DeltaWave null-testing audio comparator (beta)

[pkane2001]

23 minutes ago, Jud said:

 

Got it, thanks.  I'm pretty amazed at the pace at which you're making changes in response to requests, and grateful.  

 

What remains in order to do the comparisons I would like, as mentioned in my previous note, is only one thing not currently present in DeltaWave, and that is to enable the mixed down Reference track to be compared and saved in either the right channel or the left.  

 

One reason, as I'm sure you already understand, is to be able to determine whether any preference results from differences between left and right ears, differences in speaker-room interactions between left and right, or is due to differences in the tracks themselves.  If preference switches depending on which channel the reference is played from, it's the track; if it doesn't switch, it's ears or speakers.

 

Saving a file with the comparison, which DeltaWave already has (though not with the Reference file in the right channel as far as I can tell) allows the application of various Audirvana+ and HQPlayer upsampling filters on playback.  For me and many others who use the streaming capabilities of these players, it also allows playback through our main systems.

 

Sure, Jud. Easy enough to do, I actually wanted to do this in the last release, but simply forgot

 

[pkane2001]

12 minutes ago, STC said:

 

I wonder how many passes the so called original copy went thru before reaching Esldude. So why an additional of three needed to stabilize?  Maybe something else you are overlooking. 

 

Esldude was using his own DAC and ADC in the recording process, so likely a very different set of filters and distortions than anything that the original copy has gone through. @fas42, would be interesting to see if you can find anything through DW. There's a feature called Report Cruncher (under file menu) that lets you generate multiple file comparisons. You just organize file names in pairs, and then let it rip. Each report will contain all the statistics and charts that will help you examine the results. So you can run the comparisons between the reference and all 8 generations in one step:

 

[pkane2001]

37 minutes ago, pkane2001 said:

 

Sure, Jud. Easy enough to do, I actually wanted to do this in the last release, but simply forgot

 

 

 

@Jud, please re-download the installation zip file. You may need to uninstall the previous version, first, as I left it as v1.0.18

 

The updated DW has the Save Reference Waveform option under file menu. Sorry I missed it in the last update!

 

[Jud]

2 hours ago, pkane2001 said:

 

 

@Jud, please re-download the installation zip file. You may need to uninstall the previous version, first, as I left it as v1.0.18

 

The updated DW has the Save Reference Waveform option under file menu. Sorry I missed it in the last update!

 

 

Thanks Paul! I'll check it out this evening.

[Jud]

Hi Paul.  What does DeltaWave do with the stereo comparison if the tracks are different sample rates?

[pkane2001]

7 minutes ago, Jud said:

Hi Paul.  What does DeltaWave do with the stereo comparison if the files are different sample rates?

 

Hi Jud.

 

The default is to downsample the higher rate file to match the lower rate of the other file. In settings, you can ask it to upsample, instead. In this case the lower rate file will be upsampled to the higher rate of the other one.

 

[Jud]

15 minutes ago, pkane2001 said:

 

Hi Jud.

 

The default is to downsample the higher rate file to match the lower rate of the other file. In settings, you can ask it to upsample, instead. In this case the lower rate file will be upsampled to the higher rate of the other one.

 

 

If it gets hold of an MQA file, I'm assuming it treats it as the "folded" (lower) sample rate?

[pkane2001]

2 minutes ago, Jud said:

 

If it gets hold of an MQA file, I'm assuming it treats it as the "folded" (lower) sample rate?

 

I assume so, DW uses the reported sample rate encoded in file header.

 

[Jud]

Well now, it will be interesting to see whether an MQA file in one channel and a non-MQA file in another channel will even play back via Audirvana+ (which does the "first unfold" if set to detect MQA) or an MQA-enabled DAC; or for that matter whether a DAC or Audirvana+ would recognize an MQA stereo file mixed down to mono.

 

[pkane2001]

3 minutes ago, Jud said:

Well now, it will be interesting to see whether an MQA file in one channel and a non-MQA file in another channel will even play back via Audirvana+ (which does the "first unfold" if set to detect MQA) or an MQA-enabled DAC; or for that matter whether a DAC or Audirvana+ would recognize an MQA stereo file mixed down to mono.

 

 

I haven't been following MQA discussions closely, so can't say. It'll be an interesting experiment. I assume you don't want to resample or otherwise modify MQA track, as any bit changes will likely result in it not being 'authenticated', right? 

 

[mansr]

12 minutes ago, Jud said:

Well now, it will be interesting to see whether an MQA file in one channel and a non-MQA file in another channel will even play back via Audirvana+ (which does the "first unfold" if set to detect MQA) or an MQA-enabled DAC; or for that matter whether a DAC or Audirvana+ would recognize an MQA stereo file mixed down to mono.

The MQA encoding uses both channels. If one channel is replaced, it will no longer identify as MQA.

