same. It was assumed they did not. IMHO, history has proven this dead wrong. If I am wrong, I am wasting an awful lot of hard drive space filled with useless inaudible info. On the 16th of August, a group of DFW area hornheads will meet in my listening room to explore this topic. They range from teen age to my age and all tastes. Let's see what happens.

Dave

[Griffinator]

----------------

On 8/1/2003 2:44:12 AM James D McCall wrote:

Wow, you mean some people think CD sound is flawed? Geez, you could knock me over with a feather! They sound pretty darned good to me. Not that I don't have a number of sonic slugs in my CD cabinets, but doesn't just one really good sounding CD more or less prove that the medium itself is not flawed; at least not in any meaningful way? And I've got disc after disc whose sound quality is, IMO, only limited by the fact there's only two channels.

----------------

I'm not suggesting that the whole system is flawed. Neither is the person who wrote the paper. Quite the contrary - the suggestion here is that, while filtering systems have improved even in cheap CD decks over the last 10 years, the monitoring systems (specifically meters) in recording studios are failing to account for the reconstruction process and are thus exposing excessively "loud" discs to clipping. Your anecdotal report very much supports this idea - properly mastered discs that allow dynamic range rather than squeezing every last dB of amplitude out of them sound perfectly good on home playback systems.

[bkrop]

LP's can and did provide surround sound. Perhaps you are too young to remember the quadraphonic movement of the seventies. Two discrete channels on each side of the groove. Some records were released with matrix surround, one encoded signal on each side of the groove and processed into four channel also. Not partically hi-fi but it was out there.

[Mallette]

Griff: It's your topic. Hope you do not mind.

The last post gives me an excuse to move this from the "Home Theater" area.

Insert follows:

am posting this here because the crowd in my usual 2 Channel Forum haunt might get restless, even though the topic applies. I suspect it is too long for them publish in "Letters," but thought some might be interested or have comments here. Not sure of it's technical correctness, as it's pretty much stream of consciousness due to time constraints, as well as unedited.

To the Editor, Sound and Vision:

I couldnt help but respond to this article. First, I really enjoyed it. I lived through that original period of quadrophenia, and completely concur that it was artistically and technically a bust. I have a couple of copies of the Biggs Freiburg disc and love it. It was, without doubt, one of the few if not the only 4 channel disc of its time to have any technical or artistic merit. However, at this point, Mr. Ranada and I diverge.

I have been using a classic DynaQuad circuit ever since the first quad era, even when it wasnt cool. Material with significant out of phase information, like a Firesign Theatre or the Biggs recording, sound superb through this circuit. With no active ASP or DSP, what you get is all the music and nothing else. With nothing more than two channels of an excellent analog source sent through this device, Ive clearly heard distinct voices and sounds to the right, rear, above, and everywhere else in the room. I have the old DynaQuad demo disc and some of the material on it is downright unbelievable. Some of it is downright unlistenable as well, but excess is to be expected from the times. Granted, there is no steering or predictable positioning, but the attempts to do this in the 70s proved to be worse than the problem, and even now the processing required inevitably colors the sound beyond my threshold of acceptability.

Mr. Ranada states: The engineering philosophy behind the development of quad assumed that 360 degree imaging was both possible and obtainable with only four speakers, a sonically fatal misconception.

If this were a traditional audiophile magazine, the argument for two channels is enough would be conclusively proven, though even there signs of logical fatigue are beginning to show up. Of course, this is a predominately a home theatre magazine, so the preference for n channels is assumed. However, to make a global statement of TRVTH like the above pushes me beyond my limits. My own listening system has five speakers, two Klipschorns with a Cornwall in the center, and two Frazier Super Monte Carlos in the rear. 90% of the time, these are all driven by a single stereo amplifier fed through a DynaQuad QSD IIL decoder. While both Bell Labs in the 30s and Paul Klipsch all his life advocated a center channel, mine is there mainly because the width of the room (Klipschorns are on the long wall) is not great enough to provide an imaging sweet area without it. Ive hear plenty of 2 front only systems deliver a completely convincing image without a center.

