DeltaWave null-testing audio comparator (beta)

[fas42]

Now, here's an interesting one ... I deliberately picked one of the very worst performing loops in the Gearslutz results, Babyface_3.wav - RME Babyface (didier.brest) ... to see what DW would make of it. Well, after playing with settings a bit I got a remarkably good delta from it, and, most surprisingly, it did best with all drift parameters off. And what was particularly intriguing was that playing the difference WAV came very close to completely extinguishing the music as being recognisable - which wasn't the case when drift compensation was on.

 

Luck of the draw, an anomaly, how it's meant to work - or something else ... ?

 

DeltaWave v1.0.52, 2020-10-21T20:45:38.8851605+11:00
Reference:  Original2.wav[L] 5340666 samples 44100Hz 24bits, stereo, MD5=00
Comparison: Babyface_3.wav[L] 5335946 samples 44100Hz 24bits, stereo, MD5=00
Settings:
    Gain:True, Remove DC:True
    Non-linear Gain EQ:True    Non-linear Phase EQ: True
    EQ FFT Size:262144, EQ Frequency Cut: 0Hz - 0Hz, EQ Threshold: -160dB
    Correct Drift:False, 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=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -0.536dB   Comparison: -0.202dB
Initial RMS values Reference: -17.35dB   Comparison: -18.058dB

Null Depth=29.688dB
X-Correlation offset: -1 samples

Trimmed 0 samples ( 0.00ms) front, 0 samples ( 0.00ms end)

Final peak values Reference: -0.536dB   Comparison: -0.53dB
Final RMS values Reference: -17.346dB   Comparison: -17.346dB

Gain= -0.7283dB (0.9196x) DC=0 Phase offset=-0.022676ms (-1 samples)
Difference (rms) = -69.96dB [-82.85dBA]
Correlated Null Depth=70.49dB [78.49dBA]
Clock drift: 0 ppm

Files are NOT a bit-perfect match (match=2.44%) at 16 bits
Files are NOT a bit-perfect match (match=0.01%) at 24 bits
Files match @ 50.0014% when reduced to 11.31 bits

---- Phase difference (full bandwidth): 1.71940968179432°
    0-10kHz: 1.38°
    0-20kHz: 1.00°
    0-24kHz: 1.72°
---- Variable Group Delay. Frequency matched from 0Hz to 21.1kHz:
    1kHz = 997.2μs (358.97°)
    2kHz = 500.6μs (360.42°)
    4kHz = 251.4μs (361.99°)
    8kHz = 126.5μs (364.45°)
    16kHz = 64.1μs (369.13°)
Timing error (rms jitter): 582.4ns

RMS of the difference of spectra: -135.253077006884dB
gn=1.08746380823695, dc=8.36305684559057E-12, dr=0, of=-1

DONE!

Signature: 481ab9d56e0495a14c62f9b26db668cf

[pkane2001]

1 hour ago, fas42 said:

Now, here's an interesting one ... I deliberately picked one of the very worst performing loops in the Gearslutz results, Babyface_3.wav - RME Babyface (didier.brest) ... to see what DW would make of it. Well, after playing with settings a bit I got a remarkably good delta from it, and, most surprisingly, it did best with all drift parameters off. And what was particularly intriguing was that playing the difference WAV came very close to completely extinguishing the music as being recognisable - which wasn't the case when drift compensation was on.

 

Luck of the draw, an anomaly, how it's meant to work - or something else ... ?

 

DeltaWave v1.0.52, 2020-10-21T20:45:38.8851605+11:00
Reference:  Original2.wav[L] 5340666 samples 44100Hz 24bits, stereo, MD5=00
Comparison: Babyface_3.wav[L] 5335946 samples 44100Hz 24bits, stereo, MD5=00
Settings:
    Gain:True, Remove DC:True
    Non-linear Gain EQ:True    Non-linear Phase EQ: True
    EQ FFT Size:262144, EQ Frequency Cut: 0Hz - 0Hz, EQ Threshold: -160dB
    Correct Drift:False, 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=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -0.536dB   Comparison: -0.202dB
Initial RMS values Reference: -17.35dB   Comparison: -18.058dB

Null Depth=29.688dB
X-Correlation offset: -1 samples

Trimmed 0 samples ( 0.00ms) front, 0 samples ( 0.00ms end)

Final peak values Reference: -0.536dB   Comparison: -0.53dB
Final RMS values Reference: -17.346dB   Comparison: -17.346dB

Gain= -0.7283dB (0.9196x) DC=0 Phase offset=-0.022676ms (-1 samples)
Difference (rms) = -69.96dB [-82.85dBA]
Correlated Null Depth=70.49dB [78.49dBA]
Clock drift: 0 ppm

Files are NOT a bit-perfect match (match=2.44%) at 16 bits
Files are NOT a bit-perfect match (match=0.01%) at 24 bits
Files match @ 50.0014% when reduced to 11.31 bits

---- Phase difference (full bandwidth): 1.71940968179432°
    0-10kHz: 1.38°
    0-20kHz: 1.00°
    0-24kHz: 1.72°
---- Variable Group Delay. Frequency matched from 0Hz to 21.1kHz:
    1kHz = 997.2μs (358.97°)
    2kHz = 500.6μs (360.42°)
    4kHz = 251.4μs (361.99°)
    8kHz = 126.5μs (364.45°)
    16kHz = 64.1μs (369.13°)
Timing error (rms jitter): 582.4ns

RMS of the difference of spectra: -135.253077006884dB
gn=1.08746380823695, dc=8.36305684559057E-12, dr=0, of=-1

DONE!

Signature: 481ab9d56e0495a14c62f9b26db668cf

 

Non-linear EQ can correct both, phase and amplitude differences, including those caused by clock drift, as long as it's not excessive. So, yes, it's possible. What were the results with the drift correction turned on (and with non-linear EQ)?

 

[fas42]

All numbers a bit worse,

 

DeltaWave v1.0.52, 2020-10-22T08:28:18.7863960+11:00
Reference:  Original2.wav[L] 5340666 samples 44100Hz 24bits, stereo, MD5=00
Comparison: Babyface_3.wav[L] 5335946 samples 44100Hz 24bits, stereo, MD5=00
Settings:
    Gain:True, Remove DC:True
    Non-linear Gain EQ:True    Non-linear Phase EQ: True
    EQ FFT Size:262144, 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=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -0.536dB   Comparison: -0.202dB
Initial RMS values Reference: -17.35dB   Comparison: -18.058dB

Null Depth=29.688dB
X-Correlation offset: -1 samples
Drift computation quality, #1: Excellent (0.12μs)

Trimmed 0 samples ( 0.00ms) front, 0 samples ( 0.00ms end)

Final peak values Reference: -0.536dB   Comparison: -0.53dB
Final RMS values Reference: -17.346dB   Comparison: -17.346dB

Gain= -0.7283dB (0.9196x) DC=0 Phase offset=-0.025831ms (-1.139 samples)
Difference (rms) = -68.01dB [-72.56dBA]
Correlated Null Depth=70.52dB [74.98dBA]
Clock drift: 0 ppm

Files are NOT a bit-perfect match (match=2.03%) at 16 bits
Files are NOT a bit-perfect match (match=0.01%) at 24 bits
Files match @ 49.989% when reduced to 10.99 bits

---- Phase difference (full bandwidth): 1.84453894910125°
    0-10kHz: 1.43°
    0-20kHz: 1.14°
    0-24kHz: 1.84°
---- Variable Group Delay. Frequency matched from 0Hz to 21.1kHz:
    1kHz = 994.2μs (357.91°)
    2kHz = 497.7μs (358.33°)
    4kHz = 248.5μs (357.84°)
    8kHz = 123.7μs (356.22°)
    16kHz = 61.2μs (352.74°)
Timing error (rms jitter): 610.4ns

RMS of the difference of spectra: -118.162324190437dB
gn=1.08746687749017, dc=8.39185537705932E-12, dr=3.66E-09, of=-1.1391615581

DONE!

Signature: 851d18ca5aa3c7bbea54000d93b0e610

 

 

But what really stood out was in the playing of the delta waveform - the music component was highly intact, and stood out strongly against the 'random' noise; the subjective nulling of the former was far, far worse. The implication is that the drift correction was undoing the good things achieved by the processing relying on the other settings.

[pkane2001]

19 hours ago, fas42 said:

All numbers a bit worse,

 

DeltaWave v1.0.52, 2020-10-22T08:28:18.7863960+11:00
Reference:  Original2.wav[L] 5340666 samples 44100Hz 24bits, stereo, MD5=00
Comparison: Babyface_3.wav[L] 5335946 samples 44100Hz 24bits, stereo, MD5=00
Settings:
    Gain:True, Remove DC:True
    Non-linear Gain EQ:True    Non-linear Phase EQ: True
    EQ FFT Size:262144, 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=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -0.536dB   Comparison: -0.202dB
Initial RMS values Reference: -17.35dB   Comparison: -18.058dB

Null Depth=29.688dB
X-Correlation offset: -1 samples
Drift computation quality, #1: Excellent (0.12μs)

Trimmed 0 samples ( 0.00ms) front, 0 samples ( 0.00ms end)

Final peak values Reference: -0.536dB   Comparison: -0.53dB
Final RMS values Reference: -17.346dB   Comparison: -17.346dB

Gain= -0.7283dB (0.9196x) DC=0 Phase offset=-0.025831ms (-1.139 samples)
Difference (rms) = -68.01dB [-72.56dBA]
Correlated Null Depth=70.52dB [74.98dBA]
Clock drift: 0 ppm

Files are NOT a bit-perfect match (match=2.03%) at 16 bits
Files are NOT a bit-perfect match (match=0.01%) at 24 bits
Files match @ 49.989% when reduced to 10.99 bits

---- Phase difference (full bandwidth): 1.84453894910125°
    0-10kHz: 1.43°
    0-20kHz: 1.14°
    0-24kHz: 1.84°
---- Variable Group Delay. Frequency matched from 0Hz to 21.1kHz:
    1kHz = 994.2μs (357.91°)
    2kHz = 497.7μs (358.33°)
    4kHz = 248.5μs (357.84°)
    8kHz = 123.7μs (356.22°)
    16kHz = 61.2μs (352.74°)
Timing error (rms jitter): 610.4ns

RMS of the difference of spectra: -118.162324190437dB
gn=1.08746687749017, dc=8.39185537705932E-12, dr=3.66E-09, of=-1.1391615581

DONE!

Signature: 851d18ca5aa3c7bbea54000d93b0e610

 

 

But what really stood out was in the playing of the delta waveform - the music component was highly intact, and stood out strongly against the 'random' noise; the subjective nulling of the former was far, far worse. The implication is that the drift correction was undoing the good things achieved by the processing relying on the other settings.

 

Non-linear EQ is a strange beast. It corrects for all kinds of errors, as long as they are related to the frequency domain. In effect, it's like a graphic equalizer with a separate slider for each frequency bin in the FFT. Beyond correcting for amplitude, it can also correct phase errors. This works well when the same type of error occurs throughout the recorded track, for example a simple variable group delay. But, if the errors are not static and change over time, the non-linear EQ will not be able to correct these.

 

If you have a track that corrects better with just the non-linear EQ than with drift correction + non-linear EQ, that means that the track has very little or no clock drift, but a large non-linear error in phase and/or amplitude.

[fas42]

Just to mention again that the Manual Corrections window has lost the plot 😉, badly. Upsampling of a short bit, and both DC and Gain Factor were set to nonsensical numbers, when the only setting touched was Offset. Which means I can't trust what this processing is doing ...

[pkane2001]

1 minute ago, fas42 said:

Just to mention again that the Manual Corrections window has lost the plot 😉, badly. Upsampling of a short bit, and both DC and Gain Factor were set to nonsensical numbers, when the only setting touched was Offset. Which means I can't trust what this processing is doing ...

The fix is coming shortly :)

[pkane2001]

New version available: 1.0.53b

 

Changes are as follows:

• Added: support to process .mp4, .m4a, and .alac files
• Added: 2M FFT size for non-linear EQ settings
• Added: new Cosine and Flattop FFT windows
• Fixed: when using the manual adjustment window Gain setting was ignored and recalculated each time (@fas42)
• Changed: group delay computation changed to reduce time and memory needed to process
• Added: double-clicking on the trim label “End” or “Take” will now switch between these two modes. This setting can also be changed in Settings window

[pkane2001]

duplicate

[TomCapraro]

Congratulations Paul, it's not that the previous version was "bad" but ... here perfection has been reached (obviously if the D / A-> A / D setup is performing) since with the necessary corrections the difference signal reaches the electronic noise.
The 20khz step, present in previous versions, is smoothed out at the same level.
You can improve everything in the future if necessary ... but not the result of the difference signal

[fas42]

Another curiousity - one of the 'best' Gearslutz performers, and playing with Manual Corrections, getting some remarkable numbers,

 

 

Is this something real - note that at the end I'm playing with very fine changes of the offset, which are so precise that they don't get shown in the table; but altering the value does indeed alter the Corr Null - or some anomaly of the processing?

[pkane2001]

46 minutes ago, fas42 said:

Another curiousity - one of the 'best' Gearslutz performers, and playing with Manual Corrections, getting some remarkable numbers,

 

 

Is this something real - note that at the end I'm playing with very fine changes of the offset, which are so precise that they don't get shown in the table; but altering the value does indeed alter the Corr Null - or some anomaly of the processing?

 

Yeah, that's real. Correlated null is a very sensitive calculation. Still, I'd pay attention to RMS null much more.

 

I use Correlated Null is as an indicator that DeltaWave has done a good job of lining up the samples between two tracks. I.e., eliminating timing errors. If it's a low number, then something is likely wrong with phase or clock drift calculation, or DW didn't find the right values. 

[fas42]

Paul, could the Delta Spectrogram be in the Lock/Unlock group of Spectrograms 1 and 2?

[pkane2001]

2 hours ago, fas42 said:

Paul, could the Delta Spectrogram be in the Lock/Unlock group of Spectrograms 1 and 2?

 

Yes, Frank. I changed Delta to be separate from the other two for a reason a while back, but now can't remember why 🤷‍♂️. 

 

If I don't think of a good reason to keep them separated, I'll make the change.

 

[fas42]

Don't you hate it ... when you do something which made sense at the time - and then wonder, further along, why you did it, 😜? Of course, you go back to the old setup, with more time spent doing that - and Murphy guarantees, next day, that it becomes blindingly obvious why you went to the effort, the first time ... 🤪

[mehmethandisbudak]

Thanks for this great application pkanee2001. I can see all details when comparing my mixes even I don't understand many parameters and graphs..etc. 

I want to ask a question. Let's say you have 2 files to compare and they sound very similar. You can't hear the difference with blind test. "You can't hear the difference" doesn't mean that other people can't hear it,too. So here is my question : 

Is there anyway to make an assumption about the difference between files is hearable or not  by checking parameters and graphs? Is there any  kind of clue in those graphs, parameters that difference is in hearable zone or not?

[pkane2001]

2 hours ago, mehmethandisbudak said:

Thanks for this great application pkanee2001. I can see all details when comparing my mixes even I don't understand many parameters and graphs..etc. 

I want to ask a question. Let's say you have 2 files to compare and they sound very similar. You can't hear the difference with blind test. "You can't hear the difference" doesn't mean that other people can't hear it,too. So here is my question : 

Is there anyway to make an assumption about the difference between files is hearable or not  by checking parameters and graphs? Is there any  kind of clue in those graphs, parameters that difference is in hearable zone or not?

 

There are a few things you can use to judge whether the difference is audible or not.

 

1. The delta file, the difference between the reference and comparison files. You can play this file at a loud level (use the volume control if it's too low, but be careful not to blow your speakers!) If you have to add 50dB+ to the volume of the delta file to hear it, then the differences will not be audible. You can also listen to what remains in the delta file. If it's all noise and no music, or music is well below the level of noise, you will not hear the difference.

 

2. The RMS Null dB value, and especially dbA value can be useful. If these are below -70dB, you will not hear the difference. These are average values, so this doesn't mean there can't be one or a few spots where the differences are larger, but overall, this means the level of differences is below audibility. 

 

3. Look at the delta waveform and you can judge by eye if there's sufficiently large differences that might be audible. For example, this delta file represents something that just barely at the edge of being audible, possibly when playing really loud or in a very, very quiet environment, and with very young ears :)

 

 

 

 

 

 

[fas42]

So, all amplifiers "sound the same"? ... 😉

 

This website test, https://alpha-audio.net/2020/08/live-mass-test-amplifiers-2500-euro/, used the same setup, only changing the amplifier - and recorded the speaker output. Available in FLAC, DW is good enough to pick what's going on, and here's a shot of what the speaker was putting out, after matching to a correlated null of 51dB,

 

 

One amp was Naim, one was Yamaha - have a guess which was which?

[pkane2001]

10 hours ago, fas42 said:

So, all amplifiers "sound the same"? ... 😉

 

This website test, https://alpha-audio.net/2020/08/live-mass-test-amplifiers-2500-euro/, used the same setup, only changing the amplifier - and recorded the speaker output. Available in FLAC, DW is good enough to pick what's going on, and here's a shot of what the speaker was putting out, after matching to a correlated null of 51dB,

 

 

One amp was Naim, one was Yamaha - have a guess which was which?

All well designed amplifiers do, unless presented with loads they can’t handle. Amps designed for specific sound signature, of course, can sound different.

[pkane2001]

Updated DeltaWave version 1.0.54 is now available.

Changes in 1.0.54b

• Added: new FFT window types, including multiple Kaiser
• Added: FFT Window explorer with time- and frequency-domain plots and measurements
• Changed: Delta Spectrogram window scale is now locked to Spectrogram 1 and 2
• Fixed: under one specific combination of filter settings, filter 1 @start was being ignored

To activate FFT Window explorer, click on the question mark in settings, under Spectrum:

 

You can explore and compare other windows once there:

[fas42]

2 hours ago, pkane2001 said:

Added: FFT Window explorer with time- and frequency-domain plots and measurements

 

Very nifty !!

 

2 hours ago, pkane2001 said:

• Changed: Delta Spectrogram window scale is now locked to Spectrogram 1 and

 

 

A problem ... the Delta Spectrogram is now a sea of red, literally, on one of the best matching Gearslutz's captures,

 

DeltaWave v1.0.54, 2020-11-01T11:10:41.8126298+11:00
Reference:  Original2.wav[L] 2211840 samples 44100Hz 24bits, stereo, MD5=00
Comparison: Forssell DAC from AD245(master) bis -20 dbu.wav[L] 2277376 samples 44100Hz 24bits, stereo, MD5=00
Settings:
    Gain:True, Remove DC:True
    Non-linear Gain EQ:True    Non-linear Phase EQ: True
    EQ FFT Size:2097152, 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=72s, End=0s
Discarding Comparison: Start=73s, End=0s

Initial peak values Reference: -0.536dB   Comparison: -4.666dB
Initial RMS values Reference: -16.193dB   Comparison: -20.387dB

Null Depth=12.341dB
X-Correlation offset: 43960 samples
Drift computation quality, #1: Excellent (0.03μs)

Trimmed 0 samples ( 0.00ms) front, 0 samples ( 0.00ms end)

Final peak values Reference: -0.536dB   Comparison: -0.532dB
Final RMS values Reference: -16.175dB   Comparison: -16.175dB

Gain= -3.9992dB (0.631x) DC=0 Phase offset=996.833481ms (43960.357 samples)
Difference (rms) = -75.18dB [-83.79dBA]
Correlated Null Depth=78.17dB [81.33dBA]
Clock drift: 0 ppm

Files are NOT a bit-perfect match (match=4.29%) at 16 bits
Files are NOT a bit-perfect match (match=0.02%) at 24 bits
Files match @ 49.9712% when reduced to 12.19 bits

---- Phase difference (full bandwidth): 5.89183213693535°
    0-10kHz: 0.46°
    0-20kHz: 0.70°
    0-24kHz: 5.89°
---- Variable Group Delay. Frequency matched from 0Hz to 21.1kHz:
    1kHz = 20ns (0.01°)
    2kHz = 21.6ns (0.02°)
    4kHz = 21.9ns (0.03°)
    8kHz = 54.1ns (0.16°)
    16kHz = 80.6ns (0.46°)
Timing error (rms jitter): 556.2ns

RMS of the difference of spectra: -124.777713900001dB
gn=1.58474981021717, dc=2.94481385281932E-07, dr=0, of=43960.3565086558

DONE!

Signature: 04448bec2a835635e19278c8c2536a80

 

 

53b was working fine, here.

 

 

[pkane2001]

1 hour ago, fas42 said:

 

Very nifty !!

 

 

A problem ... the Delta Spectrogram is now a sea of red, literally, on one of the best matching Gearslutz's captures,

 

DeltaWave v1.0.54, 2020-11-01T11:10:41.8126298+11:00
Reference:  Original2.wav[L] 2211840 samples 44100Hz 24bits, stereo, MD5=00
Comparison: Forssell DAC from AD245(master) bis -20 dbu.wav[L] 2277376 samples 44100Hz 24bits, stereo, MD5=00
Settings:
    Gain:True, Remove DC:True
    Non-linear Gain EQ:True    Non-linear Phase EQ: True
    EQ FFT Size:2097152, 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=72s, End=0s
Discarding Comparison: Start=73s, End=0s

Initial peak values Reference: -0.536dB   Comparison: -4.666dB
Initial RMS values Reference: -16.193dB   Comparison: -20.387dB

Null Depth=12.341dB
X-Correlation offset: 43960 samples
Drift computation quality, #1: Excellent (0.03μs)

Trimmed 0 samples ( 0.00ms) front, 0 samples ( 0.00ms end)

Final peak values Reference: -0.536dB   Comparison: -0.532dB
Final RMS values Reference: -16.175dB   Comparison: -16.175dB

Gain= -3.9992dB (0.631x) DC=0 Phase offset=996.833481ms (43960.357 samples)
Difference (rms) = -75.18dB [-83.79dBA]
Correlated Null Depth=78.17dB [81.33dBA]
Clock drift: 0 ppm

Files are NOT a bit-perfect match (match=4.29%) at 16 bits
Files are NOT a bit-perfect match (match=0.02%) at 24 bits
Files match @ 49.9712% when reduced to 12.19 bits

---- Phase difference (full bandwidth): 5.89183213693535°
    0-10kHz: 0.46°
    0-20kHz: 0.70°
    0-24kHz: 5.89°
---- Variable Group Delay. Frequency matched from 0Hz to 21.1kHz:
    1kHz = 20ns (0.01°)
    2kHz = 21.6ns (0.02°)
    4kHz = 21.9ns (0.03°)
    8kHz = 54.1ns (0.16°)
    16kHz = 80.6ns (0.46°)
Timing error (rms jitter): 556.2ns

RMS of the difference of spectra: -124.777713900001dB
gn=1.58474981021717, dc=2.94481385281932E-07, dr=0, of=43960.3565086558

DONE!

Signature: 04448bec2a835635e19278c8c2536a80

 

 

53b was working fine, here.

 

 

 

Ah, well, that's one of the reasons I had them unlocked, originally! :) Since the plots are all tied together now, the scaling of the display is the same for all three spectrograms. You'll need to go into the adjustments menu on the delta spectrogram and set the range of values you'd like to see to be narrower than the default. Change it to say -20dB minimum and 0dB maximum:

 

[fas42]

I was thrown for a bit - because I don't often look at this window ... finally, it dawned on me 🙃. The scaling of 53b was -100 to +100, which meant I was usually looking at green, using the Rainbow palette, for 0dB, midway between -5 and +5dB ... now that I've lost the + side of things, how should I read what the display says?

[pkane2001]

11 hours ago, fas42 said:

I was thrown for a bit - because I don't often look at this window ... finally, it dawned on me 🙃. The scaling of 53b was -100 to +100, which meant I was usually looking at green, using the Rainbow palette, for 0dB, midway between -5 and +5dB ... now that I've lost the + side of things, how should I read what the display says?

 

The range (min/max) that you specify in the spectrogram adjustment window defines how the colors are mapped to values. If the range  is -100 to 100db, all the colors will be assigned to that range. Anything outside that range will remain at the same color. If you pick a range of -10 to 0db, then all the colors in the palette will be mapped to that range. Think of it as zooming in to the values in the spectrogram, the smaller the range that you define, the more colorful values in that range will become, easier to see or distinguish. But note that all the values outside the range will appear as a single color!  I guess another way to think about it as compression in audio, but applied to the spectrogram colors ;)

 

In a delta spectrogram, say the values range between +5 and -30db. If you use the -200/+200dB range, you will be so far zoomed out on the colors that almost everything will appear the same. OTOH, if you set the range to say, +10 and -35dB, you'll see a lot more details in the color, because you'll be using many more of the colors to represent a much smaller numerical range of interest.

 

[pkane2001]

Fat fingers

[fas42]

What I was getting at, to take an example with respect to 53b, if Spec1 at Position A was -100dB, and for Spec2 was -103dB; and Position B value of Spec1 was also 100dB, and Spec2 was -97dB; then the DiffSpec would show this clearly, for -3 to +3 range. In 54b, with DiffSpec range of say -6 to 0, should the exact same picture be shown as for 53b?