I don't have this fully formulated in my head yet. But I'm going to take a stab at explaining what I'm thinking, and I'd like to see if anyone has any detailed information to either support or refute this idea. I've been ruminating over this one for more than a month now, and it's time I got some other opinions. I know that there is at least one person on this BBS who knows this subject in excruciating detail, and I'd like to see if I can get other expert input on this.

Warning: This can be a rather deep subject, and I tend to get wordy. So this is going to be a long post. It'll take me a while to build up to my actual question- bear with me.


Definitions:

To make this discussion more cohesive, I'd like to get a few conventions out of the way.

- Digital Audio Extraction (DAE), nicknamed ripping, is, for this discussion, defined as the process of using a PC's CD-ROM drive to directly extract the 16-bit data from an audio CD and save it on the hard disk as a 16-bit PCM .WAV file.

- Encoding, or data-compressing that resulting .WAV file into an MP3 file, is a separate process, and for this discussion, is irrelevant. My question is about the actual extraction process and the resulting uncompressed bytes of audio data. MP3 data compression is not what I'm asking about and I'd like to leave it completely out of this discussion.

- Jitter Correction has two common meanings, one of which is technically incorrect. The incorrect meaning is more accurately called "Sector Synchronization", and refers to error-compensating a CD-ROM drive that can't do proper DAE. For the purposes of this discussion, I would like to assume that we're talking about a drive that can do a perfect rip without errors and does not need sector synchronization. However, the true meaning of jitter correction (the time-base correction of individual samples at playback time) might be germane to this discussion, so for this discussion let's use the correct meaning for the term.

Okay, with the definitions out of the way, on with the discussion.


The DAC Process:

When you play a CD, there has to be at least one digital-to-analog (DAC) conversion somewhere along the trip between the surface of the CD and the speaker cone which makes the sounds.

In the case of playing a .WAV file, the DAC process is done by the computer's sound card. When you ripped that .WAV file, there was no DAC involved: the CD-ROM drive read the bytes exactly as-is and turned them directly into a .WAV file without any intermediate conversion. The analog conversion only happens at playback time. Since a CD is a lossless medium (and since, for this discussion, we're assuming that no errors were induced at the ripping stage), the .WAV file should be bit-for-bit identical to the audio data that the mastering facility had when it pressed the CD.

In the case of listening to an audio CD directly, without ripping, the DAC is performed by the circuitry on the CD drive. If it's your computer's CD-ROM doing the playback, then there's a little audio-CD DAC circuit which turns the bits into analog audio and sends them down the little internal stereo cable into your sound card's mixer. Yes, the sound is coming out of your sound card, but the sound isn't being generated by the sound card, it's being generated by the DAC circuitry on the CD-ROM drive. The sound card just has a little analog mixer controlling how loud the CD's volume is.

In the case of listening to an audio CD on a consumer audio CD player, the DAC is performed by the audio CD player and sent out of the stereo line outputs to your amplifier. Same as if you listen to an audio CD on your computer (although probably using a different brand/model/quality of DAC circuit).


Jargon:

Now, when you purchase a consumer audio CD player, you often see jargon on the box stating things like "1-bit" or "20x oversampling" or some other set of terms. These refer to the method that the audio CD player's DAC circuitry uses to convert the bytes into sound. I only have a vague grasp of what these methods are, and that's the first place I'm going to need help.


How much do the DACs shape the sound?

It seems to me that the single most critical part of any digital audio playback is the DAC circuitry. This is responsible for the character of the music which comes out of the player.

Do consumer audio CD players deliberately use techniques which shape the sound a little bit? For instance, removing harshness or high-frequency noise? How do they do that? I know that the answer is "oversampling" but I'm a little fuzzy on the details, help me out here.

I also know that different CD players each do this a little differently. Some do it better than others, and those are the ones cherished by audiophiles. Note that this is not so much a question of the "quality" of the DACs, it's a question of what techniques do the DACs employ when turning the bytes into audio.


Then what about PC playback?

Well then, do PC sound card DACs use the same techniques as consumer CD player DACs to shape the sound, or do they just pump the bytes out as-is?

And for that matter, what about the Empeg's circuits? Do they behave like a consumer CD player or a PC sound card?