[esldude]

9 hours ago, fas42 said:

Purely done in Audacity, this is what the amplitude envelope looks like,

 

 

Zooming in,

 

 

and again,

 

 

No aspect of DW contaminated this story, - what gives??

Those amplitudes are very small.  Is this from a difference signal, was the difference done in Audacity or is it a near silent portion? If you aligned to the nearest sample in Audacity and differenced them then I would say it is happenstance.  Sometimes the ADC would have sampled so this works and sometimes you are rather in between sample times.  

 

I'm not seeing those waveforms in Audacity looking at the the raw files themselves.  

[Jud]

24 minutes ago, mansr said:

The MQA encoding uses both channels. If one channel is replaced, it will no longer identify as MQA.

 

OK, confirms what I was wondering about.  This makes good comparisons more difficult but not impossible.

[mansr]

2 minutes ago, Jud said:

OK, confirms what I was wondering about.  This makes good comparisons more difficult but not impossible.

Yes, you'll have to save the decoded MQA first, then use that for comparisons.

[Jud]

Just now, mansr said:

Yes, you'll have to save the decoded MQA first, then use that for comparisons.

 

Exactly.  Tips on how?  

[mansr]

24 minutes ago, Jud said:

Exactly.  Tips on how?  

I use the Bluesound decoder on a random ARM device. For PC/Mac users, the best option is probably one of the many virtual sound cards that save the data to a file.

[Jud]

8 hours ago, pkane2001 said:

 

 

@Jud, please re-download the installation zip file. You may need to uninstall the previous version, first, as I left it as v1.0.18

 

The updated DW has the Save Reference Waveform option under file menu. Sorry I missed it in the last update!

 

 

Hi Paul - 

 

- Would the reference file always play in the left channel, even when swapping reference and compare files?

 

- Can you think of any audible advantages that the reference file might have due to being used essentially as a template for matching the two files (other than phase drift correction, which I'd turn off)?

[pkane2001]

6 minutes ago, Jud said:

 

Hi Paul - 

 

- Would the reference file always play in the left channel, even when swapping reference and compare files?

 

- Can you think of any audible advantages that the reference file might have due to being used essentially as a template for matching the two files (other than phase drift correction, which I'd turn off)?

 

If you play it from the Play menu, then yes. If you use the Comparator Stereo XY mode, you can switch left and right through the menu Swap Reference and Comparison.

 

The results may be different when swapping Reference and Comparison files on the main screen and doing another match. That's because the algorithm will adjust everything, including level to the other file. It's possible that the result will not sound the same, even if because of a slightly different level. They will be similar, but not exactly the same. Best to use the comparator to swap.

 

[fas42]

6 hours ago, esldude said:

Those amplitudes are very small.  Is this from a difference signal, was the difference done in Audacity or is it a near silent portion? If you aligned to the nearest sample in Audacity and differenced them then I would say it is happenstance.  Sometimes the ADC would have sampled so this works and sometimes you are rather in between sample times.  

 

I'm not seeing those waveforms in Audacity looking at the the raw files themselves.  

 

The copy samples were brickwalled at 19kHz in Audacity, and aligned. So you're looking at the >19k content - which after the 3rd copy matches excellently with the previous generation. And, they come from the area where I posted the earlier clips, where the waveform mismatch was so obvious.

 

Going to change tack now - will look at full spectrum waveform, and do what's necessary to achieve best null; the raw match at 352k in DW is very poor, because of variance with frequency.

[esldude]

6 hours ago, fas42 said:

 

The copy samples were brickwalled at 19kHz in Audacity, and aligned. So you're looking at the >19k content - which after the 3rd copy matches excellently with the previous generation. And, they come from the area where I posted the earlier clips, where the waveform mismatch was so obvious.

 

Going to change tack now - will look at full spectrum waveform, and do what's necessary to achieve best null; the raw match at 352k in DW is very poor, because of variance with frequency.

I was thinking brickwalled in the opposite direction. 

[rando]

Nobody hardly complains of hearing their name in a positive light. Like many others I've been enjoying this piece of software.  Thanks. 

[fas42]

Okay, wasn't happy with the results I was getting with the Bob Marley full spectrum processing, tried doing some band splitting. Still not finding good waveform nulls, so selected a specific frequency which had a peak in the track spectrum, at 5165kHz; notch filtered out the track content at that frequency - in DW, for that, was getting "Cross-correlation found multiple phase inversions -- data may be too noisy or too dissimilar!" messages at times depending upon sample rate.

 

Final effort was to upsample to big numbers, 1.411022M and save in 24 bits - which gave me over a 4 sec clip,

 

 

 

Last zoomed in,

 

 

Note noise levels are good, so that aspect is working fine. But the delta is still only slightly better than 40dB down, even with this extreme level of sampling.

