Beta-test: DeltaWave Null Comparison software

[Rock Rabbit]

For me better this color selection on graphs, I have very limited vision

 

[pkane]

View attachment 22709
For me better this color selection on graphs, I have very limited vision

What do you think, better?

 

[Blumlein 88]

Did the graph for Delta of Spectra get flipped upside down again in the 1.06 version?

 

[pkane]

Did the graph for Delta of Spectra get flipped upside down again in the 1.06 version?

Looks like I never flipped it in the first place It's still reference spectrum minus the comparison. I'll change in the next update.

 

[pkane]

Quick update Version 1.0.7 contains the following changes:

• Added high-contrast plot display option (upper right corner)
• Reversed the direction of the plot in Delta of Spectra, so that it is now Comparison minus Reference
• Added tooltips to controls on the main, settings, and Comparator screens
• Fixed the plot refresh button as it caused certain elements of the plots to be hidden

 

[Blumlein 88]

Quick update Version 1.0.7 contains the following changes:

• Added high-contrast plot display option (upper right corner)
• Reversed the direction of the plot in Delta of Spectra, so that it is now Comparison minus Reference
• Added tooltips to controls on the main, settings, and Comparator screens
• Fixed the plot refresh button as it caused certain elements of the plots to be hidden

Cool, like the dark background.

I've a question about phase shown in the results tab and the delta phase graph. How are you doing that or maybe what is it really showing us? Sometimes it appears to be like I would expect while others it is all over the place even with a good, good null. One example is comparing two consecutive tests of the same file on the same gear. Nulling is very deep around 100 db, delta spectrum is almost completely flat, but phase will vary in wild swings of plus and minus 150 degrees. It appears to be effected by how closely matched in samples of offset at the beginning of a file is vs the reference. Or maybe I need to try it more before I can say that is true.

 

[pkane]

Cool, like the dark background.

I've a question about phase shown in the results tab and the delta phase graph. How are you doing that or maybe what is it really showing us? Sometimes it appears to be like I would expect while others it is all over the place even with a good, good null. One example is comparing two consecutive tests of the same file on the same gear. Nulling is very deep around 100 db, delta spectrum is almost completely flat, but phase will vary in wild swings of plus and minus 150 degrees. It appears to be effected by how closely matched in samples of offset at the beginning of a file is vs the reference. Or maybe I need to try it more before I can say that is true.

Delta Phase display is the counterpart of delta of spectra plot. The result of an FFT is a set of frequency bins, each representing an amplitude and a phase. Delta of Spectra plot represents the difference in magnitude, while Delta of Phase plot represents the difference in phase for each of these bins.

Delta Phase is very sensitive to small differences and can easily result in wild swings if the data is noisy or doesn't match well for other reasons. The setting of phase limit:

is used to limit the wild swings for bins where magnitude is below this limit. In other words, any phase differences for bins where magnitude is below -120dB will not appear on the delta phase plot. Try changing this number to -100dB if the RMS delta is at about that value to see if phase differences become more well behaved.

Does this make sense?

Regards,

-Paul

 

[Rock Rabbit]

Nice, very nice graphics now
The DSD demodulator seems to have the right amount of ultrasonics before the low pass filter (50 kHz or less) with -30 dB @100 kHz. For example the AK4490 makes first a digital filter to -15 dB, then an external 3rd order Bessel.
What kind of filter use Deltawave?
The demodulated DSD could make some benefits in correlation (to a PCM wave) using the 50 kHz filter?

 

[pkane]

Nice, very nice graphics now
The DSD demodulator seems to have the right amount of ultrasonics before the low pass filter (50 kHz or less) with -30 dB @100 kHz. For example the AK4490 makes first a digital filter to -15 dB, then an external 3rd order Bessel.
What kind of filter use Deltawave?
The demodulated DSD could make some benefits in correlation (to a PCM wave) using the 50 kHz filter?

Glad to hear the high-contrast display helped!

DeltaWave is using a 7 stage CIC demodulator. This was just a quick-and-dirty way for me to get DSD files converted to PCM, it was never meant to be a reference DSD implementation A 50KHz filter will help cut off all the noise-shaped ultrasonics. You can just select the 48KHz low pass filter to do this from the DW filter selector:

Here's what it looks like with the filter:

 

[Blumlein 88]

Delta Phase display is the counterpart of delta of spectra plot. The result of an FFT is a set of frequency bins, each representing an amplitude and a phase. Delta of Spectra plot represents the difference in magnitude, while Delta of Phase plot represents the difference in phase for each of these bins.

Delta Phase is very sensitive to small differences and can easily result in wild swings if the data is noisy or doesn't match well for other reasons. The setting of phase limit:

View attachment 22740

is used to limit the wild swings for bins where magnitude is below this limit. In other words, any phase differences for bins where magnitude is below -120dB will not appear on the delta phase plot. Try changing this number to -100dB if the RMS delta is at about that value to see if phase differences become more well behaved.

Does this make sense?