The real question:

If a computer's sound card uses different techniques in the DAC process than a consumer CD audio player does, then it would follow that a ripped .WAV could sound significantly different than the original CD did. The actual bytes of data are no different, just the playback circuitry.

So how much different could it sound? Just how important is this?


The implication for MP3:

So if the source .WAV file sounds different than the consumer CD player does, then how can we expect the resulting MP3 to sound as good?

We would have to make sure that the WAV or MP3 playback circuits are doing the same DAC techniques to shape the sound as the best consumer audio CD players already do. Are they? And if not, is there any processing we could do to the WAV file which would emulate what the CD player's DACs are doing?

___________
Tony Fabris

surely if DACs on CD players did some sound shaping they'd all have to do the same thing, otherwise CDs would sound different from player to player.

As I understand it, that's exactly the case: CDs do sound different from player to player because of different DAC implementations.

From a logic POV, it just doesn't make sense for the DACs to do any shaping themselves.

Well, actually, the DACs have to shape the sound to a certain extent. A 16-bit 44.1khz wave sample is just an approximation of a real wave. How the DACs fill in the "gaps" between the sample points is at the root of the question.

___________
Tony Fabris

Tony,

You've brought up a lot of technical issues that really can't be answered in a short post. However, if I understand correctly, your fundamental question is this: Does a WAV file, perfectly ripped from a CD, sound different than the CD itself?

The answer is no.

Certainly, if I play the identical CD on two different CD players or two different stereo systems, they are going to sound differnt, due to differences in the quality of the D/A conversion and the analog electronics that follow it.

However, if I pull the 16-bit samples from an audio CD, and then I burn a CD with those very samples, then the duplicate will sound identical to the original when played on an identical system.

Likewise, if I played the original CD on a CD player with an external D/A converter, and I played the ripped .WAV file through my Soundblaster Live! connected to that same D/A converter, they should sound identical.

The ripping process does not alter the sound quality at all, unless errors have been introduced in the data. All of the audible differences between a .WAV and a CD are due to the quality of the DAC and the subsequent analog components, and that's it.

Of course, the compression process does change the sound, but we all know that.

Michael Grant
12GB Green
080000266

Well then, do PC sound card DACs use the same techniques as consumer CD player DACs to shape the sound, or do they just pump the bytes out as-is?

Actually, this may be your point of confusion here, Tony. No CD player can just "pump out the bytes as-is." The digital-to-analog conversion is not a straightforward process, and no CD player or other digital audio device can do it perfectly. All of the terms like "18-bit D/As", "8 times oversampling", etc. are just different techniques used to mitigate the distorting effects of imperfect D/A conversion. They all make different compromises, and play different tricks, in the effort to produce the best possible output at their target price point.

Your sound card, your Empeg, and your CD player all perform fundamentally the same task, but with different levels of quality. Comparing these three products from an audio quality standpoint is no different than comparing the audio quality of three "standard" pieces of stereo equipment.


Michael Grant
12GB Green
080000266

The ripping process does not alter the sound quality at all, unless errors have been introduced in the data. All of the audible differences between a .WAV and a CD are due to the quality of the DAC and the subsequent analog components, and that's it.

Right. Exactly. I'm aware of everything you just said. But this cuts right to the issue I'm trying to get at (and probably didn't word it correctly before):

The jargon-names that you see on consumer audio CD players... all of those different DAC techniques have various advantages and disadvantages. As I understand it, the ones that are most prized by audiophiles are the ones that make the CD sound its smoothest and most natural, with the least amount of high-frequency noise, or the least amount of jitter, or the least amount of harshness.

So when we're playing back our MP3s or WAVs on our PC sound cards or our car MP3 players, are these devices using those same, audiophile-pleasing techniques to turn the data into sound?

___________
Tony Fabris

all of those different DAC techniques have various advantages and disadvantages. As I understand it, the ones that are most prized by audiophiles are the ones that make the CD sound its smoothest and most natural, with the least amount of high-frequency noise, or the least amount of jitter, or the least amount of harshness.

Yep, you're right... But it's important to keep in mind that higher quality components can often go a lot farther for improving sound quality than many of these "tricks" can. For example, the initial reason people started using the "1-bit D/A" solution was cost: it was cheaper to design a decent sounding 1-bit D/A than an equally decent 16-bit D/A. And yet, when we get up to audiophile quality, the 1-bit D/A solution may very well have other limits (like clock jitter, for example) that remove it from consideration. (Don't tell these guys that, though; check out their "Millenium" digital amp.)

