DeltaWave null-testing audio comparator (beta)

TomCapraro · March 26, 2019

what can cause passing the precision of the offset from 18 to 30?

pkane2001 · March 26, 2019

1 minute ago, TomCapraro said:

found...!!!
if the precision of the offset from 18 to 30 is imposed, the problem occurs
if instead I leave it at 18 everything returns to post
this is why it worked for you and I didn't

Nice! Thank you. Now, I need to spend some time to understand why

TomCapraro · March 26, 2019

... all the joy over
I close the program again and even at 18 it is not bit perfect

TomCapraro · March 26, 2019

the problem is eliminated in this way
I launch the program and it is not bit perfect
active drift correction and is bit perfect
disable drift correction and it is incredibly bit perfect

TomCapraro · March 26, 2019

leaving the files loaded and making the comparison repeated times over 10 times 5 times is bit perfect and 5 times not perfect
provides random results

TomCapraro · March 26, 2019

1 hour ago, pkane2001 said:

Lasciatemi provare a disattivare il nuovo algoritmo di calcolo della deriva, poiché è l'unica differenza che posso vedere tra queste versioni. Pubblicherò una nuova versione da provare a breve.

2019-03-26 14:52:23.5003|DEBUG|Wave.WaveForm|Settings:
   Gain:False, Remove DC:False
   Non-linear Gain:False   EQ FFT Size:524288, EQ Frequency Cut: 0Hz - 0Hz, EQ Threshold: -160dB
   Correct Drift:False, Precision:30
   Upsample:False, Window:Hann
   Spectrum Window:Hann, Spectrum Size:65536
   Spectrogram Window:Hann, Spectrogram Size:32768, Spectrogram Steps:1024
   Dither:False
   Trim Silence:False

2019-03-26 14:52:33.8994|DEBUG|Wave.WaveForm|Progress [NOT Bit Perfect ] 0,63% Gain matching disabled. Phase offset=0ms Difference (rms) = -84,8dB [-89,19dBA] Correlated Null Depth=68,78dB [69,82dBA] Clock drift: 0 ppm, , 100%
2019-03-26 14:52:33.9150|INFO|Wave.WaveForm|Signature: bff25344e75ac416a03691702340650a

TomCapraro · March 26, 2019

2019-03-26 15:03:25.2516|DEBUG|Wave.WaveForm|Settings:
   Gain:False, Remove DC:False
   Non-linear Gain:False   EQ FFT Size:524288, EQ Frequency Cut: 0Hz - 0Hz, EQ Threshold: -160dB
   Correct Drift:False, Precision:30
   Upsample:False, Window:Hann
   Spectrum Window:Hann, Spectrum Size:65536
   Spectrogram Window:Hann, Spectrogram Size:32768, Spectrogram Steps:1024
   Dither:False
   Trim Silence:False

2019-03-26 15:03:29.3058|DEBUG|Wave.WaveForm|Progress [BIT PERFECT!] Gain matching disabled. Phase offset=0ms Difference (rms) = -300dB [-300dBA] Correlated Null Depth=300dB [300dBA] Clock drift: 0 ppm, , 100%
2019-03-26 15:03:29.3058|INFO|Wave.WaveForm|Signature: a6dd658afbae9719e993a86ddbdbaa2a

pkane2001 · March 26, 2019

1 hour ago, TomCapraro said:

leaving the files loaded and making the comparison repeated times over 10 times 5 times is bit perfect and 5 times not perfect
provides random results

Very strange, and I still can't reproduce it, even changing precision, reinstalling, etc. Random is even harder to explain.

Let's try the change I suggested earlier. Please uninstall the current version, then download and install this update:

https://drive.google.com/open?id=1gLXAumjPOekHLuE3UvuO6K0E3HaL8fvA

Please let me know if this solves the problem.

TomCapraro · March 26, 2019

25 minutes ago, pkane2001 said:

Very strange, and I still can't reproduce it, even changing precision, reinstalling, etc. Random is even harder to explain.

Let's try the change I suggested earlier. Please uninstall the current version, then download and install this update:

https://drive.google.com/open?id=1gLXAumjPOekHLuE3UvuO6K0E3HaL8fvA

Please let me know if this solves the problem.

now everything works perfectly
thank you...!!!

fas42 · March 27, 2019

Doing some exploring, with the latest drive.google version, and noted a peculiarity a couple of times: after some switching between panels, and zooming, the results panel redraws, with all info gone, apart from Phase differences. Can't get it to restore, even with Refresh button.