 

Trying the same exercise with 44.1k/16 bit, and got

 

 

... barely better than the original waveforms - so sample rate, etc, of input waveforms is critical to determining what one gets.

[pkane2001]

1 hour ago, fas42 said:

Okay, wasn't happy with the results I was getting with the Bob Marley full spectrum processing, tried doing some band splitting. Still not finding good waveform nulls, so selected a specific frequency which had a peak in the track spectrum, at 5165kHz; notch filtered out the track content at that frequency - in DW, for that, was getting "Cross-correlation found multiple phase inversions -- data may be too noisy or too dissimilar!" messages at times depending upon sample rate.

 

Final effort was to upsample to big numbers, 1.411022M and save in 24 bits - which gave me over a 4 sec clip,

 

 

 

Last zoomed in,

 

 

Note noise levels are good, so that aspect is working fine. But the delta is still only slightly better than 40dB down, even with this extreme level of sampling.

 

Trying the same exercise with 44.1k/16 bit, and got

 

 

... barely better than the original waveforms - so sample rate, etc, of input waveforms is critical to determining what one gets.

 

Frank, I like that you are testing these different scenarios, but I'm not clear what it is you are trying to achieve. Can you describe what it is you are trying to get to? What nulls are you getting, and with which files?

 

One thing to remember: when you upscale you are not adding any additional information to the waveform. It's a form of averaging and interpolation that, in effect, produces some smoothing of the data by inserting artificially computed values (new samples).

 

The null value is computed as an average (RMS for difference, mean for correlated null). What that means is if you add more values that are 'smoothed' or averaged in between the real ones, you are increasing the number of matching samples, but they are not real. So it's not very surprising that null results improve with upscaled data -- the smoothed, interpolated data is skewing them.

 

By the way, that's why downsampling is the default in DeltaWave for matching sampling rates -- upsampling has an impact on the resulting match.

[fas42]

Paul, I'm trying to see if DW can give me information as to what's different in two versions of an audio, that I can relate to what I'm hearing - just getting a result that two files are a certain percentage different is of little value in doing that.

 

Upsampling is producing a waveform that my ears actually hear, the 44.1 version has the information to produce that, but if one could reproduce that 44.1 waveform, as a stepped output, it would sound pretty awful ...

 

There's no point in DW producing results which are meaningless to my ears - so I believe it has to process the input so that it's dealing with what the waveform looks like in the audio sections of the circuitry - not the digital version.

 

In the spirit of that intent, I'm now going to deliberately contaminate Bob Marley B in various ways, and see if DW can pick up what's going on. If it can't, then its only value is in distinguishing whether there's a variation, expressed as as a single, meaningless number.

 

 

[fas42]

And problems straight away - I merely attenuated the B wave by 0.38 dB, and this was the result:

 

DeltaWave v1.0.18, 2019-03-20T13:01:47.5458208+11:00
File 1:  Bob Marley B.wav[L] 237568 samples 44100Hz 16bits, MD5=00
File 2: Bob Marley B,-0.38.wav[L] 237568 samples 44100Hz 16bits, MD5=00
Settings:
    Gain:True, Remove DC:True
    Non-linear Gain:False    EQ FFT Size:262144, EQ Frequency Cut: 0Hz - 0Hz, EQ Threshold: -160dB
    Correct Drift:True, Precision:18
    Upsample:False, Window:Hann
    Spectrum Window:Blackman, Spectrum Size:524288
    Spectrogram Window:Lanczos, Spectrogram Size:32768, Spectrogram Steps:1024
    Dither:False

Discarding: Start=20s, End=6s
Discarding: End  =20s, End=6s

Initial peak values File 1: -2.568dB   File 2: -2.948dB
Initial RMS values File 1: -17.94dB   File 2: -18.32dB

Null Depth=27.301dB
X-Correlation offset: 0 samples

Final peak values File 1: -2.568dB   File 2: -2.565dB
Final RMS values File 1: -17.916dB   File 2: -17.916dB

Gain= -0.3842dB (0.9567x) Phase offset=0ms (0 samples)
Difference (rms) = -90.05dB [-96.02dBA]
Correlated Null Depth=91.71dB [85.86dBA]
Clock drift: 0 ppm

Files are NOT a bit-perfect match (match=31.79%) at 16 bits
Files match @ 50% when reduced to 15.08 bits

Phase difference (full bandwidth): 0.992471064268582°
                 0-10,000Hz: 0.0759831135659693°
                 0-20,000Hz: 0.748855122275202°
                 0-24,000Hz: 0.992471064268582°
                 0-44,100Hz: 0.992471064268582°
                 0-48,000Hz: 0.992471064268582°
RMS of the difference of spectra: -167.16135102673dB
DONE!

Signature: 70007d5e4084977dd4c109c0b908bf45