Regards,

-Paul

Okay, so this works about like I thought it did. I'd didn't know about the setting and what it was intended to do. You might want to label it db as I wondered what it meant to be set at 120 degrees. That is probably what is happening getting phase differences that wander wildly when there isn't much of anything in some bins.

 

[Blumlein 88]

Finally got around to messing with it some more. My phase setting was on 160 so no wonder it was chaotic. Tried a few settings and ended up leaving it on 96 which makes it work like you'd expect.

 

[Blumlein 88]

Is it just me, seems the volume boost control for applying gain to the difference track playback is not working in 1.07 as it no longer increases or effects playback level. Was working before.
@pkane

 

[Pluto]

seems the volume boost control for applying gain to the difference track playback is not working in 1.07

It works if you set it before commencing playback. I agree, it used to work on-the-fly but doesn't in 1.07.

I'm sure this small regression will be fixed.

 

[pkane]

It works if you set it before commencing playback. I agree, it used to work on-the-fly but doesn't in 1.07.

I'm sure this small regression will be fixed.

Just saw that. Didn’t think I touched that logic, but obviously something I changed as part of the comparator implementation broke the volume control. I’ll fix it ASAP.

 

[pkane]

Version 1.0.8b posted. Please try it.

The updates are as follows:

• Fixed volume control not working during playback
• Fixed driver selection for WASAPI drivers
• Added Log frequency axis display option

 

[Pluto]

The updates are as follows:

• Fixed volume control not working during playback – Excellent. I suggest you make sure that this always defaults to 0dB, especially when switching between ‘real’ audio and Δ audio. Accidental playback of the former with 40dB excess gain might be unsettling
  
• Fixed driver selection for WASAPI drivers – Not sure what you have done here, as it seemed to work before but, as you are working in this area, could you please pick up the ‘friendly names’ for the audio devices. On my system, your audio selection drop-down shows

whereas Foobar, for example, shows the full friendly names the various devices have been given.



I'm sure you appreciate the problem here. ALSO, I suspect that the chosen audio output device is not being correctly remembered. My speakers are at entry number 7 on your drop-down. When the program is closed and re-run, entry number 4 comes up pre-selected (same name, wrong device).

• Added Log frequency axis display option - useful. The frequency labels could use a small tidy-up, as you can see here between 673 and 841.

Some rounding would help, I suspect. Also, when switching between linear and log axes, I notice you re-calculate the graphs before you adjust the state of the check-box. Mightn't doing the check-box first be better as this acknowledges receipt of the instruction and prevents ‘keep-pressing-it’ syndrome!

Keep it come -- coming along nicely!

 

[pkane]

• Fixed volume control not working during playback – Excellent. I suggest you make sure that this always defaults to 0dB, especially when switching between ‘real’ audio and Δ audio. Accidental playback of the former with 40dB excess gain might be unsettling

One thing that may not be obvious is that the volume is set for all three selections independently.

The volume set for Reference or Comparison or Delta tracks applies only to that track. You can see what it’s set to for each by clicking in the box with the name of the file for each of the tracks — the volume control will change to reflect the setting for just that track. It’s OK to raise the volume for Delta track, it will not affect the other two.

When a track is playing, the volume control only changes the volume for that track - the others remain unchanged. Volume selections are remembered across sessions, so that could potentially result in a surprise the next time, if not paying attention!

• could you please pick up the ‘friendly names’ for the audio devices. On my system, your audio selection drop-down shows

whereas Foobar, for example, shows the full friendly names the various devices have been given.

Yes, looks like I was using 'Device friendly name' as opposed to just 'Friendly name'. Didn't realize there was a difference

The frequency labels could use a small tidy-up

Rounding and display of decimal places is made more difficult by the log axis. I'll see what I can do!

 

[pkane]

Latest update posted: version 1.0.9

Changes in version 1.0.9b

• Fixed setting correct audio driver after restart
• Changed driver list to display Friendly Name
• Changed DSD conversion process from CIC demodulator to a configurable low-pass filter, added DSD conversion settings:
  
  
  
• Improved Log axis label formatting
• Added support for drag-and-drop of one or two audio files into DW main window from Windows Explorer
• Added a swap option to File menu

Here's an example of DSD conversion in action (blue is DSD128, white - PCM192):

 

[amirm]

What do you think, better?

Much better to my eye.

 

[pkane]

Latest update posted: version 1.0.9

Changes in version 1.0.9b

• Fixed setting correct audio driver after restart
• Changed driver list to display Friendly Name
• Changed DSD conversion process from CIC demodulator to a configurable low-pass filter, added DSD conversion settings:
  
  View attachment 23007
• Improved Log axis label formatting
• Added support for drag-and-drop of one or two audio files into DW main window from Windows Explorer
• Added a swap option to File menu

Here's an example of DSD conversion in action (blue is DSD128, white - PCM192):
View attachment 23008

More of the results of the same DSD128 vs PCM192 matching analysis as above:

Phase difference:

Difference of spectra:

Spectrum of the difference: