I agree with something BigShot alluded to in a slightly later post...
A major part of the problem with discussing this sort of thing is that audiophiles often resort to inaccurate or vague terms to describe what they hear.
As a result, different people may use the same words to describe very different things, and it is often difficult to equate sound with measurements.
Also, much as some people prefer to disagree, exactly how certain measurable differences relate to AUDIBLE differences is still not always clearly defined.
To use the example I put forward as... an example...
To many people I know, Sabre DACs (especially those of the vintage of the 9018 chip), seem audibly to boost high frequencies.
SUBJECTIVELY, compared to DACs of many other brands, it sounds as if the Sabre DACs boost the upper midrange by about 1 dB.
(To offer the converse; if I take a component that uses an AD1955 DAC, and apply a 1 dB boost cenetered around 7 kHz, the result sounds more like an ESS9018 without the boost.)
However, when we measure the ESS9018 DAC, we find that its frequency response is exceptionally flat (like most other good DACs).
Therefore, an actual difference in frequency response DOES NOT account for the difference in sound... because no such measurable difference exists.
(Therefore, the difference in frequency response that people claim to hear does not actually exist.)
However, if we measure the impulse response of the filters used in the Sabre DACs, we find that they ARE in fact measurably quite different.
This would seem to suggest at least the possibility that the differences in filters, which we can measure, may account for the subjective/audible differences.
I'll also offer a more well known example of how sometimes it's simply a case of "not measuring the right thing".
Class B amplifiers commonly have a known flaw - known as "crossover notch distortion".
On an oscilloscope, this sort of distortion literally appears as a "notch" in the sine wave which appears right before and right after the zero crossing point of the waveform.
As it so happens, the actual distortion is largely independent of the amplitude of the music signal... for a given circuit, adjusted a given way, the notch will be a certain size.
As a result of this, the PERCENTAGE of distortion will vary depending on the signal level.
(With a large signal, the distortion will be a very small average percentage of the total; wheres a tiny signal may be almost entirely distortion.)
Basically, with a large signal, you have a very high level of distortion, but for a very small percentage of the time.
(In a poor design, with a certain input signal, you can quite literally have a signal with an AVERAGE THD of 0.5%, but with a THD of 20% for 0.5% of the TIME.)
Therefore, studies about the level of AVERAGE THD which is audible really do NOT apply to this situation (unless some controlled study has shown that they are similar).
(Many people seem to agree that excessive crossover notch distortion sounds "harsh"... but the amount which is required to be audible is not widely agreed upon.)
I'll offer another very simple example.....
Let's assume I have a speaker that exhibits ringing......
When I apply a 1 second burst at 1 kHz to this speaker, it continues to produce sound at -30 dB for 20 msec after the applied tone burst ends.
Now consider the THD of that speaker during the time interval from 5 msec to 15 msec AFTER THE APPLIED TONE BURST ENDS.
During that time interval, we have no input signal, but a clearly measurable output signal... so all of that output signal is "distortion".
So, for that time interval, the THD is 100%... but the AVERAGE THD is much lower (and it will be different depending on what arbitrary measurement interval you choose).
A similar situation exists with DACs....
The filters in DACs create some amount of ringing before and after transient signals.
Therefore, for some very short period of time, they produce an output signal that is "pure distortion".
The duration of this signal is very short, and most of the energy it contains is at "ultrasonic frequencies", but it also expends energy that is "borrowed" from the audible spectrum.
(When the signal is "spread out in time", much of the extra distortion is inaudible, but it also takes energy away from the "intended signal".)
Many people, including myself, believe that it is this ringing and its variants that accounts for the differences people claim to hear between different DAC filter choices.
And, as far as I know, very little actual controlled research has been done to confirm or deny this hypothesis.
Errrr.... yes and no.
Saying "plain old resampling of the signal" is an oversimplification.
In general, altering the sample rate involves some filtering.
If you look at the software you're using to convert between sample rates, you'll
USUALLY find options for what sort of filtering to use (in R8Brain, or Izotope, you can choose both the sharpness and the cutoff frequency of the filter).
And, yes, you will hear slight differences in the output when you change those settings (those differences are usually pretty subtle).
And, yes, some of those differences involve audible differences in high-frequency response (which can affect "the tonal balance").
And, yes, some pros DO prefer one converter over another "because of how it sounds".
If you want to see a comparison of a whole bunch of sample rate converters.... check out this website:
http://src.infinitewave.ca/
You'll see that many sample rate conversion utilities do deliver virtually identical outputs...
But a surprising number actually DO cause significant alterations to the signal...
I would agree that, with a DAC, assuming that the data remains the same, then the resulting analog audio should sound identical.
However, apparently, on some DACs, the various
TYPES of inputs have considerably different performance... particularly when it comes to rejecting jitter present on the incoming source.
(Feel free to argue for another ten pages about whether the difference should be audible or not... but you can readily measure it.)
I would also add that, subjectively, many people I know agree that Sabre DACs tend to emphasize high frequencies, resulting in a sound that seems more detailed, yet slightly unnatural. I'm not at all interested in arguing about "how we must be imagining it" - nor do I know anyone who has done proper tests to reasonably confirm or deny it. HOWEVER, in their early product literature, ESS (who makes Sabre DACs) claimed that, rather than design their DACs for the most accurate response, they used focus groups to "choose the output filter that most people preferred the sound of". They at least claimed, at that point, not that their DACs were accurate, but that they "sounded different and better". They have notably omitted this claim in recent years.