Also, the us checkbox - I presume microsecs, but somehow 4.3 secs translates to 0:42:517.7006 as a label.

pkane2001 · March 27, 2019

5 hours ago, fas42 said:

Doing some exploring, with the latest drive.google version, and noted a peculiarity a couple of times: after some switching between panels, and zooming, the results panel redraws, with all info gone, apart from Phase differences. Can't get it to restore, even with Refresh button.

Also, the us checkbox - I presume microsecs, but somehow 4.3 secs translates to 0:42:517.7006 as a label.

The results are cleared any time the plots are refreshed -- I'll fix it so that it preserves them when possible.

Somehow there was an extra zero in the seconds computation causing 10x larger # of seconds! I'll fix that, ASAP.

Regards,

-Paul

TomCapraro · March 27, 2019

I have seen that the algorithm that calculates the clock drift has been deactivated
Why not make it enabled and disabled through an option?
In this way we can verify the drift at will before making the comparison.
In any case, the software is fine.

Arpiben · March 27, 2019

Hi @pkane2001

Please do you mind bringing more details regarding:

Phase difference information:

The phase values are not provided above 48 kHz, any reason or limitation?
Phase difference (full bandwidth): 0,00147477068670476°
0-10,000Hz: 0,00423158331862641°
0-20,000Hz: 0,0030270942735478°
0-24,000Hz: 0,00277569901502977°
0-44,100Hz: 0,0021632420867899°
0-48,000Hz: 0,00201385547662473°

Matching Gain:

When selected the Reference Spectrogram (N°1) may be altered when it is not supposed to. Only comparison file or spectrogram N°2 is supposed to be modified. Chart inversion or samples truncation ?
Unmatched gain:
Matched gain:

Thanks.

pkane2001 · March 27, 2019

26 minutes ago, Arpiben said:

Hi @pkane2001

Please do you mind bringing more details regarding:

Phase difference information:

The phase values are not provided above 48 kHz, any reason or limitation?

Phase difference (full bandwidth): 0,00147477068670476°
0-10,000Hz: 0,00423158331862641°
0-20,000Hz: 0,0030270942735478°
0-24,000Hz: 0,00277569901502977°
0-44,100Hz: 0,0021632420867899°
0-48,000Hz: 0,00201385547662473°

Matching Gain:

When selected the Reference Spectrogram (N°1) may be altered when it is not supposed to. Only comparison file or spectrogram N°2 is supposed to be modified. Chart inversion or samples truncation ?

Unmatched gain:

Matched gain:

Thanks.

Reference should not be changed. Let me take a look.

pkane2001 · March 27, 2019

37 minutes ago, Arpiben said:

Hi @pkane2001

Please do you mind bringing more details regarding:

Phase difference information:

The phase values are not provided above 48 kHz, any reason or limitation?

Phase difference (full bandwidth): 0,00147477068670476°
0-10,000Hz: 0,00423158331862641°
0-20,000Hz: 0,0030270942735478°
0-24,000Hz: 0,00277569901502977°
0-44,100Hz: 0,0021632420867899°
0-48,000Hz: 0,00201385547662473°

These are just RMS values of phase differences between the two signals, for the specified frequency range.

The first number (full bandwidth) includes all the frequencies up to Nyquist. The other ranges are just the ones that I wanted to see I can certainly add others, if there's interest.

Arpiben · March 27, 2019

5 hours ago, pkane2001 said:

These are just RMS values of phase differences between the two signals, for the specified frequency range.

The first number (full bandwidth) includes all the frequencies up to Nyquist. The other ranges are just the ones that I wanted to see I can certainly add others, if there's interest.

Well if the partial ones are only a linear interpolation of the full bandwidth phase difference, IMO, it is useless adding ranges.

In the contrary, it helps characterising the different types of Frequency Offsets and/or Time/Phase Errors (cf picture)

I am not expecting DW to retrieve the exact ADCs' Quartz Oscillator clock drift at track or even at mix level...

But your tool can be interesting not only for audio digital comparisons.☺️

Rgds.

pkane2001 · March 27, 2019

38 minutes ago, Arpiben said:

Well if the partial ones are only a linear interpolation of the full bandwidth phase difference, IMO, it is useless adding ranges.

In the contrary, it helps characterising the different types of Frequency Offsets and/or Time/Phase Errors (cf picture)

I am not expecting DW to retrieve the exact ADCs' Quartz Oscillator clock drift at track or even at mix level...

