CD DAC Technology

There is thread argument on this subject.... good luck!

http://www.avsforum.com/avs-vb/showt...0#post10301240

1) One simple reason they don't sound the same is because their output levels vary. As I've harped on, technical measurements have found, and others who've taken the time have discovered, there is no guarantee that a player will output 2.0000 volts precisely. The CD spec is 2 volts and if you recall from grammar school regarding rounding and signficant digits, a player can output 1.5001 volts and meet the spec of 2 volts. That in and of itself can lead to people reporting subjective differences that have nothing more to do than simple output level differences. So, unless you are prepared to take the steps necessary to normalize the output of two players, you may never be able to tell if there is something else that's going on.

2) Well yes, but you can have players that...

a) Have some problems with RFI coming down the cable.
b) Some manufacturers play with the reconstruction filter and you can get additional information (distortion?) mixed in with the original signal.
c) There are some other issues, but I'd tend to focus on the above first.

3) It's incredibly cheap as you've noted. Some manufacturers though look to push the technological envelope, eagerly embracing methods and circuit design to make things as perfect as possible. Some look to have players that can play a multitude of media ranging from CD to other formats (wma, mp3, recordables, rewriteables, etc.) Others take other approaches and make heavy aluminum cases as their technological advance.

4) Yes, CD is a bit different as has been noted above. Many though do use their CD-ROMs as players and invest in a decent soundcard. It's not unrealistic whatsoever to find the performance of high end consumer cards or pro cards to technically roundly clock one-box offerings.

5) The adding of "color" is not unknown and can be done subtly or painfully obviously. The end result of which is better is up to you. There's no evidence to support that more expensive CD players have less errors. The error encoding scheme for CD is Reed-Solomon and is quite robust, with virtually all errors being corrected perfectly. IOW, no guessing, no interpolation...the missing bit(s) are restored to their original without the listener ever knowing there were problems. Only when discs are damaged pretty badly might you run into problems.

If something new sounds better than what you have, it may be because it truly does sound better. OTOH, it may mean that over time, your old player is no longer meeting its original performance specifications and measurements. The thing is, without trying to find out if the latter is true, your reasoning and internal justification for preferring a new player may be completely off base.

Way back when in the early 90s, at a meeting of the New Jersey Audio Society, we investigated the effects of digital errors on CD sound quality. A member had a player containing a chip that would output a pulse every time an uncorrectable error was encountered. I designed a pulse stretcher to flash an indicator light when an error was encountered.

Playing through several disks, the indicator light never came on!

OK ... the circuit isn't working. We physically mutilated a disk with a few light scratches. The light still didn't come on.

OK .. the circuit still isn't working. We made deeper scratches. The light began to flash and if you listened carefully you could hear snapping sounds.

So much for those pads you glue to the disks and that green stuff they sold to eliminate internal reflections from the laser and also those shock mounts for the player. Even back then, bit recovery was not an issue with CD sound quality.

Myth busted.

--- CHAS

If you were a manufacturer, charging a bit more and then some for some product, would it be in your best interest to provide statistical studies that said your player sounded different or better than some other arbitrary one? Would it be in your best interest to compare the waveforms from the analog outputs to some other player (btw, there's a downloadable product that lets you do that), normalize the data and subtract one from another? Or would it just be better to focus your attention on some minutae? Some triviality? A brand of resistor or capacitor? Would it be in your financial best interest to just be vague, repeat the old 'everyone hears differently' chant, toss in some break-in jargon, say science doesn't know everything, and direct you to reviewers and other happy users?

I believe the implementation of the filter that reconstructs the waveform after the Digital to Analog converter does its thing has the primary influence on CD sound.

My first Sony player used an analog 'brick wall filter' to remove the digital artifacts above 20 kilohertz. These filters worked fine for frequency but suffered from abysmal phase response. As the frequency approaches the sharp cutoff of 'brick wall' filter, the energy takes longer to get through the filter. This is akin to the trebles coming out of the filter after the music stops. You are not supposed to hear these small delays so why did those players sound so terrible?

