29/01/2012-00:11:16
RE: acquisition et numérisation d'une sortie ligne
Interview: Daniel Weiss - PAROXYSMAL DISCHARGE
If you had a room and a system and choice of music that you have full control over, do you think you could tell a €200 CD player from a €15000 transport feeding one of your DAC's?
I wouldn't expect to hear a difference actually.
Because of the jitter suppression?
Yes, provided of course that the data is read correctly.
...
Given the Nyquist theorem and the limits of human hearing, is the only reason we sometimes can tell 16 bits 44,1kHz from hi-res, just the reconstruction filters?
Because theoretically we shouldn't hear the higher resolution.
I don't know, maybe it has to do with non-linearities in the ear, so high frequencies above 20kHz can end up at lower frequencies through intermodulation.
Maybe there's something to that, I don't know.
But then the speakers have to do that, and the microphones etc.
A problem is that very few microphones record that high frequencies.
If you analyze a 24-bit, 96kHz recording of classical music, it might not have anything above, say, 25kHz.
But still people think it sounds much better.
So could it be the bit depth, the extra dynamics?
It's different in the D-A converter, with the frequencies it can transmit and the anti-aliasing or the reconstruction filter which can be much flatter, so it gives you less artifacts from the filter.
Your equipment supports up to 24 bits, 192kHz. Is that completely overkill?
Yes. Bob Stuart of Meridian once gave a talk at AES and his proposal was to keep it at 60 kHz.
The closest one we have as a standard is 88,2, but more equipment supports 96.
96 is coming from the studio standard, doubled from 48. But 88,2 or 96 will be plenty.
So anyone claiming to hear a difference between 96 and 192 would either be hearing placebo or in how it was converted to that sampling rate?
That is a point of course.
I know of professional people having done tests with these conversions and they can hardly hear a difference, even if it's up to 192 and down to 44,1 again.
Especially if they use Saracon!
Yes, they used Saracon.
That seems to have become a bit of a standard?
There are some competitors, but it's one of the better ones.
There's a website actually, comparing all kinds of sampling rate converters.
And saracon does very well.
Yes, luckily (laughs).
If you could choose only one format, would it be 24 bits, 44,1kHz or 16 bits, 88,2 kHz?
Probably 16/88.
At least in theory, our ears can manage more than 96dB of dynamics, but we can't hear much above 20kHz.
So according to that, the extra dynamics should be better.
Professional people usually go with larger word lengths than bandwidth.
That's because they will do processing with the sound, so they want that margin.
As an end format, you can do a very decent encoding in 16 bits.
I think the dynamic range is enough, you don't have such dynamic range in a listening situation usually.
Buy you have a finite number of steps within those 16 bit, so it's not infinite precision.
Would you ever need a finer resolution than that?
The question is whether you need the signal to noise ratio basically.
It's 96dB at 16 bits, which is huge.
You have maybe a 30dB dynamic range in the music, and that is already quite a lot actually.
So you still have a 60dB lower noise floor.
I don't think you hear the noise of a CD.
At normal listening levels, do you hear the noise floor?
No.
See, that's what I mean.
If you had a room and a system and choice of music that you have full control over, do you think you could tell a €200 CD player from a €15000 transport feeding one of your DAC's?
I wouldn't expect to hear a difference actually.
Because of the jitter suppression?
Yes, provided of course that the data is read correctly.
...
Given the Nyquist theorem and the limits of human hearing, is the only reason we sometimes can tell 16 bits 44,1kHz from hi-res, just the reconstruction filters?
Because theoretically we shouldn't hear the higher resolution.
I don't know, maybe it has to do with non-linearities in the ear, so high frequencies above 20kHz can end up at lower frequencies through intermodulation.
Maybe there's something to that, I don't know.
But then the speakers have to do that, and the microphones etc.
A problem is that very few microphones record that high frequencies.
If you analyze a 24-bit, 96kHz recording of classical music, it might not have anything above, say, 25kHz.
But still people think it sounds much better.
So could it be the bit depth, the extra dynamics?
It's different in the D-A converter, with the frequencies it can transmit and the anti-aliasing or the reconstruction filter which can be much flatter, so it gives you less artifacts from the filter.
Your equipment supports up to 24 bits, 192kHz. Is that completely overkill?
Yes. Bob Stuart of Meridian once gave a talk at AES and his proposal was to keep it at 60 kHz.
The closest one we have as a standard is 88,2, but more equipment supports 96.
96 is coming from the studio standard, doubled from 48. But 88,2 or 96 will be plenty.
So anyone claiming to hear a difference between 96 and 192 would either be hearing placebo or in how it was converted to that sampling rate?
That is a point of course.
I know of professional people having done tests with these conversions and they can hardly hear a difference, even if it's up to 192 and down to 44,1 again.
Especially if they use Saracon!
Yes, they used Saracon.
That seems to have become a bit of a standard?
There are some competitors, but it's one of the better ones.
There's a website actually, comparing all kinds of sampling rate converters.
And saracon does very well.
Yes, luckily (laughs).
If you could choose only one format, would it be 24 bits, 44,1kHz or 16 bits, 88,2 kHz?
Probably 16/88.
At least in theory, our ears can manage more than 96dB of dynamics, but we can't hear much above 20kHz.
So according to that, the extra dynamics should be better.
Professional people usually go with larger word lengths than bandwidth.
That's because they will do processing with the sound, so they want that margin.
As an end format, you can do a very decent encoding in 16 bits.
I think the dynamic range is enough, you don't have such dynamic range in a listening situation usually.
Buy you have a finite number of steps within those 16 bit, so it's not infinite precision.
Would you ever need a finer resolution than that?
The question is whether you need the signal to noise ratio basically.
It's 96dB at 16 bits, which is huge.
You have maybe a 30dB dynamic range in the music, and that is already quite a lot actually.
So you still have a 60dB lower noise floor.
I don't think you hear the noise of a CD.
At normal listening levels, do you hear the noise floor?
No.
See, that's what I mean.
"Voici mon secret. Il est très simple: on ne voit bien qu'avec le cœur. L'essentiel est invisible pour les yeux." Antoine de Saint-Exupéry