The other 10% of the time, the rear channels are amplified discretely to play the 24/192 recordings I make on location. Yes, 24/192. I built my own machine for this purpose, and have learned more about audio reproduction from it that in the rest of my life combined. An LP recorded to 24/96 is excellent, and good enough to not make one run for the record when you hear it. An LP recorded to 24/192 is indistinguishable. As to live recordings, thats where it gets exciting. In July a friend of mine and I drove to Austin, Texas and recorded a show band called the Asylum Street Spankers live. They are so acoustically oriented they dont even mic their vocals. Further, their show is heavily audience participatory, similar to a Rocky Horror crowd. The fans know the music and the choruses, and join right in. My mic plan was 4 ribbon mics near coincident, back lobes of the band mics over the crowd, back lobes of the rear channel mics over the band. When I got home with my 36 gigs of data, I could hardly believe my ears. With the rear channels properly balanced, the sense of reality was downright disturbing. Never having heard anything like that before, I kept straining to see if there was some sort of edginess, grain, or other distortion. No, just so real as to cause a bit of spatial and temporal doubt. Once I relaxed, I began to re-experience the evening in a way Id never heard from any recording.

So are the points:

The need for a center channel is dependent upon a number of factors, individually or collectively. These include the type of material, quality of the original mic plan, fidelity of the recording, and type and spacing of speakers. There are other potential factors, but these are the most important.

Modern surround, 5.1 and up, is specifically designed for its intended use and deviation from that use results in audio chaos. While Ive heard superb and engaging surround mixes, most sound quite contrived and steered. For instance, forcing of the voices of the aforementioned Asylum Street Spankers recording to the center would completely destroy the image.

Every producer should choose the number of active channels based on the music without having to feel like something has to be on each and very available channel. Two examples:

The choral tradition know as Singing the Sacred Harp or fasola uses voices facing each other forming a square. What do you do with the center channel here? What do you put on the .1 channel? The same would be true for environmental recordings. Do you put the wind on the center channel, or what?

Finally, Mr. Ranada suggests that the Biggs recording produced daring listener-in-the-middle effects more typical of multichannel pop productions. So, in spite of the sonically fatal misconception the one recording that used the medium properly worked. While I can certainly see a .1 channel for pipe organ, Ive no idea what one would do with the center channel with four sources, other than ruin the image.

This was an excellent and fun look back at the past of surround. I suspect my quibbles above can be explained by Mr. Ranadas assumption of a home theatre crowd predominating. This column is, after all, something of a branch of the main theme of the magazine. Ive a friend who is an HT fanatic. 3 kilowatts of SVS subs, 6.1 speakers all Klipsch Heritage, top of the line everything. On the very finest surround material, it is truly extraordinary. On the rest, the flaws stick out painfully.

Every medium has its place, and every medium can be abused.

OK, so Ive rambled on well past the limit for a letter to the editor. Perhaps you will find some of the above to at least provide a bit of office chatter.

Best regards,

David A. Mallett

[ben.]

----------------

On 8/1/2003 10:02:07 AM Griffinator wrote:

Let me ask you a question:

How is "resolution" an attribute of audio?

Audio is made up of three fundamental components - amplitude, time, and frequency. "Resolution" does not fit into that list.

----------------

Griff-

To be honest, I didn't read the article or follow the link. However, resolution is a HUGE component of digital audio. It has great impact on the amplitude portion of sound. I'm sure I don't need to explain that to you.

I will agree that not slamming levels when mastering is in general a key to better sound. It could also be considered an artistic decision. The Flaming Lips' recent releases are a good example of slamming levels with a Finalizer as a pre-mastering process. They sound very loud, but IMNSHO still sound fantastic. It would be interesting to compare a vinyl release to see how much is playback and how much can be attributed to the pre-mastering technique.

[sunnysal]

griif, I think you misunderstood some of the discussion here..."Audio is made up of three fundamental components - amplitude, time, and frequency. "Resolution" does not fit into that list." you just described a waveform. to sample it, and then recontruct it, you have to decide on how many points of sampling and reconstruction you want to use to digitize and reconstruct the waveform. that is why resolution is so important. More points sampled and used to reconstruct more potential accuracy in the resultant waveform, les points, less accurate. that is what he was saying. that cannot be disputed. tony

[Mallette]

bkrop et al:

I just thought about something relevant to the Nyquist theory and the old quad system. Only three of the six I recall could be called "discrete," and one of those is debateable: There was discrete 8 track quad (let's don't go there!), 4 track reel-to-reel (wish I owned an example), and the (I forget which) JVC or RCA LP that multiplexed the rear info above the normal tracks. Minimum cartridge had to extend to at least 44khz. Remind you of something!!!!

Didn't work. Tough for cartridge out there, and even the slightest wear was terribly audible in the rear. Of course, the other three systems, QS, SQ, and DynaQuad worked to some degree or the other, but certainly were not discrete.

Dave