Regardless, in the end high-end CD players and external D/A converters do indeed sound better than a $200 Fry's special, on average; and that's largely due to the higher-quality D/A and analog stages.

So when we're playing back our MP3s or WAVs on our PC sound cards or our car MP3 players, are these devices using those same, audiophile-pleasing techniques to turn the data into sound?

Sure they are. The rule is the same for sound cards as it is for CD players: cheap sound cards sound like crap, and you may have to spend some $$$ to get excellent sound quality. Sound cards might have some additional issues to deal with, too, because of interference caused by the surrounding electronics.

I'd say that the average PC sound card is not audiophile quality. The Soundblaster Live is generally well-regarded. Here's a rather extensive analysis of the audio quality of various PC sound cards.

As for the Empeg, well, I don't think I'm quite sure what components they're using. Perhaps one of the empeg boys can address that.






Michael Grant
12GB Green
080000266

The DACs are internal to the DSP we use; the DSP is a custom in-car part (16 bit sign-magnitude, fourfold oversampling digital filters with noise shaping), and we use good quality components in the audio path (all tantalum caps in the mk2, giving better stability) - and things like the burr-brown amps for the 4v outputs.

John did some analysis on the empeg's audio output and the frequency response was pleasingly flat - I think it turned out to be better than the SB live.

Really though, the empeg just sounds good to our collective ears. We could fit some nice BB dacs too, but we're very happy with how it sounds as-is.

Hugo

(16 bit sign-magnitude, fourfold oversampling digital filters with noise shaping)

Cool- that's the sort of information I was interested in. Now, like I said, I'm unclear on what this jargon means exactly. I have read some articles about it and I have a fuzzy understanding of it. But how does this compare to the DAC conversion that consumer CD players do? Is is exactly the same as your average consumer CD player or does it differ?

___________
Tony Fabris

I have no idea what most of it means :)

DAC technology evolves over time; some companies prefer certain methodologies (1-bit - bitstream, was very popular when I bought my home CD player in around 1990, and Philips like it). "Noise shaping" was something which used to appear on a lot of panasonic gear, but I have no idea what it means. Oversampling just means outputting the DAC value multiple times to average across multiple sets (though how this would make a difference at the sample/hold stage in such a short time period, I don't know).

Generally, PC cards sound bad simply because they have cheap DACs, cheap DAC PSUs (not generally independent ones from the "dirty" digital PSU), cheap PCBs (not multilayer usually), cheap audio path components and cheap connectors. You get the same effect with very cheap consumer CD players (like the ones which cost $60). When you get to $150+, you're no longer just paying for a decent CD mechanism that doesn't skip for a pastime and you start paying to improve the PSU, improve the DACs, get some decent capacitors in there - that sort of thing.

The empeg has independent linear PSUs to drive the audio section, a good multilayer PCB, and good components: we didn't cut any corners. That's not to say we couldn't have made a more expensive audio output section (by putting offchip DACs in there) but we believe the output is about as good as it gets from the DACs we use. In short, the empeg audio section was designed as a consumer audio part, not as an afterthought on a computer - as most laptop audio circuitry is.

Hugo

Oversampling just means outputting the DAC value multiple times to average across multiple sets (though how this would make a difference at the sample/hold stage in such a short time period, I don't know).

That's not quite how oversampling works. In order to do 4x oversampling, 3 zeros are inserted between between each consecutive pair of samples. Then, the result is low-pass filtered to smooth it all out. If your filter is designed right, the original samples are usually untouched, and the 3 intermediate zeros are replaced with a combination of several of the nearest samples. (That's a conceptual description of it; in practice, the calculations are combined together and reduced so that the three "zeros" are never actually generated.)

Now the nice thing is that this low-pass filtering is a necessary part of the D/A process, in order to prevent aliasing distortion. Doing some of that filtering digitally (i.e., before the D/A) enables you to use a less aggressive analog filter stage to finish the job. Doing all the filtering in analog (i.e., after the D/A) often introduces magnitude and phase distortion that you just don't want. The digital filter response can be tailored much more exactly than an analog filter can.