But your tool can be interesting not only for audio digital comparisons.☺️

Rgds.

Partial frequency ranges are not interpolated, these are actual phase differences in the frequency domain, but only including those frequencies in the ranges given. The reason that might be useful is to judge the accuracy of the alignment of the two signals, and since we are talking about audio, proper phase alignment up to around 48KHz might be significant, although below 24KHz is what's really important. Also, phase errors at higher frequencies are larger. Measuring phase error between 0 and 192KHz may be an interesting academic exercise, but not useful in judging the audible quality of waveform alignment.

DW already measures clock drift error, although only linear is corrected. I thought about adding a chart showing the actual clock drift visually, since I already compute it.

Quote

But your tool can be interesting not only for audio digital comparisons.☺️

True. But I want to get that basic function working first

fas42 · March 27, 2019

52 minutes ago, pkane2001 said:

DW already measures clock drift error, although only linear is corrected. I thought about adding a chart showing the actual clock drift visually, since I already compute it.

Yes!! .... a big Yes!

BTW, DW rejects 8 bit files - stops immediately on loading.

pkane2001 · March 27, 2019

18 minutes ago, fas42 said:

Yes!! .... a big Yes!

BTW, DW rejects 8 bit files - stops immediately on loading.

Didn’t think anyone would want to process 8-bit files! 🧐

TomCapraro · March 27, 2019

16 minutes ago, pkane2001 said:

Non pensavo che qualcuno avrebbe voluto elaborare file a 8 bit! 🧐

😊

fas42 · March 27, 2019

24 minutes ago, pkane2001 said:

Didn’t think anyone would want to process 8-bit files! 🧐

Trying some exercises to have DW process large input files - for my laptop it starts to choke on the amount of data, and the m/c effectively grinds to a halt. So, seeing if reducing the amount of data along one axis will help to get the processing to complete.

pkane2001 · March 27, 2019

3 minutes ago, fas42 said:

Trying some exercises to have DW process large input files - for my laptop it starts to choke on the amount of data, and the m/c effectively grinds to a halt. So, seeing if reducing the amount of data along one axis will help to get the processing to complete.

8 bit files won't help -- it's limited by the number of samples and the amount of available memory. 1 bit or 8 bits or 24 bits all get converted to 64-bit internal format.

fas42 · March 28, 2019

Just to make you feel good, Paul, I did some simple manual processing of the Orig and 1st Gen Marley files in Audacity, and can confirm that the clock drift was indeed 0.63 ppm ... , .

fas42 · March 29, 2019

Another little exercise to see what the impact of alignment was - took the 32 bit, original Marley B, at 44.1k, chopped off the >20k content, upsampled to 88.2k, deleted a single sample at the start, and resampled to 44.1k. So, identical track, misaligned by exactly 0.5 sample - DW got very close, the delta Waveform was of the order of 80dB down - what did it sound like? The volume had to be boosted by about 90dB in DW - and there was the Bob Marley track, with the organ most heavily damaged, but otherwise easily identifiable. Was I listening to a true difference between the files - no, merely the limit of the processing capacity of the software at the moment to more precisly analyse what was going on ... unfortunately.

pkane2001 · March 29, 2019

20 minutes ago, fas42 said:

Another little exercise to see what the impact of alignment was - took the 32 bit, original Marley B, at 44.1k, chopped off the >20k content, upsampled to 88.2k, deleted a single sample at the start, and resampled to 44.1k. So, identical track, misaligned by exactly 0.5 sample - DW got very close, the delta Waveform was of the order of 80dB down - what did it sound like? The volume had to be boosted by about 90dB in DW - and there was the Bob Marley track, with the organ most heavily damaged, but otherwise easily identifiable. Was I listening to a true difference between the files - no, merely the limit of the processing capacity of the software at the moment to more precisly analyse what was going on ... unfortunately.

No, Frank, you were listening to the true difference between the files. The software removes unimportant differences such as volume level, clock drift, and phase offset to help with the comparison. By the way, in the next version I'm adding a feature that will help you prove this to yourself.

Remember what we audiophiles normally do for comparison... listen to two different tracks, at different levels, on a different day, sometimes at a different location. I think DeltaWave is just a tad better tool for precise comparisons, no?

By the way, if you know of any other "unimportant" differences that DW should remove, I'm all ears

DeltaWave null-testing audio comparator (beta)

Recommended Posts

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Create an account or sign in to comment

Create an account

Sign in