Beta-test: DeltaWave Null Comparison software

[pkane]

Version 1.0.43 download is now available.

Changes in 1.0.43b

• Fix for phase and spectrum plots showing very small differences between runs
• Fix for volume going down after a second of play from the main window

The fix for the first item involved redoing a lot of the FFT logic, as the library I'm using didn't support this change (and, in fact, caused this problem). Let me know if you find any issues!

 

[zermak]

Thanks for the hard work!
I don't use it as deeply as others do around here but it is a great piece of software!

 

[Earfonia]

Version 1.0.43 download is now available.

Changes in 1.0.43b

• Fix for phase and spectrum plots showing very small differences between runs
• Fix for volume going down after a second of play from the main window

The fix for the first item involved redoing a lot of the FFT logic, as the library I'm using didn't support this change (and, in fact, caused this problem). Let me know if you find any issues!

I've been looking for this kind of app, definitely will try it. Thanks a lot for the hard work!

 

[pkane]

An experimental version for your testing pleasure

Version 1.0.45 introduces a new option to enable the alignment of simple, repetitive waveforms, such as sine-waves, multi-tone, square waves, etc. The option is enabled in settings:




The result is a better alignment of Comparison to Reference for simple test captures. For example, two 1kHz sine waves, raw, both captured with Apogee firewire audio interface in loopback mode:



And now aligned:




For single sine-wave tone captures, this also produces some measurements in the Results tab that includes THD, THD+N and Dynamic Range values for the comparison file. These numbers need to be verified, they seem to mostly match what I've been able to measure through other means (REW), but I've not done a lot of testing with different devices:

---Measurements (for a simple sine-wave only)---

Comparison DR = 106.6dB
Comparison THD+N = -108.41dB
Comparison THD   = -117.39dB
    H1 (1000Hz) = -1.4dB
    H2 (2000Hz) = -123.24dB
    H3 (3000Hz) = -118.64dB
    H4 (4000Hz) = -139.14dB
    H5 (5000Hz) = -126.02dB
    H6 (6000Hz) = -150.53dB
    H7 (7000Hz) = -139.52dB
    H9 (9000Hz) = -152.41dB

Here's an example of a multi-tone waveform, raw:




And now aligned using the new setting:




Here's the spectrum of the above waveforms:

 

[hyperknot]

First of all thanks for developing this nice piece of software! I have a probably beginner question to ask, with a not so beginner answer:
How can I prove or disprove that different audio players sound different on my Macbook.

My idea was to take a 30 second real world audio sample sample and play it 2x in different programs. iTunes (or any Core Audio) and in Amarra / Amarra SQ / a few other specialised players. Playing 2x is so that I'd have a reference and see what are the results of identical hardware/software.

Then I'd need an other computer with a sound card and a line in with a medium or better quality ADC. I have a Windows PC with a PCI soundcard like this, is this card adequate to record this (RMAA DAC/ADC loopback)?

What should I use for recording settings if my playback consists 100% of 16/44.1 material? Should I record in 24/44.1?

Also, should I record real world music or something like RMAA samples? (My theory is that DeltaWave cannot work on sweep tones for example, as there a frequency change can be understood as a non-linear time delay, so real world is better.)

 

[pkane]

First of all thanks for developing this nice piece of software! I have a probably beginner question to ask, with a not so beginner answer:
How can I prove or disprove that different audio players sound different on my Macbook.

My idea was to take a 30 second real world audio sample sample and play it 2x in different programs. iTunes (or any Core Audio) and in Amarra / Amarra SQ / a few other specialised players. Playing 2x is so that I'd have a reference and see what are the results of identical hardware/software.

Then I'd need an other computer with a sound card and a line in with a medium or better quality ADC. I have a Windows PC with a PCI soundcard like this, is this card adequate to record this (RMAA DAC/ADC loopback)?

What should I use for recording settings if my playback consists 100% of 16/44.1 material? Should I record in 24/44.1?

Also, should I record real world music or something like RMAA samples? (My theory is that DeltaWave cannot work on sweep tones for example, as there a frequency change can be understood as a non-linear time delay, so real world is better.)

Yes, basically you need to record the analog output of your Macbook. To do that, you'll need to feed into an ADC input on another PC or an external ADC with a digital connector to you Macbook (USB, etc.). That sound card looks good enough for testing.

Use playback sample rate and bit rate that you normally use for listening, as that's what you want to compare. For ADC/recording, I prefer to record at 24 bit with sample rate of 96kHz. This covers everything up to about 48kHz so you can see any issues/artifacts just beyond the audible spectrum.

