DACs and reproduced sound


I am trying to understand how to think of DACs effecting reproduced music (I am new to the hobby). I think of a DACs "role" as taking a digital input (1s and 0s) and a cleanly as possible processing that digital signal to an analogue output - goal is not impart anything on the processed data. The difference between a good and bad DAC seems like it should be on how well it does that. Or, said another way, from a review of a Benchmark DAC:

"The old regulars know exactly my position regarding the stupidity of ascribing a “character” to the sound of an utterly neutral signal path. Oohing and aahing over the vast improvement in soundstaging, front-to-back depth, bass delineation, or treble sweetness obtainable with this or that electronic component may sell high-end magazines but is totally unscientific and delusional. What the Benchmark DAC1 HDR adds to or subtracts from its input signal is borderline unmeasurable, so the sonic character of its output is obviously the sonic character of its input. It’s as simple as that. It has no sound of its own."

I sort of think of amplifiers and speakers (I am digital only listener)as being more important in "imparting" a particular musical flavor (warm, bright, etc.).

I am a bit new to the hobby so I would like any insights or be educated on DACs some more.
Ag insider logo xs@2xdangelod
Dangelod - I prefer neutral gear but others might like particular sound (warm, sweet etc.) - nothing wrong with it.

There is no "unimportant" and the signal processing might completely change the sound. Some things might be more "important" in terms of best money allocation but it is personal preference.
I think the quoted portion of your post is vacuous. Every component in the chain processes the signal it recieves, digital or analogue. Each has an analogue output which has its own sound which is not going to be an exact copy of the input signal (If you could even describe what the input signal sounded like! I believe you can't - its just a WAG unless you happen to be in a recording studio comparing two different reproduction systems with the mic feed). The most obvious differences, not necessarily all though, are the result of the design of the analogue output stage in a DAC or CDP or whatever. IMHO of course.

BTW an 'utterly neutral signal path', a straight wire, is not 'utterly neutral', or at least so say many wire enthusiasts. :-)
In simple terms, a DAC's analog section is no more or less compromised than any preamp or amplifier circuitry, except for the presence of high-frequency analog filters. In contrast, the adverse signature of the digital domain and digital/analog converter sections, results from jitter, compromises in digital filtering, and characteristics inherent in the DAC chipset. If your DAC has treble glare, hard or flat treble and upper midrange, has a smeary non-resolving quality, sounds synthetic, or becomes fatiguing, you are most likely hearing weaknesses in the digital section and digital/analog conversion. To say that "the sonic character of its output is obviously the sonic character of its input" is just Benchmark marketspeak.

Finally, there are the inherent limitations of the RBCD medium itself. Sadly much RBCD mastering is done poorly and is over-compressed, leading to some of the above symptoms regardless of playback equipment. However, with the right front end, much RBCD can approach SACD and vinyl.

Fortune cookie say that until you get right with CDP or have a good vinyl front end to compare to your CDP, you will experience much restlessness and waste of time & money swapping downstream components.
Yes, the job of the DAC is to take the 1's and 0's which are the digital building blocks needed to construct an analog waveform and construct it.

Its like building with Legos. The bits (1s and 0s) are the lego blocks. The DAC is the device that takes the blocks and does the construction (of the analog waveform needed by the amp and other components downstream).

No two DACs are exactly the same, so each will sound different, some a lot different and some not so much, but different nonetheless.

The waveform building process requires a clock to tell the DAC when the time is right to use the next set of bits/blocks available to construct the next portion of the sound wave. The clock is often provided upstream from the DAC, either by the transport or some other device that can provide the clock. Sometimes, like with upsampling DACS, like the BEnchmark, I believe, the clock needed is on board with the DAC, so the "DAC" box has everything it needs to build a waveform. REgardless, the clock + the bits/blocks plus the DAC, which determines the analog equivalent of each set of bits presented to construct the waveform, are the basics things you need to convert digital signals to analog ones.
Mapman - very good description and very helpful.

With the clock, if I have a realtively low end CDP or a sonos music streamer, each with their own internal DAC (and, therefore clocks) and I run them through the benchmark, does the benchmark "re-clock" the "onced clocked" signal, making it more accurate ?
Dangelod,

I believe that is true in the case of the Benchmark, but is not always the case in that not all DACs reclock data.

