[stv014]

Null testing/difference extraction is very sensitive to frequency response and delay differences. That is why audio recorded from actual analog DAC outputs, which do not have perfectly flat response, and also have small random delay variations due to the clock frequency not being entirely constant (basically, very low frequency jitter), will produce a relatively high difference signal. Notice that the sound card recordings all have a notch in the difference signal at the 1500-2000 Hz range, because that is what the frequency of the tones used for matching the levels and delay was.
 
In the case of D.wav, the minimum phase filter has a group delay that is zero at DC but increases towards the Nyquist frequency, and therefore the difference is also greater in the high frequency range. This file also has the highest group delay in the treble range.

 

[stv014]

Quote:

Originally Posted by UltMusicSnob /img/forum/go_quote.gif
 
I extended the range out to 96 kHz, because these are 192/24 wav's, and for the chance to see any artifacts that pop up, but I seriously doubt any energy up there made any difference in playback, unless my headphones have secret ultrasonic capabilities I didn't know about.

 
Most of the content in the 48-96 kHz range is from noise shaping by the delta-sigma DAC and ADC.
 
Quote:

Originally Posted by UltMusicSnob /img/forum/go_quote.gif
 
A.flac is playback of B.flac recorded from the Xonar Line Out?

 
Yes. It is from another Xonar, though, with a different DAC chip.
 
Quote:

Originally Posted by UltMusicSnob /img/forum/go_quote.gif
 
C appears to have a difference peaked right at the 22.5 kHz level, symmetrical around it. C is also the only one with the extra difference energy ~40 kHz, so maybe these are results of the slow roll-off instead of fast filter?

 
Yes, the slow roll-off filter mode of the DAC has worse image rejection, but lower group delay, and shorter ringing.
 
Quote:

Originally Posted by UltMusicSnob /img/forum/go_quote.gif
 
E.flac is also playback of B.flace, recorded from Xonar Line Out, but with volume adjusted before playback, instead of at playback like A.flac.

 
No, the volume was only adjusted for the Xonar D1 after recording (which is not clipped in this case even at maximum playback volume) to match it with the other files. This is only a minor detail, though, since all the adjustments were small. All files have reduced volume compared to the original by a factor of 0.93 (-0.63 dB), with which D.wav has a peak level just under 0 dBFS.

 

[UltMusicSnob]

Quote:

Null testing/difference extraction is very sensitive to frequency response and delay differences. That is why audio recorded from actual analog DAC outputs, which do not have perfectly flat response, and also have small random delay variations due to the clock frequency not being entirely constant (basically, very low frequency jitter), will produce a relatively high difference signal. Notice that the sound card recordings all have a notch in the difference signal at the 1500-2000 Hz range, because that is what the frequency of the tones used for matching the levels and delay was.
 
In the case of D.wav, the minimum phase filter has a group delay that is zero at DC but increases towards the Nyquist frequency, and therefore the difference is also greater in the high frequency range. This file also has the highest group delay in the treble range.

I was wondering about picking up additional signal changes due to the Xonar's playback out the analog line--I take it the analyses posted on page 7 are consistent with that. It makes the desired outcome--comparing 44.1 to 192 by comparing encodings of the analog playbacks of both, both captured at 192/24--inherently problematic. By removing one variable, sample rates at playback, more variables from the playback card (here, Xonar) are introduced.
 
From my post #46:
 
"On the one hand, this removes the effects of any difference purely on my end in DAC handling of differing playback rates at playback time--because both in this test are in 192_24.
On the other hand, it adds two new factors---not that they are known to have effects, just that these are differing experimental conditions:
     the DAC processing on stv014's device to send the 44.1 file to analog playback,
plus,
     the effect of A/D on stv014's 192-encoding capture."
 
I'm persuaded by all the above that the fine 'tuning' of the available parameters in resampling is important. Subtle, of course, but the differences are there.

 

[stv014]

Well, it would also be possible to capture the 192/24 file as well, I did not do that to avoid any complaints about the recording equipment hiding the difference due to its "limited resolution" or whatever other reason (although I do not believe that would be the case). So, comparing an upsampled file to a full D/A-A/D loop is expected to give results that are biased towards finding an audible difference. Nevertheless, a positive result is still surprising given the measured performance of the playback/recording loop, and so far it is not reproduced by others.
 