Players with digital processing translate digital artifacts to higher frequencies so the filter does not to cut off as fast and the phase response is much less a factor. My first generation Magnovox was an instant improvement. Cymbals sounded more like cymbals rather than garbage can lids clashing together.

This whole digital thing is based upon the work of Nyquist who worked for Bell Telephone Laboratories. He found you can reproduce a waveform if you make snapshot samples of its amplitude at twice the maximum frequency in the composite waveform. But for things to play back properly, a reconstruction filter must respond to each sample with small amplitude oscillations that increase to a peak and then diminish back to small oscillations. The filter response process subtends several sample intervals so all the amplitudes must add together exactly to reproduce the original waveform. If I remember correctly, this is called a Sin(X)/X impulse response.

If you don't get it correct, there is distortion and other kinds of hash that ruins the sound. Put your ear up to the midrange driver or tweeter and listen for the hash. I find it to be worse on some recordings then on others. You need to get filters correct on the recording end also.

Anyone out there with a PhD in information theory?

--- CHAS

Well, did they all sound terrible? It was not an opinion shared by everyone other than a rather small contingent and some of those were and still are Luddites.
Laying around somewhere in my house, and god knows where, there's an old player which I'll try and find do some comparative listening. However, consider that way back when there are other reasons why stuff may've sounded 'terrible'.

1) How it was mastered. IOW many applied similar techniques used in vinyl reproduction to this new medium.

2) Practically speaking, vinyl is a medium that has a far smaller bandwidth, greater noise floor, varying levels of distortion, crosstalk issues, less dynamic range, and a bunch of other things that CD simply doesn't have. Well, CD does have it if you make a digital recording from your vinyl, but that's another matter. Frankly, people who heard the "same" two recordings on vinyl and CD literally heard two distinct and different presentations, with CD quite potentially revealing information that wasn't audible on vinyl. Not everyone liked this and frankly invented all sorts of techno and pseudo-babble to rationalize their preference. Further, if these phase characteristics were so objectionable, why are they not objectionable in the following scenarios?

a) Examine some of the phase errors in many tube based amplifiers.
b) Consider the phase errors that are introduced in the recording process.
c) Phase errors are rather signficant with phono cartridges.

In a general sense HIPAR, I do though agree that some early, non-oversampling had issues with both the anti-aliasing and anti-imaging filters. Given the inherent variances in manufacturing, desire to saturate the market, and ignorance, spending more on or designing technically better performing analog based filters didn't always happen. Moving to oversampling and allowing the bulk of the work to be done in the digital domain made life easier and costs cheaper.

A couple of decades is a long time though. Might not be a bad idea though to hold on to older units given the way DRM is being pushed by the recording industry and legislated by those who need 40 pieces of silver.

During the last 25 years there have been many attempts to quantify what's important for digital sound reproduction. Is it:

a) Is it the number of bits for the sample?
b) The linearity of the A to D - D to A conversion process?
c) Bit error rate
d) The reconstruction filter?
e) Clock jitter?
f) The quality of components?
g) Electolytic capacitors in the signal path?
h) Power supply robustness?

All of these have been throughly investigated and the still debate rages on. The technicians and listeners can't seem to agree and I've been involved with both approaches.

Here's another 'way back when' digital audio test story that I would like to redo with 2007 technology:

Vanguard records once produced a vinyl test record recording the Haydn Military Symphony. The output from the mixing console was divided into two equal paths. One path went to an analog tape recorder and the other path went to a digital recorder. One side of the record was mastered from the digital tape and the other side was mastered from the analog tape. According to the liner notes, great care was used to operate the cutting lathe assuring equal quality pressings for each side of the record. So this was a test of analog vs digital recording rather than CD vs vinyl.

The audio society gathered to determine who could hear a difference between analog and digital recorded audio. No one was allowed to hear the record before the test began. The turntable was shielded from view and operated by a disinterested party. The same passage from each side was played and the panel was asked to vote on which they liked better. Almost everyone (much better than a 50/50 guess) chose the analog mastered material. The results were announced.