Assuming a low end player relative to the BEnchmark, or also perhaps even just a poor connection from the digital source to DAC, the BEnchmark reclocking should result in lower jitter (errors in the timing) and better accuracy and detail.
Dangelod - jitter is just only one of the elements and it is much more complicated than that. Benchmark asynchronously upsamples incoming digital data about million times to filter it in digital filter and output to 24-bit DAC at 100kHz rate. Filtering itself affects the sound and some prefer no oversampling and filtering at all (NOS DACs). Recently reviewed in Stereophile magazine Meridian CDP uses different filtering scheme and sounds much better (but costs $16k). DAC IC itself comes in different varietes to overcome resolution limitations. Benchmark is using Analog Devices Sigma-Delta DAC (similar scheme to SACD) that has 24 bit resolution (not obtainable in traditional DACs) but timing errors limit resolution anyway. Others use traditional DACs - some with statistical averaging (like DCS RingDACs) to improve resolution. Harmonic distortion play also important part. Benchmark updates DAC ,capable of 192kHz,at only 100kHz because harmonic distortions are much lower at this point. Important to remember is that there is no perfect DAC and everything is a compromise (and very complicated). Listening test will give you much more than reading spects.
I agree that high-end magazines are delusional and unscientific. Nowdays it's very commercial rather than acheiving something scientific.
DAC despite processing 1's or 0's is a complexed logical device (especially today's advanced ones).
Our todays' CD digital format has 16-bit on 'vertical Y axis' and 44.1kHz samplig freequency on 'horizontal X axis'.
Giving example of the highest range of audiable freequencies such as 20kHz will roughly have only 2...3 digital samples * 16bit will yield 36bits of overall resolution.

First DACs processed one bit at a time. The advantage of these DACs is simplicity of the design and logic. The disadvantage is high probability of error meaning to give instead of 5-bits amplitude 3 or 7 bits amplitude since the 'receiving one-bit register' only cares about signal presence or absence i.e. '0' or '1'. Hence the quality of a digital equipment was 90% depended on the analog process which is filtering and buffering garbage.

Nowdays DACs are much more advanced they process sample by sample which may contain upto 24 bits amplitude and upto 192kHz sampling freequency. The proccessing by 'word' or taking the whole sample onto the register allows to logically analyze 'condition' of a particular word hence reducing drastically the probability of error. The DAC ICs are just like jewelry can be dirt cheap and can be pricey depending on what capability they all have and certainly newer and advanced models (such as Benchmark's relocking) would run at the higher price tag but soon will be reduced as the time passes by just like with PCs.

The listening difference vs. old processors is substantial especially on higher freequency details and also dynamics. Less listening fatigue and more natural sound.

The corresponding price of a digital equipment is another point of convincing price paid vs. performance, but often it's not true and not scientifically proven where I agree with poster since the fully functional digital unit that supposingly has the best DAC IC(s) and all possible functions, read all possible high resolution formats can't be priced $10k. Hence I don't think that there's something possibly more advanced and perfect than Grace Design M902 DAC-Preamp-Headphone amp.
anyone that says dacs don't sound different are full of crap. Anyone who says the benchmark gear is neutral or adds no "color" to the reproduced sound is even more full of crap. The thing upsamples, that in and of itself changes the sound, particularly of 44.1 redbook data. And it's a bright sounding dac. though a very clean sounding dac with tons of detail.
The comment was surely about the analog preamp. A good preamp should indeed reproduce the input signal as closely as possible and good preamps should sound nearly identical, as they reproduce what they are fed. It s very hard to hear the difference between one excellent neutral preamp and another. No surprises there.

DAC's do sound different as they process the signal although differences are also very small but many people can hear these slight differences when listening critically and switching back and forth on the same track.
The Benchmark has some good things as well as some budget compromises inside. There is an active aftermarket of modifiers offering upgrades in areas like soft recovery diodes, Bybees, op amps, and chassis sheilding. Modified units have received good reviews.

http://www.odysseyaudiohk.com/cgi-bin/yabb/YaBB.pl?num=1246198439
I have a strong personal policy of not making "ad hominem" arguments, i.e., debating a point by attacking the person making the point, rather than the point itself. However, in this case it should be pointed out that the comment quoted in Dangelod's initial post was made by Peter Aczel, The Audio Critic.

Twenty-five or so years ago, Mr. Aczel's reviews routinely and intensively delved into the subtle sonic nuances and differences which he perceived as characterizing the sound of electronic audio components. In more recent years, however, he metamorphosed to the extreme opposite end of the audiophile ideological belief spectrum, maintaining very assertively that all electronic components meeting basic measurable standards of performance will sound identical.

Mr. Aczel is a very intelligent person, and he writes extremely well and very persuasively. His "day job" long ago was that of an advertising writer. But nevertheless he can often be completely wrong, as evidenced for starters by these diametrically opposed positions he has taken about electronic components over the years.