I believe that "noise shaping" is related to the digital filtering process. Due to the way the digital filtering is performed, the 3 intermediate samples are likely calculated out to 24-32 bits. Now they have to be chopped down to 16 bits in order to send them out of the D/A, so that is going to introduce some noise into the signal. "Noise shaping" is a way to control the frequency content of that noise. You really can't change how much noise that you're introducing; however, you can shape it so that most of the energy sits in frequency bands that you can't hear very well anyway. (Or for that matter, since the D/A converters are running at 176.4kHz, you can shape much of the noise to sit in frequencies that you can't hear at all.) In a sense that fits in real well with the MP3 concept...


Michael Grant
12GB Green
080000266

Edited by mcgrant on 6/9/00 00:49 AM.

I have no idea what most of it means (...) (1-bit - bitstream, was very popular when I bought my home CD player in around 1990, and Philips like it). "Noise shaping" was something which used to appear on a lot of panasonic gear, but I have no idea what it means.

Right, that's the sort of thing I'm asking about. What's the difference between these technologies, and, from an audiophile point of view, how important are they?

You sounded ever-so-slightly defensive in that last post, Hugo. Please understand I wasn't trying to cast aspersions upon the Empeg's output stage. I'm just trying to (a) understand the differences between various DAC implementations, and (b) find out what can be heard, if anything, as differences between the various implementations.

___________
Tony Fabris

OK, I'm not a great expert but I've read a bit about how these things work, so here are my (partial) definitions of those terms that are confusing you. Some of my answer is taken from the rec.audio.pro FAQ at http://recordist.com/rap-faq/current.

First, to get our knowledge base even, the DA conversion stage takes a clock signal, and on each clock the input from the digital side is converted to an analog voltage. How does it do this? Most converters work by storing chunks of electrical charge in capacitors and then emptying the appropriate capacitors into the output stream on the clock signal. CD audio is two 16-bit words at a rate of 44.1KHz (44100 samples per second).

Now, there are a couple of ways to do this conversion. One is to take your (sixteen-bit) lump of data as one chunk (sixteen inputs) and have sixteen capacitors with 1, 2, 4, 8, and so on up to 32,768 'units' of charge (where a unit is 1/4,096V - i.e the largest bucket holds 2 volts - all of them summed together produce 4 volts).

Now the largest capacitor in this series has to have a tolerance of about 0.01% so it doesn't end up holding 32,126 or 40,000 units of charge, which makes it rather expensive to manufacture. However, you can treat a stream of 16 bit chunks coming at you at 44.1KHz as a stream of single bits coming at you at 11.3MHz. This is what a 1-bit DAC does, essentially. I'm not exactly sure how it works from there - the FAQ doesn't say much - but I thought it worked like a kind of finite state machine, so that after sixteen clocks at 11.3MHz the signal would have 'bounced around' inside the chip enough to get the right voltage on the analog out. But I'll research this further.

Each clock cycle, the output is forced to be some arbitrary voltage, which may not (in the case of very loud percussive hits) be very close to the last voltage output. Don't forget here that it's a two-way street - whatever's on the end of the analog output has its own ideas of electrical conductance, capacitance and inductance. So the DAC does its best to drag the signal on its output stage to where it should be, which may make the signal jerk around a bit.

Think of it like an analog clock with hands on a ratchet mechanism - at 2:59 the hour hand is still pointing at the 2, and one minute later it's pointing to the 3. As the last second ticks over the clock mechanism has to move that hand fairly quickly, so it might jerk a bit when starting to move and shudder a bit before it comes to rest. This sort of thing does happen (in a very minute way) on the audio line.

Noise shaping is basically the process of filtering the jerks and shudders out of the signal that comes out, so it looks (relatively) smooth to the receiver of the analog signal. MGrant's explanation is much better than this, but I'm sticking to it because it may help the people who didn't read the R.A.P. FAQ.

Oversampling can refer to two things, here. Firstly, it could mean reading the bits off the CD a couple of times and working out your best guess of what they actually are. The audio is encoded with an error-correcting system here, which is designed not so much to eliminate the possibility of errors as to work out a good guess of what the value might have been when you do get a bad byte. By re-reading it a couple of times, the microscopic wobble in the CD that caused the phase change that caused the misread might have gone now.

On the other hand, it can mean that you try several different analog outputs of the same digital input and average them to work out the most likely original signal. After all, those capacitors can sometimes not be quite filled, or the clock might have been slightly skewed and have hit the output latch (the bit that tells the converter stage to send on its results to the output line) before the rest of the chip was ready. So oversampling can work here as well.

I'll get to the bottom of 1-bit DACs and get back to you.

Save the whales. Feed the hungry. Free the mallocs.

Edited by PaulWay on 6/9/00 02:52 AM.

This is great info. Just the sort of thing I was looking for. I've actually got a pretty good handle on what oversampling is, but I know I'd just embarass myself if I tried to explain it so I'm hoping someone else can jump in with a proper explanation.

I'm wondering what audible differences you'd find between a 1-bit DAC with oversampling as compared to 16-bit DAC using some other techniques I don't completely understand.

___________
Tony Fabris

Paul,

If you have an imperfect D/A converter, oversampling can help some---providing, say, 3dB more SNR every time you double the sample rate. But generally speaking, oversampling is primarily used to simplify the filtering process that has to occur after D/A conversion. The FAQ that you posted has a good explanation of this.

In order to mitigate the effects of an imperfect D/A converter, the two most common approaches are
1) to 'overspec' the part---for example, to use an 18- or 20-bit converter even when only 16 is required. In my work (not audio), we have a 14-bit A/D converter that we expect to see 12 "effective" bits from; a 12-bit D/A that we expect 11 "effective" bits from; etc.
2) to go completely the other direction, and use the 'reduced' bit techniques (like 1-bit sigma/delta or delta/sigma D/As) and crank up the sample rate accordingly.

One very good point that FAQ makes is that any of these techniques can be employed in a lousy product. I think the 1-bit techniques were first employed to make cheaper products; but now the Sony Super Audio CD (SACD) format uses a highly oversampled 1-bit storage format. Weird.

For another example of a high-quality 1-bit solution, check out the TACT Millennium digital amplifier. Basically, it uses a high-power 1-bit D/A converter for each audio channel. There are no analog amplification stages! Increasing the volume simply changes the voltage levels for the 1s and 0s. In the end, most of the quantization noise is shaped way up into the inaudible range, and a very gentle low-pass filter is all that is needed to clean up the signal. WOW! (OK I'm a nerd, I like that stuff)


Michael Grant
12GB Green
080000266

However, you can treat a stream of 16 bit chunks coming at you at 44.1KHz as a stream of single bits coming at you at 11.3MHz. This is what a 1-bit DAC does, essentially. I'm not exactly sure how it works from there

The 44.1KHz stream of 16 bit samples (pulse code modulated, PCM), gets converted into a digital stream which is high (logic 1) and low (logic 0) for lengths of time in proportion to each sample (effectively, pulse width modulated, PWM).

So when one of your 44.1KHz samples is, say, 15000, then during that particular 1/44100 of a second the bitstream will be high for 15000/65536 of the time and low for, er, 50536/65536 of the time.

Then you get this bitstream to dump, or not, the charge from a single capacitor onto the output. As there's only one capacitor, you don't need to get the range of very accurately spaced capacitor values of the "traditional" solution.

The resulting signal is, of course, very very noisy at high frequencies, but you just filter off everything above audio frequency.

Peter

The 44.1KHz stream of 16 bit samples (pulse code modulated, PCM), gets converted into a digital stream which is high (logic 1) and low (logic 0) for lengths of time in proportion to each sample (effectively, pulse width modulated, PWM). So when one of your 44.1KHz samples is, say, 15000, then during that particular 1/44100 of a second the bitstream will be high for 15000/65536 of the time and low for, er, 50536/65536 of the time.

Yes, if you're willing to do 65536x oversampling, that would be fine. The only problem is that your new sampling rate is 65536 x 44.1kHz = 2.9GHz! Yes, that's billions of samples per second. I don't think there's a single audio system that actually does that.

For example, the TACT 1-bit digital amplifier uses a master clockrate of 90.3kHz, which is only 2048x oversampling. So you would think that you could only get 11-bit precision out of that. In fact, it achieves >110dB SNR and >130dB dynamic range over the audible frequencies, which is more like 20 or 21 bits of precision if my calculations are right. (Of course, a 16-bit CD can't even use all that precision.) What the heck!?

