I would like to add 2 things to Javiers remark.
It should be noted though that there
can be quite audible differences between DAC's even expensive ones compared to cheap ones.
The reason for this is the reconstruction filter that may be lacking in DAC's that can 'handle' various bitrates.
an example for this can be found here:
archimago.blogspot.nl/2013/05/measurements-teac-ud-501-pcm-performance.html(this plot is made by Archimago and is used without asking permission)
These are plots of one and the
same DAC that is considered decent by most (the TEAC UD501).
What clearly can be seen here is the large
roll-off in the highs.
This is because it emulates a NOS DAC with no 'correct' digital or analog reconstruction filter that is REALLY needed to accurately reproduce signals above 10kHz.
The same FR results can be had from all other NOS DAC's around that are 'suited' for multiple samplerates.
These plots are only this rolled off when
44/16 is played (
most files are this format) but NOT when that is either upsampled to 192/16 or other high-res files are used.
I can assure you this roll-off is
VERY audible and can easily be picked out in a blind test thereby showing that these (usually expensive) NOS DAC's sound different.
And since some people prefer a slight roll-off in the highs (because a lot of recordings sound a bit 'sharp' or 'digital' will make some think the more expensive DAC's or the ones with a 'soft' or 'smooth' digital filter feel it sounds better where in reality it sounds different.
Some will even go as far as stating that although the roll-off is measured it actually isn't there at all, based on the fact that they hear 'smooth' highs.
The reason for this roll-off is because of the
sample rate is never
synchronised with an
audio signal and the higher the frequency of the audio signal
the less samples are taken.
When a (sharp) reconstruction filter is used that reconstruction filter simply 'swings' the analog output signal to its original level (and is why the post ringing is there).
Below some illustrative material of this effect.
The thin line is the original (high frequency) input sinewave which would also be constructed VERY SIMILARLY when a proper (post ringing) reconstruction filter is used.
The thicker average line, connecting the dots where the actual samples were 'taken' in time, is the
average level a NOS DAC without reconstruction filter will look like.
In reality the output will differ yet again as an output of a DAC chip is a 'sample and hold' construction.
A sample and hold is when a sample is 'taken' from a signal (the sinewave in this case) at a certain time and that 'sample value' is held to the same level (the hold in sample and hold) till the next sample value overwrites this and from that moment that sample is held.
This is shown below.
The very thick line I drew in (did not do a good job as I did it qickly) shows what a NOS DAC (or a real ladder DAC BEFORE the reconstruction filter) will actually output.
As you can see these DACs do
squarewaves extremely well but unfortunately there are no squarewaves in audio signals, just sinewaves.
It also shows that some signals are lower in amplitude while others may be at or near maximum.
The average is thus lower in amplitude.
This
gets worse the higher the frequencies are, this causes 'variable' roll-off (thus depending on the audio frequency)
The
NOS DAC advocates are all thrilled with the nice squarewave reproduction with no post or pre-ringing and will gladly show off the 'super fast' rise and fall times of those squarewaves. At the same time they think they are showing how incredible fast transients must be based on this. In reality transients in music aren't that fast at all and it is a load of BS.
NOS DAC manufacturers, however, will gladly show how well a squarewave looks but WON'T post pictures of a 10kHz sinewave for obvious reasons.
This
doesn't mean a NOS DAC
can't sound good. It can and also doesn't NEED to show this roll-off in the highs. For these DACs to do that the original 44.1 data needs to be upsampled at least 4x or high res material must be used. In this case a NOS DAC would be hard to distinguish from other DACs even though NOS DACs are technically still inferior. Fortunately for NOS DAC owners the hearing isn't as good as considered which irons out these last effects and they will be convinced the improvement in 'attack' will be there for sure.
One may ask why people speak of 'digitis' (the typical digital glare) when using 'standard' DACs while stating that NOS sounds 'better' even when the actual output signal looks NOTHING like the original signal (when compared on an analog level for the higher frequencies).
The answer is quite simple... either the average roll-off for higher frequencies is appealing to them or the hearing isn't as good as some think human hearing is.
This LAST part is the biggest issue between the 'subjectivsts' and opposite camps and cannot ever be bridged because of these fundamentally different POV's.
In any case....
If the distortion is low (below 0.01% = -80dB), the bandwidth is flat (within 0.1dB) from 40Hz to 16kHz, has little to no phase shifts in that band and has a (sharp) reconstruction filter, be it digital or analog, then chances are you won't be able to pick them blind.
Most DACs easily comply to those minimal requirements for 'transparency' as also mentioned by NwAvGuy.
Also the
output level is of importance when
comparing DACs directly.
A DAC that has 0.1 to 0.3dB higher output voltage will ALWAYS be picked as 'preferred' in a (badly performed) blind test.
When one considers that 0.1dB = around 1% it isn't hard to imagine that with a blind test with direct comparison a DAC that has 2.2V out and one that has 2.22V out can be picked out in a blind test where the higher one is likely to be preferred.
That
level difference may be the ONLY difference.
It is very easy to find DACs that have a marginally different output voltages so when comparing any DAC make sure the output levels are matched well within 1% when trying to perform a 'valid' blind test.
This (together with 2 DACs outputting the same (splitted) signal at the same time is the difficult part.
Certainly when one considers the output voltage is NOT happening at the same time the bits are entered in the DAC.
In a DAC the DAC chip receives a serial signal that is processed, this happens in a number of time steps (clock pulses) which may differ from DAC to DAC.
In general I think most (if not all people) will be very hard pressed to find ANY differences between properly designed DAC's in a real and well performed blind test but are perfectly able to pick 'high-end' miracle DAC's as in these the actual output signal simply differs from the recorded signal, however nice that alteration may sound.
Sorry for the long rant.... and tried to keep it as simple as possible.