Next, several passages were played with the record side chosen randomly by rolling dice. Again the listening panel overwhelmingly chose the correct answer. I scored 100%!

No .. I couldn't hear tape hiss! I was surprised that a phonograph system has the resolution required to demonstrate this.

In a way I'm sorry I participated because I have had an anti-digital bias since with an attitude that digital masters are inherently flawed. It follows that both CDs and vinyl records mastered digitally are flawed. Unfortunately, in todays digitally crazed world, that's almost everything.

I read a more recent story on the matter that contends listeners can hear a difference until the sampling rate is increased to 100 kHz with a bit depth of 24 bits. I don't know how that test was conducted.

--- CHAS

No, it's also a comparison of playback approaches. Have you ever taken a record and carefully prepared a CD for comparison?

They do.

You wish.

The answer to the question is that because CD technology has actually yet to peak. One of the issues with music as opposed to pure data, is that while it's all 1s and 0s, time becomes relevant as to when those 1s or 0s turn into other 1s or 0s; any vibrations, etc. can throw off the proper retrieval of those 1s and 0s from a time variable perspective; essentially, we're talking about jitter. Minimizing jitter is a problem that continues to plague CD manufacturers and is one of the reasons that not all CD players sound the same. Further, the quality of the components inside the CD player matter, like which DACs are used, power supply isolation to minimize jitter and noise, etc. Also, different CD players employ differing algorithms to reconstruct the data pulled off the CD (i.e. different techniques of upsampling aka re-sampling).

Not really. I generally only wish for things that are not already the case.

Keep dreaming.

You and your alter ego can keep running around these forums spamming every thread putting down things you can't afford; I guess if you guys try hard enough you'll finally be able to convince each other that there is no difference

To the OP, trust your ears, not a bunch of posters in denial because they can't afford superior equipment.

If superior equipment means equipment that's prone to any sort of mechanical or air-borne vibration, I'll pass. You really should've taken me up on the offer for that jitter paper by Benjamin & Gannon of Dolby Labs. Then you could chase something else.

A pretty silly idea, considering that I have upwards of $40,000 worth of CDs (not to mention an additional $20,000 worth of DVDs.)

What did you do, sign up for the Columbia House CD and DVD clubs under a bunch of names, like you do here?

I've said it before and I say it again. After twenty plus years of engineering and alchemy, everyone has run out of options to make the overall digital reproduction process using 16 bit samples and 44.1 khz sampling work properly. It's a bit starved media that was introduced prematurely before the the state of the digital recording art allowed things to be done correctly.

a) 16 Bits is not enough to reproduce the lower levels of the composite waveform. A log A-D transfer function that assigns more bits to lower levels might have helped

b) 44.1 kHz is too close to the Nyquist sampling criteria to allow a realistic implementation of the reconstruction filter for frequencies above about 10 khz.

I have commented earlier on the bit recovery myth. Also I listed the technical areas that enthusiasts have claimed to make a difference. I have delved into some; without notacable improvement in performance.

Do you all know the Von Karajan story? I think Sony and Phillips did the best they could with their constraints during the late 70's. Now we are stuck with a legacy system.

Rappers, rockers and popists are all perfectly happy with CD quality sound. I have found that orchestral recordings break it. This is tyranny of the majority because the new digital audio formats have failed in the mass marketplace.

The problem with digital quality anything is that no one complains about it anymore.

--- CHAS

Why don't you PM the people you hallucinate are me and tell them so. I bet you will get some interesting responses (maybe including finally being put in your place as you so richly deserve.)
I buy my CDs in stores. I hate Columbia House and other clubs.
I chose to only spend $2500 on my stereo because I got the sound I wanted for that, and paying more to get allegedly "superior" stuff would give me no benefit and be a waste of money. It has nothing to do with affording it.

Why do you hate these clubs?

To everyone else: did anybody ask Pulliamm how much he spent on his system? I didn't.

He must be feeling inadequate........again.