That's exactly the non-intuitive---and frankly, amazing---part about noise shaping. If you measured the noise across the entire frequency range (0-45.2 MHz), you'll find precisely the amount of total noise power that you would expect to see from a 1-bit quantizer: a lot. But by controlling very precisely when the unit switches back and forth between the 0s and 1s, the effect is to shape the frequency response of that noise---so that most of it lies well above the audible range, and is easily filtered out.

You can begin to see this with your 65536x oversampling example. The easiest example is right in the middle of the range---0.5. Obviously I should send out half 1s and half 0s to get that. If I send out 32768 1s followed by 32768 0s, I'll see noise power starting at 44.1kHz---still inaudible, but I'll still need a very strong analog filter to attenuate it. On the other hand, if I send out 1,0,1,0,1,0 instead, I've shifted the noise power all the way up to the gigahertz range. No problem filtering that out!

So that's why 1-bit D/A's don't actually have to go all the way up to 65536x oversampling in order to achieve 16-bit precision in the audible range. I seem to remember some units getting away with only 256x oversampling.

By the way, I have no connection to TACT, nor have I even listened to one, I just think the technology is cool.

Michael Grant
12GB Green
080000266

That's exactly the non-intuitive---and frankly, amazing---part about noise shaping. If you measured the noise across the entire frequency range (0-45.2 MHz), you'll find precisely the amount of total noise power that you would expect to see from a 1-bit quantizer: a lot. But by controlling very precisely when the unit switches back and forth between the 0s and 1s, the effect is to shape the frequency response of that noise---so that most of it lies well above the audible range, and is easily filtered out.

Ah, now that's the bit I never knew (I designed one of these as a degree project -- it didn't get built 'cos it was too many gates for the FPGA, but it wouldn't have made a 44.1kHz DAC anyway as the chip was only clocked at a few MHz).

Good scheme this, post all sorts of nonsense on the BBS and have people who actually know about stuff point out the places where I don't. Cool!

So your 2048x oversampling DAC basically has a lookup table of 65536 2048-bit patterns, one for each possible PCM input, and just shifts out the right one at 90Mhz? And the cleverness is in designing the 2048-bit patterns?

Peter

Found some relevant articles:

http://www.audiomedia.com/archive/features/us-0999/us-0999-dac/us-0999-dac.htm
http://www.audiomedia.com/archive/features/us-1099/us-1099-da2/us-1099-da2.htm

Hugo

So your 2048x oversampling DAC basically has a lookup table of 65536 2048-bit patterns, one for each possible PCM input, and just shifts out the right one at 90Mhz? And the cleverness is in designing the 2048-bit patterns?

No, actually, that wouldn't work either---you'd still be limited to 2048 voltage levels. The patterns are not simple functions of the input level at all. Instead, these devices use nonlinear feedback within themselves to figure out when to switch back and forth.

The technology is called "sigma/delta modulation". The confusing part is that, in a sense, you're doing analog to digital conversion---because you're turning a higher-bit signal into a 1-bit signal. Here you can find a few diagrams that might prove helpful, but this introduction is very incomplete. Here is an on-line paper called "A Fundamental Introduction to the Compact Disc Player" which discusses many aspects of digital audio, including MASH and sigma-delta modulation. I'm sure a search of "sigma delta modulation" on various search engines would dig up more.


Michael Grant
12GB Green
080000266

Thanks, Hugo. Cool articles. They do answer some of my questions.

___________
Tony Fabris

Thanks, Michael and Peter, for your explanations. As I said, I don't understand it perfectly. You've filled in the gaps in my knowledge and enlightened everyone else too. Thanks!

I'm assuming this is going into a FAQ somewhere, Tony?

Save the whales. Feed the hungry. Free the mallocs.

I'm assuming this is going into a FAQ somewhere, Tony?

I dunno if it falls under the umbrella of an Empeg FAQ...

And I'm still not sure if I've gotten to the root of my original question yet, which is: How different would these different DAC implementations sound? How different can I expect a ripped WAV file played on my PC to sound from a consumer CD player? And in exactly what ways would it sound different? I've seen a lot of technical details of what the implementations do (which is great, don't get me wrong, that was part of my question), but nothing so far that says "Compared to a 16-bit DAC, a 1-bit DAC will sound yada yada yada..."

I think one of my main questions has been answered which is: Does a consumer CD player attempt to smooth out the sound at all? In other words, does it attempt to alter the sound a little bit, rather than just reproducing the bits as accurately as possible? I think the answer to that question is no: A CD player tries to reproduce the bits with as much detail and resolution as possible, without any deliberate shaping or smoothing of the output.

I was trying to get at the root of a nagging feeling that ripped files sound harsher than the original CDs. But I have no common frame of reference equipment on which to play both pieces. Once it's in WAV format, the CD player won't play it. And if I burn it back onto a CD, then it's just the identical thing again. I was thinking that consumer CD players might be discarding certain data that the ripped version doesn't. From what I've read so far, this doesn't seem to be the case, so any harshness I'm noticing is the result of the equalization, accuracy, and response curves of the playback equipment. If this is true, then anything (other than compression artifacts) that I'm hearing in a rip should be correctable by careful equalization.

Still, definitely keep the information on the DACs coming, this is great stuff.

___________
Tony Fabris

If this is true, then anything (other than compression artifacts) that I'm hearing in a rip should be correctable by careful equalization.

Well, that's true within limits. If your MP3s sound bad because they are too compressed, you can always reduce the compression level, and increase your hard disk capacity to compensate. But if your MP3s sound bad because your sound card sucks, there's not much you can do about that.

One particular problem in sound cards, from my understanding, is "harmonic distortion." You may have seen measures for "total harmonic distortion" in the specs for stereo equipment. Harmonic distortion occurs due to nonlinearities in your analog chain (or the D/A). If you feed a 1kHz sine wave into your sound card, harmonic distortion will show up as unwanted components at 2kHz, 3kHz, 4kHz, etc. Well, the problem with harmonic distortion is that it can't be eliminated by equalization. Once it's there, it's there.

Needless to say, one of the things you pay for in more expensive audio equipment is a design that minimizes noise, provides a flat frequency response, and minimizes nonlinear distortion. And that is really the important point: evaluate your sound card just like you would evaluate a piece of stereo equipment. That may mean considering the numbers: SNR, THD, dynamic range, and so forth. That may meen going on recommendations of computer audiophiles you trust. But that may mean listening, since quality is so often in the ear of the beholder. If possible, listen to a sound card connected through your favorite stereo, playing a variety of music, before you commit to buying it. I admit that's quite hard to do without a solid return policy from the computer store.

So when you say your .WAV files sound harsh compared to your CD player, I'd definitely finger your sound card as the culprit---if, of course, you're feeding your sound card's output to the same stereo system as the CD player.







Michael Grant
12GB Green
080000266

Does a consumer CD player attempt to smooth out the sound at all? In other words, does it attempt to alter the sound a little bit, rather than just reproducing the bits as accurately as possible? I think the answer to that question is no:

I think I have to disagree with you here, Tony.

In my limited and entirely empirical experience, there is a really big difference from one CD player to the next, with everything else (i.e., same CD, same amps, same speakers, same listening environment) being equal.

I have now had four different head units in my car: A Sony CD player, an expensive Panasonic CD player, an inexpensive Panasonic CD player, and the Mark 2 empeg. Each has shown markedly different sound characteristics.

The Sony was probably the most accurate. Because of the acoustic characteristics of my setup (speaker size/brand/location, shape of the interior, etc.) it was a continual battle to get enough mid-range, to the point I had to always keep the "loud" setting on.

The expensive Panasonic was a disaster. It was one of the "gee-whiz" players, aimed at teenagers who wanted to impress their friends more than listen to music: all crispness and definition was gone, mid-bass and bass seemed to blend into an amorphous whole, and the whole effect was one of unobtrusive smoothness -- sort of Montovani-like even when playing Rush. That CD player stayed in my car for two days, to be replaced by...

The inexpensive Panasonic CD player. All the crispness, definition, and punchiness was back. Plus, so much mid-range that not only did I have to turn the "Loud" off, but I had to install L-pads to attenuate the three mid-sized speakers in the front of the car.

I may be mistaken, but I believe the primary design differences among these three units was the DACs. I'm not technically knowledgeable to discuss things like 8-bit, 16-bit, 20-bit etc., but I do believe the designers of the DACs in those three players each had their own ideas about what the output was supposed to sound like, and they implemented those ideas.