And (to Shadorne), no he was not talking about just the analog preamp section. Here is a quote from his concluding paragraph, to cite just one of many statements in the review which make that clear:

All in all, the Benchmark DAC1 HDR is damn close to a perfect piece of equipment. Neither its digital performance nor its analog performance could be meaningfully improved. That’s really all that needs to be said.

Kijanki said:

Important to remember is that there is no perfect DAC and everything is a compromise (and very complicated). Listening test will give you much more than reading specs.

I second this 100%.

To get an idea of the kinds of complexity and subtle technical factors which are involved, begin by flipping through this 39 page datasheet for a family of advanced dac chips:

http://focus.ti.com/lit/ds/symlink/pcm2707.pdf

And then take a look at these excellent whitepapers from Ayre:

http://www.ayre.com/PDF/Ayre_MP_White_Paper.pdf

http://www.ayre.com/PDF/Ayre_USB_DAC_White_Paper.pdf

And consider that the dac datasheet is just for an integrated circuit chip, and doesn't begin to address the complexities and potential design pitfalls of the surrounding circuitry. And consider that the least significant bit of just a 16-bit dac provides a resolution of 1 part in 65,536, which is 0.0015% of the upper limit of its output range ("full scale"), and the least significant bit of a 24 bit dac provides a resolution of 1 part in 16,777,216, which is 0.000006% of full scale. There are innumerable ways in which the accuracy which that resolution can theoretically provide can be and will be vastly degraded, both within the dac device, and in the surrounding circuitry. Real-world devices simply aren't that accurate.

Regards,
-- Al
Real-world devices simply aren't that accurate.

Agreed. However at around 1% distortion we start to have trouble to hear or identify distortion. At best speakers may have 0.3% distortion over their entire frequency operating range at full power (if they are exceptional quality).

Perhaps the point Peter is making is that extra ultra precision in modern line level analog equipment can quickly approach "diminishing returns" given the acuity of our hearing and the relatively crap performance of speakers and room acoustics in general...

Agreed, "perfect" is completely the WRONG word... as Kijanki points out NOTHING IS PERFECT... (except my wife and kids)

"Good Enough given the sad inadequacies of other audio system equipment" is more like it but "Perfect" ....certainly and absolutely NOT.
I think that this is an interesting technical discussion, lots of information. I keep on thinking that if God came down with the perfect DAC, none of us would probably like it. The whole Analog to Digital Recording process, plus the Digital to Analog conversion, has consistently removed Harmonic Content and Decay. Some of the best DACs have tried to reproduce what little Harmonic Content there is on the CD Disk. I believe that some utilize their Analog Output Stages, to overemphasize the little Harmonic Content there is. I don't bear any grudges to the changes from the original, because the Digital Process is so largely damaging to the Harmonic Content, that any exaggeration in the opposite direction is an improvement, within reason of course. This is why, I believe that if God did provide the perfect DAC, it would be the perfect magnifying lense to all of the huge warts of the Digital Recording Medium. More accuracy only exposes less and less Harmonic Content that is inherit in the Digital Recording itself. You might have the perfect DAC, but you probably wouldn't like it!
It's a panglossian notion that the failings of one area of a system can be ameliorated by the failings in another. Not IME.
hi shadorne:

if no component is perfect it follows that no preamp is neutral.

can you identify an allegedly neutral preamp ?

it is my hypothesis that after some duration, a component's deviation from neutrality will become evident.
if no component is perfect it follows that no preamp is neutral.

By neutral I mean it does not add coloration deliberately. There are plenty of examples of audio equipment that add coloration deliberately to create an effect or a particular sound.
It's a panglossian notion that the failings of one area of a system can be ameliorated by the failings in another. Not IME.

Agreed. My point was that worrying about the performance benefit of an incremental reduction of 0.0001% distortion seems like worrying about a pimple on a cows butt when you are drinking the milk. Considering that speakers already add far far more distortion why worry.
My idea with pangloss was to question the thought that a "perfect DAC" would be too revealing of the underlying flaws of CD format. IME the idea that there are ruthlessly revealing components is generally misconceived. Any such observation about a component more likely reveals a problem in the component than it does flaws inherent in RBCD.

Also, measured distortion is only one among many criteria that determine sonics, whether in loudspeakers or in CDPs.