HIPAR...

a) I'm unclear as to what you're saying here. Is this a dynamic range thing you're looking at or something else?

b) Why? What about oversampling?

As far as the von Karajan story, rumor has it that he said a CD must be long enough to contain his performance of Beethoven'ts 9th symphony. To think though that a bunch of people consulted him before doing their own work on the medium is preposterous although I'd think he was fond of blowing smoke up everyone's arse. It's one of those legends. You know, like the moon landings were all faked and pictures of ghosts.

IMO, if anyone is to blame, it's the recording engineers and the nature of much of the popular music as you pointed out. If hi-rez in whatever format sounded better, I think it's because more care and constraints were placed upon them. 'Course, the engineers don't blame themselves. They're artists.

They didn't ask, they assumed (as always.) Now they can stop assuming.

1) Except on those rare occasions when I like their selection, I have to send the card in to cancel it or they send it anyway.
2) Almost all their stuff is mainstream/top 40. Hardly any of the obscure/indie stuff I like.
3) I sent BMG two payments on one check. They only credited me for one, and hounded me for something like 3 years (long after I had cancelled.)

I'm willing to bet it would sound different. For better or worse, each individual would have to decide for their self.

I personally have a substantial number of CDs ripped losslessly to a 500 gig external hardrive. I then hook my laptop to my system and wirelessly stream the music from my PC with the attached external hardrive. So far it has sounded decent.

To be honest though I usually use this arrangement for convenience, such as a party when I want to create a large playlist.

When I want to sit back and do some critical listening from the sweet spot I have always continued to use my player, a Sony CDP-CA70ES.

Thanks for the reply. Sarcasm evidently escapes you though.

Chu Gai

The bit arguement goes something like this:

Since the A-D uses a linear transfer function, a small amplitude sine wave is represented by just a few bit steps from the converter. A maximum amplitude sine wave gets all of the bits so larger amplitude waves are better defined by more bits. On the D-A end of the chain, if you observed the voltage envelope the maximum amplitude sine wave would be better defined before analog reconstruction. The problem is exasperated as one approaches the Nyquist cutoff frequency because the wave also has fewer samples per cycle.

A scheme was presented to even out the encoding-decoding process by using a log transfer function to devote more bits to lower amplitudes at the expense of bits and poorer resolution to higher amplitudes. Both systems would exhibit 100+db dynamic range.

I don't know if it was ever tried or if it would sound better. But it certainly could not be implemented after the fact.

My take on oversampling:

I don't believe that oversampling actually creates data that isn't there. It's not like oversampling can magically make source material sampled at 44.1 kHz identical to what it would be if it were originally sampled at 88.2 kHz.

Jacob

I have never been able to verify that real time recovery of the bit stream has ever been a problem. Also note that the CD digital bit stream is not a lossy compressed format so none of the source bits are lost during the transfer process. This is in contrast to MP3 or MPEG.

On recording:

I'm a believer that the best sounding orchestral recordings were produced back around 1960. They used minimal mike configurations. Non of those huge mixing boards you see today; just a few recorded tracks with very little artsy interpretation by an 'engineer'. They even sound presentable after transcription to CD format. I have a few reissues from Chesky Records (I once meet the Chesky brothers) that demonstrate what I'm contending. Also check out reissues of Mercury Living Presence.

Most pop recordings today have nothing to do with realistic reproduction. Watch videos of recording sessions on MTV. They are always singing to prerecorded tracks while listening through headphones. Then someone manipulates 32 tracks to get the best 'sound'.

--- CHAS

The one good thing about things digital is that you can move the bits around without error. Going from a CD drive to a hard drive and then through WiFi to a remote WiFi D to A should provide (almost) the same sound as listening directly from the Sony player.

You can theoretically send the bit stream through the internet to NASA, have them send it 200 Million miles to Mars and play it back there without loss of fidelity.

Also you cannot hear the difference between the original and 100th generation copy of a digital recording.

--- CHAS

It is time to stop that silly idea. Makes you look foolish. Or, that is a desired effect one is seeking.

So you got a freebie, so what.