Now, the fourth unit, the empeg, is a different case altogether, because there are a few extra variables thrown in, most importantly the fact that it is reading compressed files. That aspect aside, the empeg is like the Sony, but more so. With flat equalization, my mid-range is deficient, and there is a definite harshness to the sound -- I think "strident" is a very good descriptive word. My stereo shop man says this is not surprising -- in his experience, the more signal processing that is done, the more "sterile" and bright the sound, and apparently the empeg does a lot of signal processing.

In other words, CD manufacturers do manipulate the sound output of their players according to the whims of their engineers. It is up to the end users to compensate for these manipulations with whatever tools are available -- speaker placement, amp gains, equalizers, etc. Depending on the preferences of the listener, these compensations may be large or small.

tanstaafl.

"There Ain't No Such Thing As A Free Lunch"

One thing: the empeg does no sound processing at all (apart from, well, the decompression of the sound file!) - if eq is flat & loudness is off, volume at 0dB, there is no change to the bitstream at all before it reaches the DACs.

This is one of the advantages of digital sound controls - when they're off, they're really off :)

Hugo

Hmm. What about the DSP? Doesn't that fall after the DACs in the signal chain?

(For Doug's benefit, in case he didn't know this: The Digital Signal Processor is responsible for the tone, EQ, and loudness controls. It's this very thing that allows the EQ to be so flexible.)

As I understand it, some consumer devices use a similar (perhaps even the same) DSP to shape the sound. Perhaps that's germane to this discussion? Maybe products which don't use DSPs have some sort of analog circuitry in the output stage which makes the music sound less strident?

___________
Tony Fabris

DSPs are always before the DACs when you have digital source media - it's not a lot of point getting sound to analogue when you then have to sample it to digitally process it...

Hugo

DSPs are always before the DACs when you have digital source media - it's not a lot of point getting sound to analogue when you then have to sample it to digitally process it...

(Slaps forehead) Duh, of course. Okay.

In that case, does the DSP enter into this discussion at all anyway?

___________
Tony Fabris

Hmm. What about the DSP? Doesn't that fall after the DACs in the signal chain?

All of the processing the DSP performs (tone, equalization, loudness, etc.) comes immediately before the DAC. In the case of the Empeg (if I remember correctly) the DACs in the Empeg are actually contained within the DSP chip itself. Nevertheless, they are the final step in the process before the analog processing. In many other cases, the DACs are completely separate components.

As I understand it, some consumer devices use a similar (perhaps even the same) DSP to shape the sound.

You bet. There are quite a few DSPs out there to choose from, and quite a few processing steps they could perform. In particular, many portable players provide equalization, bass boost, loudness, spatialization, etc. I even used to have one that did dynamic range compression like rjlov's volume adjusting kernel.

Home players tend not to do extra signal processing, however, leaving that task up to the preamplifier/receiver. So all of their digital signal processing tends to be limited to interpolation and noise shaping (for oversampled systems only), and maybe DAC preemphasis.

Perhaps that's germane to this discussion? Maybe products which don't use DSPs have some sort of analog circuitry in the output stage which makes the music sound less strident?

Sure, that could very well be the case. All digital music devices have some analog circuitry in the output stage that will affect the sound quality.

But it's important not to try to pin the blame on any one stage in the signal processing chain. The conversion from digital data to output voltage is a long process with many steps, each of which can contribute in different ways to distort or color the sound. When comparing two CD players, you're comparing systems, not just DACs or DSPs alone.


Michael Grant
12GB Green
080000266

The DACs are actually in the C-DSP. It is a large multi-function chip with ADCs on the input, a DSP core pre-programmed to carry out certain functions (such as implement the Equaliser in software) and 4 DACs on the output side. Everything in one.

So yeah, it does enter the discussion.

One of the few remaining Mk1 owners... #00015

Hmmm. OK, so I completely misunderstood what you were saying. [FX slaps himself about the head several times]. Of course DACS do wave shaping - in building the wave (however they decide to do it), they shape it.

I had misread your post and assumed that you were talking about the DAC acting as a DSP too (i.e. changing the wave form rather than shaping it).

Sorry again. Busy busy busy and I should be spending my spare time sleeping, not sitting infront of my home computer - I'm doing too much of that at work at the moment .

Nick.


--
18Gb blue - s/n 080000299 (original queue position 8724)