You can record whatever sound file works for you, although some large-scale orchestral music is probably a good idea. Gearslutz forum has a long-running thread where they are recording and reporting results of loopback of DAC/ADC converters using the same sound file:

https://www.gearslutz.com/board/showpost.php?p=14093948&postcount=1776

It might be interesting to use the same file for your own testing, because you can then also compare your results to what others have measured. The actual file used for testing is referenced in that post: http://www.mediafire.com/file/5hg6wl6ygql7217/Original2.wav/file

Record two to three captures using each player, as this will make it easier to see if the differences are random or between different software players.

Good luck!

 

[ofrappier]

Hi,

2 files one played with software player, one recorded with virtual cable : 100 % bit perfect
but...
Phase problem

how is possible ?

rapport :

DeltaWave v1.0.45, 2019-10-07T23:59:31.2100670+02:00
Reference: TEST-DJ-SOFT-ORIG.wav[L] 1851392 samples 44100Hz 16bits, stereo, MD5=00
Comparison: 20191007 2326 Recording.wav[L] 2211840 samples 44100Hz 16bits, stereo, MD5=00
Settings:
Gain:True, Remove DC:True
Non-linear Gain EQ:False Non-linear Phase EQ: False
EQ FFT Size:65536, EQ Frequency Cut: 0Hz - 0Hz, EQ Threshold: -160dB
Correct Drift:True, Precision:30
Non-Linear drift Correction:False
Upsample:False, Window:Hann
Spectrum Window:Hann, Spectrum Size:32768
Spectrogram Window:Hann, Spectrogram Size:4096, Spectrogram Steps:2048
Dither:False
Trim Silence:False
Enable Simple Waveform Measurement: False

Discarding Reference: Start=10s, End=10s
Discarding Comparison: Start=10s, End=10s

Initial peak values Reference: -0,045dB Comparison: -0,045dB
Initial RMS values Reference: -9,92dB Comparison: -9,821dB

Null Depth=8,208dB
X-Correlation offset: -78131 samples
Drift computation quality, #1: Excellent (0μs)


Trimmed 0 samples ( 0,00ms) front, 0 samples ( 0,00ms end)


Final peak values Reference: 0,578dB Comparison: 0,578dB
Final RMS values Reference: -9,921dB Comparison: -9,921dB

Gain= 0dB (1x) DC=0 Phase offset=-1771,678005ms (-78131 samples)
Difference (rms) = -316,18dB [-317,68dBA]
Correlated Null Depth=306,29dB [302,03dBA]
Clock drift: 0 ppm


Files are a bit-perfect match at 16 bits


---- Phase difference (full bandwidth): 20061,6930700685°
0-10kHz: 11082,30°
0-20kHz: 18815,75°
0-24kHz: 20061,69°
Timing error (rms jitter): 0sec

RMS of the difference of spectra: -356,230520281272dB
gn=1, dc=0, dr=0, of=-78131

DONE!

Signature: b50373af40eb5472abbe4753d76f36cc

 

[pkane]

Hi,

2 files one played with software player, one recorded with virtual cable : 100 % bit perfect
but...
Phase problem

how is possible ?

rapport :

DeltaWave v1.0.45, 2019-10-07T23:59:31.2100670+02:00
Reference: TEST-DJ-SOFT-ORIG.wav[L] 1851392 samples 44100Hz 16bits, stereo, MD5=00
Comparison: 20191007 2326 Recording.wav[L] 2211840 samples 44100Hz 16bits, stereo, MD5=00
Settings:
Gain:True, Remove DC:True
Non-linear Gain EQ:False Non-linear Phase EQ: False
EQ FFT Size:65536, EQ Frequency Cut: 0Hz - 0Hz, EQ Threshold: -160dB
Correct Drift:True, Precision:30
Non-Linear drift Correction:False
Upsample:False, Window:Hann
Spectrum Window:Hann, Spectrum Size:32768
Spectrogram Window:Hann, Spectrogram Size:4096, Spectrogram Steps:2048
Dither:False
Trim Silence:False
Enable Simple Waveform Measurement: False

Discarding Reference: Start=10s, End=10s
Discarding Comparison: Start=10s, End=10s

Initial peak values Reference: -0,045dB Comparison: -0,045dB
Initial RMS values Reference: -9,92dB Comparison: -9,821dB

Null Depth=8,208dB
X-Correlation offset: -78131 samples
Drift computation quality, #1: Excellent (0μs)


Trimmed 0 samples ( 0,00ms) front, 0 samples ( 0,00ms end)


Final peak values Reference: 0,578dB Comparison: 0,578dB
Final RMS values Reference: -9,921dB Comparison: -9,921dB

Gain= 0dB (1x) DC=0 Phase offset=-1771,678005ms (-78131 samples)
Difference (rms) = -316,18dB [-317,68dBA]
Correlated Null Depth=306,29dB [302,03dBA]
Clock drift: 0 ppm


Files are a bit-perfect match at 16 bits


---- Phase difference (full bandwidth): 20061,6930700685°
0-10kHz: 11082,30°
0-20kHz: 18815,75°
0-24kHz: 20061,69°
Timing error (rms jitter): 0sec

RMS of the difference of spectra: -356,230520281272dB
gn=1, dc=0, dr=0, of=-78131

DONE!

Signature: b50373af40eb5472abbe4753d76f36cc

Hi,

Can you please try turning on 'Auto-trim' setting in options? I suspect that because the two files are initially misaligned, there are parts of the track at the start or end where they don't overlap. This can cause significant differences in computation. If you turn on trim, it should remove these 'unmatched' parts of the track(s).

 

[ofrappier]

Hi,

Can you please try turning on 'Auto-trim' setting in options? I suspect that because the two files are initially misaligned, there are parts of the track at the start or end where they don't overlap. This can cause significant differences in computation. If you turn on trim, it should remove these 'unmatched' parts of the track(s).

Hi,

It work with autotrim , with a full classical file.
It doesn't work with autotrim with single waveform.

thanks

 

[hyperknot]

Some bug reports related to Report Cruncher:
1. The - (minus) button doesn't work
2. The generated reports are not saved during the process. I started a batch and wanted to see how is the first one but there was nothing in the folder even when it was processing the second one.
3. The * and > are not clear what they do.
4. Once I added a file there is no way to remove it (double clicking doesn't overwrite it and there is no way to remove it).
5. The sorting by columns don't work. (It sorts some, but not all of the rows).

 

[hyperknot]

+1, very important for Report Cruncher, it stops at "Residual drift error too large, couldn't find a solution, do you want to try an alternative drift..."

 

[pkane]

Some bug reports related to Report Cruncher:
1. The - (minus) button doesn't work
2. The generated reports are not saved during the process. I started a batch and wanted to see how is the first one but there was nothing in the folder even when it was processing the second one.
3. The * and > are not clear what they do.
4. Once I added a file there is no way to remove it (double clicking doesn't overwrite it and there is no way to remove it).
5. The sorting by columns don't work. (It sorts some, but not all of the rows).

Report cruncher has been dormant for at least 6-8 months, as I was working on other parts of DW. I guess it's time for me to get back to it Thanks for reporting, I'll fix these ASAP.

 

[hyperknot]

Report cruncher has been dormant for at least 6-8 months, as I was working on other parts of DW. I guess it's time for me to get back to it Thanks for reporting, I'll fix these ASAP.

Thanks. Maybe 2. is not an issue, right now it saves files continuously.

 

[hyperknot]

BTW, why is a report 23 MB? I wanted to post them for quick checking but they are as big as the source audio files!

 

[pkane]

BTW, why is a report 23 MB? I wanted to post them for quick checking but they are as big as the source audio files!

Spectrogram images don't compress well....

 

[ofrappier]

Hi Pkane,

Thanks to your software, i have made a digital crossover for my 2 subwoofer, with no phase, perfect linear phase :

For the 2 SAT :

For the 2 subwoofer :

This Fabfilter plugin allow brickwall crossover without phase change, with linear phase option, low or medium (same results), low great latency :

routed in a MOTU 8A and OSX : Output 1/2 and Output 3/4 of MOTU 8A

And the result of this crossover : no phase change and good spectrogram !

Superb CROSSOVER

Thanks to your software, that allow to calibrate DIGITAL SOFTWARE CROSSOVER.

And thanks your software, i know that OSX audio engine, loopback and audio hi jack software are bit perfect before the crossover (100%)

 

[hyperknot]

@ofrappier which tools are you using to apply filter on system wide sounds in macOS? I'm currently trying SoundSource from RougeAmoeba (https://rogueamoeba.com/soundsource/) to add AudioUnit filters and it seems to work. Are you using an AudioUnit filter for crossover?

 

[ofrappier]

@ofrappier which tools are you using to apply filter on system wide sounds in macOS? I'm currently trying SoundSource from RougeAmoeba (https://rogueamoeba.com/soundsource/) to add AudioUnit filters and it seems to work. Are you using an AudioUnit filter for crossover?

Hi,

Loopback : (to the main sound to the virtual DSP with Audio hijack)
https://rogueamoeba.com/loopback/
Audio hijack : (settings with low latency)
https://rogueamoeba.com/audiohijack/
Fabfilter pro Q-3 : equa. with linear PHASE set to low (for great latency)
https://www.fabfilter.com/products/pro-q-3-equalizer-plug-in
MOTU 8A : you can set in AUDIO HIJACK, the output : sat to 5/6 and sub to 5/8

MOTU 8A USB3 Routing :

Regards,

Olivier

 

[hyperknot]

Wow, that MOTU panel looks complicated. But in theory you could do this with any 6 or 8 channel sound cards, like an Creative Sound Blaster, couldn't you?

 

[ofrappier]

Hi,

I use Motu 8a because it's USB 32 bits float good driver in OSX, and sound perfect to me, with high balanced output level.
I would avoir all creative Sound blaster products.