I did include D.flac for the purpose of comparing different reconstruction filters without passing the signal through any analog hardware, however. So, perhaps a few more similar upsampled files could be created with different filter parameters, as well as a loopback recording of B.flac.

 

[dizzyorange]

I think I can hear a small difference when using foobar's oversampling component but I actually prefer the sound with it off.  When it's on the sound feels slightly "smoother", if you will.  But I like the slightly more "raw" sound of 16/44.1.  Maybe it's just all in my head.

 

[stv014]

So, why not try some ABX tests to find out ?

 

[dizzyorange]

I don't have an easy way to do it blindly.  I've been comparing the sound by activating and deactivating the component in foobar, which is instant, but I know which is which.  I know about the ABX foobar plug-in, but I don't know how to upsample a song and save it in the upsampled state.  I suppose I could look it up but it doesn't seem worth the effort at the moment.

 

[wakibaki]

@ UltMusicSnob
 
Your username seems designed to evoke distrust. For myself I don't like snobs, particularly where music is concerned. Having been a musician for some 45+ years, I prefer to jam with people with an inclusive attitude.
 
Why don't you take your flag over to hydrogenaudio.org, they are widely recognised to be the most critical analysts of claims such as yours. Run it up the pole and see if anybody salutes it. Frankly, I think they will tear it to shreds p.d.q., i.e. if you are hearing a difference it is an artifact of the processing, not evidence of a superior SQ, which I think is what an ultimate music snob would be trying to demonstrate. I mean, otherwise, what are you doing here?
 
w

 

[UltMusicSnob]

Quote:

So, why not try some ABX tests to find out ?

@dizzyOrange: Yes, ABX, always a good idea. There are test files of various origins around here and other forums.

 

[UltMusicSnob]

Quote:

@ UltMusicSnob
 
Your username seems designed to evoke distrust. For myself I don't like snobs, particularly where music is concerned. Having been a musician for some 45+ years, I prefer to jam with people with an inclusive attitude.
 
Why don't you take your flag over to hydrogenaudio.org, they are widely recognised to be the most critical analysts of claims such as yours. Run it up the pole and see if anybody salutes it. Frankly, I think they will tear it to shreds p.d.q., i.e. if you are hearing a difference it is an artifact of the processing, not evidence of a superior SQ, which I think is what an ultimate music snob would be trying to demonstrate. I mean, otherwise, what are you doing here?
 
w

Sorry about the "snob" part, it's only meant ironically. As a tag, at least it's memorable.

 
I know it's unreasonable to expect folks to read through long threads, so the short version is:
   I came looking for folks with lots of ABX experience (found 'em, so thank-you's to all), and hoping for significant results on upsampling ABX's by other folks (found one, maybe).
   I *do* think it's an artifact of processing; specifically, I think it's an artifact differential of different processing for files of different sample rates.
   ...what's 'SQ'?

 

[Baxide]

Quote:

I think I can hear a small difference when using foobar's oversampling component but I actually prefer the sound with it off.  When it's on the sound feels slightly "smoother", if you will.  But I like the slightly more "raw" sound of 16/44.1.  Maybe it's just all in my head.

Upsampling is like adding water to wine. The taste is still there but far smoother. Vocal and instruments separation is also less defined and dynamic impact can be far more toned down.

 

[stv014]

Quote:

Originally Posted by wakibaki /img/forum/go_quote.gif
 
if you are hearing a difference it is an artifact of the processing

 
It would be interesting to find out what those artifacts exactly are, however.

 

[Baxide]

Quote:

 
It would be interesting to find out what those artifacts exactly are, however.

They are either 0 or 1. The exact choice in the data stream depends on whatever mathematical calculation is done by the upsampler. The data is padded with extra 0's and 1's to increase the data length from say 44.1KHz to 192KHz. That's roughly a five fold increase. So around 20% of the data will be original, and 80% will be plucked out of thin air.

 

[stv014]

That is not a very helpful post. Also, I know how the upsampler used in most of the tests works, since I wrote it. The question is exactly what differences UltMusicSnob (and so far no one else) hears in the ABX test between files that should in theory sound the same.

 

[Baxide]

Quote:

That is not a very helpful post. Also, I know how the upsampler used in most of the tests works, since I wrote it.

The truth always hutrts. It's not designed to be helpful, but factual. Consumers have the right to know what they are buying,