As the price of good digital gear has come down, some opinions are returning to the early days of RBCD, in which conventional wisdom was that there were few if any audible differences between CDPs. I agree that there have been improvements, but there is still a fairly wide gap between budget and upmarket CDPs. At local audio club meetings it never ceases to surprise me how much effort is put into systems building in all areas except for the digital source. You can hear the difference.
Also, measured distortion is only one among many criteria that determine sonics, whether in loudspeakers or in CDPs

Agreed good point - like phase for example - you can get a big difference in soundstage/imaging even though timbre may be the same and distortion negligible.
if all components are imperfect, they exhibit coloration.

the fact that coloration is unintentional and a consequence of many factors implies that configuring a stereo system is essentially creating a balance among colorations.
Dear Dgarretson,
I wasn't claiming that a "perfect DAC" would be too revealing of the underlying flaws of the CD Format. Every Recording, Analog-Digital Conversion, and Playback peice of Equipment damages the original to varying degrees. They all still damage the original never the less. There are going to be some that greatly damage the original signal magnitudes greater than others. There are huge Sound differences between DACs; however, there is even a greater disparity in Sound Quality from one CD to another, that is clearly discernable even with the worse DAC. The only explanation for this huge disparity, is that the Analog/Digital Conversion does most of the damage to the Harmonic Content of the Recording. Magnifying this damage by increasing the accuracy of the DAC ten fold, might ameliorate some of the damage that the DAC causes, but won't even put a dent in the damage that the original Analog/Digital Conversion has caused. Such Damage is so severe, that the only solution might be to err in the opposite direction in the Analog Output Stage. Certainly, this means modifying the original, and not being ever more faithful to the accuracy of the severly damaged original signal. Being ever more faithful to the original signal, could even end up distorting the Music even more severe, by compounding the damage that was done during the Analog/Digital Recording Process. It may sound counter intuitive, but I believe that there is more going on in our Analog Output Stages, than just being faithful to the original signal.
Pettyofficer - Many analog tapes were digitized with jittery clocks and less than perfect A/D converters (long time ago). Recorded jitter cannot be suppressed and the only remedy is to digitize it again with better equipment (if analog tapes still exist). Recorded jitter is pretty much the same as hiss on analog tapes (jitter=noise in time domain). I only hope we're going in right direction.
Pettyofficer, in the long course of modifying my CDP I found that every RBCD sounded better than before-- including some '80s masters that I had written off completely. Various magazine reviewers have made similar observations regarding the better players. I can't recall a reviewer ever stating that a top player made a CD sound worse than before. Admittedly there are wide variations in the quality of A/D and D/A. However, good luck finding one CDP that can "euphonize" the defects of a poorly mastered CD, while sounding resolving with a well-mastered one. A player should be both highly resolving and musical.

When the discussion turns to the theoretical "damage" done in A/D and D/A conversions, I fall back on my empirical results modding the SCD-1. There have been some big leaps forward with this machine to the point that RBCD rivals vinyl. And yet of all the changes, the one that matters the least is toggling between the five digital filter options. So to the degree at least that the stock player allows variability in the way digital is handled, it is much ado about nothing.

" I can't recall a Reviewer ever stating that a top player made a CD sound worse than before ", does this mean that they are all perfect, or does it mean that only the perfect players were the ones that were reviewed. Your point of a 110% perfection, is suspect in itself.
I have also heard a huge improvement in each new generation of Digital playback equipment. The problem is that your best CDs sound better, while your terrible sounding CDs got even worse. This is all of the proof that I need that all of the huge advances in playback Digital/Analog Conversion still cannot turn a sows ear into a silk purse. There is inherit distortions in the Digital Recording that the best DAC in the World cannot ameliorate. The common perception, is that Jitter is the culprit that effectively vacuums up all of the Harmonic content in the Digital Recording. Henseforth, vanishing low amounts of jitter appear to be the selling point of top DACs. Harmonics are simply multiples of the cycle of the original Musical Event, usually at a lower level. If jitter isn't within the threshold of human hearing to distort the original Musical Event, why would it completely vacuum up the Harmonic Content at lower levels? The explaination is that Harmonic Content is more complex than the digital system that is trying to reproduce it. So, with all of the advances in Digital Recording and D/A Conversion playback, with vanishing low Jitter, do Digital Recordings still suffer from a vacuum of Harmonic Content? It doesn't seem to matter what digital equipment is used in the Recording or playback, Digital still suffers from a lack of Harmonic Content when compared to the best in Analog. You won't find a Digital System that sounds as good, with Harmonics, as a comparative Analog System. You might get a little closer, and a little closer, but you will never eventually get there. I fall back on the Empirical results of 30 years of development in Digital Recording, and Playback, with vanishing low Jitter, and Higher Sampling Rates, and still we haven't reached the level of Harmonic Content that a 60 year old system could easily reproduce. Perhaps Jitter is not the only problem?