Here's the problem... with DACs in particular.
With virtually any modern DAC (except for a few goofy ones) all of the standard numbers are going to be ridiculously
GOOD.
THD will be really low, S/N will be really high, and frequency response will be very flat.
(I really doubt anyone can hear the difference between 0.02% and 0.003% THD, or between a S/N of 115 dB and one of 120 dB, or between a frequency response of 20 - 20kHz +/-0.1 dB and one of +/-0.05 dB.)
They will all be so good that it's really unlikely anyone can hear the difference.
Unfortunately, because of this, we have no choice but to conclude that the audible differences are due to some of the more
UNCOMMON measurements.
Also unfortunately, with many of those parameters, there is no clearly defined definition of what's "good" or "bad".
(Think: "What's the BEST color to paint your living room?").
Dynamic range is a derived number..... it's more or less another way of describing the S/N rating.
(Things like speakers can cause "dynamic compression" under certain circumstances - DACs cannot; it just doesn't happen.)
The same holds for terms like "pace, rhythm, and timing".
(
NO DAC alters the speed of your audio to an audible degree.)
Sometimes, if you alter the frequency response, or add some type of distortion, a file will sound "bouncier" or "like it drags" - but, if you actually measure it, you will find that the speed has not changed.
(Things with motors, like tape players and turntables, can run at the wrong speed; the speed variations in things like DAC clocks are so much smaller than what might possibly be audible it isn't even funny.)
Input type is irrelevant - since all we're talking about are
NUMBERS here.
Different inputs on a particular DAC may perform better than others; but, inherently, they all deliver the same exact numbers.
(If there was such a thing as a "perfect" input of each type they
WOULD sound identical - but some types are more likely to be closer to perfect than others.)
Impulse response is the opposite..... it varies a lot, and varies between different filter choices on the same DAC (if it offers choices).
Some filters have more or less ringing; some filters have symmetrical ringing, while others have more or less pre- or post- ringing.
While some people prefer one or the other, they're all actually compromises, and there is no specific "right" or "better" to aim for (although there are a few characteristics that should be avoided).
You can look at the 'scope pictures, and the differences are obvious, but how those differences correlate with how the DAC sounds is not at all well understood.
(With a lot of experience, you can make some broad correlations; for example, DACs with less ringing, and less pre-ringing, tend to sound smoother.)
Jitter is a horrifically complex and widely misunderstood topic.
Jitter is simply a term for "short term timing variations in the data clock".
Accurately describing jitter would require you to spell out several different parameters; many of which are difficult to measure.
You also need to differentiate between the jitter a DAC itself
GENERATES, and how well it avoids being influenced by jitter produced by other components.
Some DACs may have a lot of their own internal jitter; others may produce almost no jitter of their own, but be especially sensitive to jitter produced by the source they're connected to.
Even worse, all of the parameters involved depend on other parameters; for example, a DAC could produce a lot of low frequency jitter, produce very little high frequency jitter, but be very sensitive to high frequency random jitter on its input source.
Luckily, the
RESULTS of jitter can be quantified much more easily.
The
RESULTS of jitter, when it is an issue, are extra noises that shouldn't be there, and that show up as a sort of noise or distortion (properly called "jitter-induced sidebands").
(You'll see graphs that look like very sharp narrow peak with grass growing around it. Theoretically, the peak should be a single vertical line, and all of the "grass" is stuff that shouldn't be there.
Some of it is probably due to jitter. If you feed in an input signal with a lot of jitter deliberately added to it, and you see more grass, or taller grass, then your DAC is probably more sensitive to jitter.
Unfortunately, the exact shape of the tufts of grass, and where they're located, affects what you'll hear. Beyond saying "less grass and shorter grass are generally better" it gets rather complictaed.)
Most jitter tests are only somewhat meaningful.
For example, the widely used J-test.
Unfortunately, that test is actually testing several different things at once, and
ONLY in combination.
That test applies a sort of "torture test" signal to the input of the DAC, and looks at the output for sidebands that are caused by it.
(Sidebands that are there with the stress signal, but not with a plain test tone, are probably caused by jitter.)
This is directly analogous to putting a microphone inside a luxury car, driving it down the railroad tracks, and reporting how much noise your microphone picks up.
It tells you the combination of how much the car resists rattling and how well the sound proofing blocks the rattles that do occur... but it does
NOT tell you how well the soundproofing will block music being played outside the car.
(A car with no soundproofing that doesn't rattle because it's solidly built might measure the same as a car that rattles like crazy but has excellent sound proofing.)
DAC vendors, and manufacturers of assorted DAC tweaks, are also very fond of quoting jitter specs on specific components (like that special "ultra-low-jitter clock chip").
However, without knowing how that jitter affects the performance of the rest of the circuitry, that really isn't telling you much at all about the final result.
(If the DAC is super-sensitive to jitter, then the clock may make a huge difference; if not, then you may have just wasted a lot of money on a "premium part" that doesn't actually make any difference.)
I've got a
LOT more experience with DACs than almost anybody I know.... but I still can't give you anything simple to look for that would assure good sound quality.
However, here are a few tips:
1)
Look for low THD (any decent DAC should be able to deliver THD of 0.02% or less)
2)
Look for a good S/N ratio (anything above about 110 dB is
REALLY quiet - don't waste your time beyond that)
3)
Look for a pretty flat frequency response (there's no excuse for a DAC to not be 20 - 20 kHz +/- 0.5 dB or better)
4)
Filters.....
- generally short filters, with little ringing, sound relatively clean
- generally filters with longer post-ringing and shorter pre-ringing make plucked strings sound
VERY clean, but tend to make the high end sound a bit smoothed over
5)
I've never noticed a general difference between how Delta-Sigma DACs and R2R ladder DACs sound
6)
MOST unusual designs, like "non-oversampling" DACs, and DACs with strange filters, in general don't sound great (or sound good with one type of music and not so good with another)
The guys who design DAC chips really do know what they're doing (and so products that are built using their chips and "cookbook designs right out of the App notes" often work
VERY well.
Conversely, many manufacturers end up expending a lot of effort designing custom filters and circuitry that
DON'T sound as good as the chip maker's "cookbook" designs.
Bluntly, there are many
VERY expensive DACs that don't sound good at all (at least to me).... so, more with DACs than with other audio components, do
NOT assume that more expensive ones are actually better
I would put up Sabre DACs as an example of a few of my comments.
They have a distinctive sound.
Some products that use them have more or less of that distinctive sound.
However, when it's there, their distinctive sound is... well... distinctive.
Some people like it, and some do not, and some don't notice it at all (but most do).
While it seems to be related to their filter characteristics, I have never been able to point to the specific cause of their characteristic sound.
(Overall, the characteristics of their filters don't seem to be unique from everyone else's.)
Hi Folks,
Something I wanted to sort for myself for a long time and would like your opinion.
Among the different qualities and measurements we have for processors/dacs/etc, what would be the likely common audible thresholds for a property?
When does the difference become considerably less audible or perhaps only audible in the most resolving and expensive systems/headphones?
For example with Sample Rate since they correlate to frequencies we know the usual range for our ears and other equipment so a difference between 8kHz and 44kHz will be very audible but the difference between 192kHz and 384kHz will be considerably less audible (if at all
)
What about dynamic range?
Input Type? Jitter?
THD?
SNR?
Which threshold numbers for a given spec exceed the known "human spec"?
Which threshold numbers are
less likely to make an audible difference considering other analog/mechanic (Amps/Woofers) are involved?
Not looking for holy wars here, just to understand which factors can be prioritized for a limited budget
) 
