Hans asked me to post directly. Not expecting influence I guess... It was a few posts ago.
To recall, I could not really hear any significant difference between source and the up-sampled 88.2 versions (up-sampled with the 4-letter word software I keep forgetting)
Only after I up-sampled with Roon I could hear a difference with the original 44.1 source (I liked the the original better here)
so feel free to post your observations directly here?
To clarify for everyone. What we are investigating in these digital-filter experiments is, whether typical interpolator design/implementation is what's responsible for the characteristic difference in sound between OS playback, and NOS playback. Toward that question, the 88.2 experiment seeks to answer whether those upsampled files sound a) the same as, or b) worse than, or c) better than the their respective 44.1 source files when both a played via an NOS DAC.
Doede's results are interesting in that, the 88.2 PGGB OS interpolated files sounded as good to him as their respective 44.1 source files played back with no OS interpolation (meaning, pure NOS).
Last edited:
I did some check today. I made a FS48kHz track with 12kHz square wave and played this with ROON on my DDDAC NOS DAC
No surprise the wave form was nice and square (checked with scope of course)
Than I up-sampled with the Roon DSP to 192kHz and it became a very crude "sine wave": just a couple of steps: one for every wordclock periode, so in stead of one big step, it was 4 smaller steps looking like a "sine".
Using smooth/minum phase the curve became something like an e-curve smoothing (just a like RC filter after a square wave)
Ken, could you have this file up-sampled by the PGGB software? I would be curious what comes out....
Dropbox - To Share - Simplify your life
.
No surprise the wave form was nice and square (checked with scope of course)
Than I up-sampled with the Roon DSP to 192kHz and it became a very crude "sine wave": just a couple of steps: one for every wordclock periode, so in stead of one big step, it was 4 smaller steps looking like a "sine".
Using smooth/minum phase the curve became something like an e-curve smoothing (just a like RC filter after a square wave)
Ken, could you have this file up-sampled by the PGGB software? I would be curious what comes out....
Dropbox - To Share - Simplify your life
.
Hi DDAC,
I up-sampled your 12Khz/48Khz file with my PCM1792 to 192Khz, and the first image came out.
Then I instructed JRiver to up-sample to 192Khz, and the second image came out.
Very strange, but interesting to know.
I never liked JRiver doing the up-sampling.
Hans
.
I up-sampled your 12Khz/48Khz file with my PCM1792 to 192Khz, and the first image came out.
Then I instructed JRiver to up-sample to 192Khz, and the second image came out.
Very strange, but interesting to know.
I never liked JRiver doing the up-sampling.
Hans
.
Attachments
Hans asked me to post directly. Not expecting influence I guess... It was a few posts ago.
To recall, I could not really hear any significant difference between source and the up-sampled 88.2 versions (up-sampled with the 4-letter word software I keep forgetting
Only after I up-sampled with Roon I could hear a difference with the original 44.1 source (I liked the the original better here)
so feel free to post your observations directly here?
Hi, Doede, I wish to try and reserve as much of the patience of the author of the PGGB resampling/remastering software as possible. Who has graciously been using his time to resample our test files for us.
He does offer download of the PGGB software for a 30 day free trial. In which you are free to resample as many many files as you wish, including various dithering options and such. After trial, should you become interested to purchase a permanent license, let me know. ZB, the author, has generously offered a 50% discount through me to anyone participating in our thread.
PGGB - Offline remastering
Last edited:
Can you please elaborate on what was measured here? The output of the DAC in it's normal, intended OS mode?
An both times the DAC was fed 192kHz PCM, once upsampled by PGGB and once by Jriver?
The correct answer would be a sine wave: all odd harmonics of 12 kHz alias to 12 kHz, so you end up with 12 kHz as the only frequency component below Nyquist.
Do you get the same waveforms when you attenuate the signals before converting the sample rate? It could be an intersample overshoot issue.
Last edited:
Suppose you have a discrete-time signal with samples +1, +1, -1, -1, +1, +1, -1, -1 and so on. The only signal band-limited to the Nyquist frequency that can produce those sample values is a sine wave at fs/4 with amplitude sqrt(2) and phase pi/4 rad at the first sample. Hence, that's the correct reconstruction.
If the reconstruction or interpolation filter clips at +1 and -1, a part of the sine wave gets cut off, producing a kind of trapezium.
If the reconstruction or interpolation filter clips at +1 and -1, a part of the sine wave gets cut off, producing a kind of trapezium.
Last edited:
With the files fed from Audirvana into an S.M.S.L. M100 DAC (using the AK4452 DAC, a device capable of 768KHz PCM) there was perceived a common character difference between all the 44.1 KHz files to their counterparts upsampled to 88.2 KHz.
In a subsequent test comparing the 88.2 KHz files to ones I believe were up sampled and then down sampled back to 44.1KHz (in the previous test) the character differences appeared less obvious.
In a subsequent test comparing the 88.2 KHz files to ones I believe were up sampled and then down sampled back to 44.1KHz (in the previous test) the character differences appeared less obvious.
Hierfi,
You preferred all 4 originals in the previous test, and now you heard the least differences between the 88.2 and the processed 44.1,true ?
That leads to the logical conclusion that you prefer the original 44.1 to the 88.2.
Your M100 has a huge value for money, but good to know is that it isn’t a NOS Dac and suffers slightly from low level linearity and jitter, but your results are valuable anyhow.
Hans
You preferred all 4 originals in the previous test, and now you heard the least differences between the 88.2 and the processed 44.1,true ?
That leads to the logical conclusion that you prefer the original 44.1 to the 88.2.
Your M100 has a huge value for money, but good to know is that it isn’t a NOS Dac and suffers slightly from low level linearity and jitter, but your results are valuable anyhow.
Hans
Thanks Hans,
The original 44.1 files were preferred over the 88.2 files for reasons that their limitations were more palatable. This is in reference to substantively more realistic files from Tidal HiRez (downgraded to exclude MQA) through the M100 at 44.1. This suggests that the manner of the files under test are in question in the manner of their recording, that the M100 (as notwithstanding its limits or merits) can nevertheless resolve.
Gerrit
The original 44.1 files were preferred over the 88.2 files for reasons that their limitations were more palatable. This is in reference to substantively more realistic files from Tidal HiRez (downgraded to exclude MQA) through the M100 at 44.1. This suggests that the manner of the files under test are in question in the manner of their recording, that the M100 (as notwithstanding its limits or merits) can nevertheless resolve.
Gerrit
Finally I found some time for the 44,1 to 88,2 test. The difference is so subtle that the verdict could easily be of no difference. But I think I sense something like more dryer sound not meaning a degradation though. In the original 44,1 files in NOS mode, the bass is kind of soft. In the up/downsampled files with audacity, the bass becomes more punchy, close to boomy. Same effect with the PGGB up/downsampling but not so intense. PGGB only upsampling to 88,2 does nice! However it's not practical to process an entire music library...
With the experience gained from all this I went back to listen to these particular files for a NOS vs OS test. Using PGGB only for upsampling at 88,2 compared to 8x on chip oversampling, I would say PGGB is slightly better than OS yet NOS has something more... Maybe it's my system.
With the experience gained from all this I went back to listen to these particular files for a NOS vs OS test. Using PGGB only for upsampling at 88,2 compared to 8x on chip oversampling, I would say PGGB is slightly better than OS yet NOS has something more... Maybe it's my system.
Thanks, for that report, Kostas. I'm particularly interested to read more about your above statement. When you have some time, would you elaborate on the, "yet NOS has something more", feeling which you have? It's alright if you suspect that it may only be your system. Different song files, or a higher upsampling rate (which, the creator of PGGB has recommended since early on) than 88.2KHz, may cause you to hear things differently, so it's okay, whatever you feel that you hear right now.
In any case, your opinion on where you feel that NOS, perhaps, is still subjectively superior in certain regards is very valuable to our investigation.
Last edited:
Hi Ken,
I really strive to define that "NOS character" myself... Hard to say. It feels like a sound stage integrity beyond micro details. Think about dynamics that don't rule the whole, palpating but not attack, or an ambience reverberation that should not be focused. I usually avoid these funny expressions... I'm planning for another test using tracks which I'm well familiar with, upsampled with audacity in lieu of better software. If anything interesting comes out then perhaps we could ask for a PGGB process. Do I understand correctly that I'll have to upsample to 352,8kHz to match oversampling 8x?
I really strive to define that "NOS character" myself... Hard to say. It feels like a sound stage integrity beyond micro details. Think about dynamics that don't rule the whole, palpating but not attack, or an ambience reverberation that should not be focused. I usually avoid these funny expressions... I'm planning for another test using tracks which I'm well familiar with, upsampled with audacity in lieu of better software. If anything interesting comes out then perhaps we could ask for a PGGB process. Do I understand correctly that I'll have to upsample to 352,8kHz to match oversampling 8x?
There are more than one characteristic which could be used to describe NOS which I would also use. Among those, I think the most significant to me is the fatigue free playback charater of NOS. I feel that long term listening fatigue or boredom is the greatest sin committed by typical OS playback. Have you a feeling about any difference in listening fatigue/boredom of NOS versus the PGGB 88.2 upsampled files?
Correct, 44.1KHz x 8 = 352.8KHz.