Beta-test: DeltaWave Null Comparison software

[pkane]

Repeat, new recordings, with No Ceiling Fan, a little higher signal to ambient noise level...

Files loaded without the previous "residual drift" error...

View attachment 23998

View attachment 24001
The waveform difference is about 3 to 4 times the ambient/mic noise level:

View attachment 23999

System/ambient/mic noise at beginning of recording:

View attachment 24000

Spectrum of the noise only:

View attachment 24002

Looks about the same as the first graphic above.

By the way, I did find the original 'array index out of bounds' error with your data when processing L+R (stereo to mono mixdown). This is interesting, but somehow there was one less sample in the right channel than in the left. I've made adjustments to DW to handle this more gracefully, so thank you for reporting the problem!

Regards,

-Paul

 

[RayDunzl]

Playing with it some more - using in-room measures, which is where I listen...

A 500Hz tone... perhaps oddly, no 500Hz tone seems to show in any of the spectograms...

Cepstrum.... I plead ignorance and have no idea what I'm looking at...

A swept tone... Residual drift error again...

 

[pkane]

Playing with it some more - using in-room measures, which is where I listen...

A 500Hz tone... perhaps oddly, no 500Hz tone seems to show in any of the spectograms...

Cepstrum.... I plead ignorance and have no idea what I'm looking at...

A swept tone... Residual drift error again...

I don’t think DW will do well with microphone captures, as it relies on a very close phase relationship between the two waveforms. Any phase differences that are not linear will throw it off. Best to turn off drift correction for these comparisons.

I’ll add another way to compute drift that isn’t as sensitive to phase variations in the next version.

Strange that you don’t see the 500Hz line in spectrograms. How far above the level of other frequencies is it?

 

[RayDunzl]

Strange that you don’t see the 500Hz line in spectrograms. How far above the level of other frequencies is it?

About -7dB on the recording, no other frequencies except ambient/system noise, -40dB or less.

Spectrogram

 

[RayDunzl]

I’ll add another way to compute drift that isn’t as sensitive to phase variations in the next version.

Don't break the precision for electrical measurements... I just like to "see what happens" sometimes.

 

[pkane]

Don't break the precision for electrical measurements... I just like to "see what happens" sometimes.

I like that you're testing the limits, Ray

DeltaWave has problems with generated sounds, such as simple sine waves. This is because there are multiple points (sometimes hundreds of thousands) where they will align perfectly. Picking the one "right spot" isn't easy in this case! This is definitely something I would like to address in a future update, but it has not been the focus of the software up until now.

By the way, uncheck the Log Axis checkbox on that spectrogram and see if you can see the 500Hz line then. The log axis option doesn't work right on spectrograms... yet, even though the numbers seem to change.

 

[PierreV]

I don’t think DW will do well with microphone captures

I am quite happy with microphone captures. Have had some fun with five amplifiers, three DACs and two pairs of speakers, and plan to do more tests with the big systems soon. I did pay attention to level matching, no interference, and repeatable measurement conditions though. For big differences, I can correlate subjective impressions with measurements. Smaller differences are harder to catch though and, tbh, I don't see measurements settling the discussions about things such as "more slam", "lifted veils", "airy scene" in any unambiguous way.

But this is fun!

 

[RayDunzl]

By the way, uncheck the Log Axis checkbox on that spectrogram and see if you can see the 500Hz line then.

Yes, it does show up.

GUI Comment: Although you can scroll on the value axis to zoom, you can't grab the display and move it up and down as you can with, say, the spectrum dispalays.

 

[pkane]

Version 1.0.20b is now available.

This incorporates fixes for reported issues, including everything in v1.0.19, as follows:

• Better nulls with Audacity-processed 16-bit files
• Read 24- and 32-bit integer WAV formats correctly
• Display correct offset value in Results tab
• Added silence trimming option (on by default)
• Added scale display and other enhancements in the spectrogram windows, including annotation support
• Right-click now removes the last annotation added to the chart
• Changed the waveform Y axis to display in dB rather than 1 to -1 floating point by default
• Added the option to switch between dB and floating point view of Y axis
• Fixed the 100% that’s really a ‘nearly 100%’ display
• Fixed the installer so previous versions are removed from the list of installed packages
• Handled different number of samples in left and right channels when working on L+R mono mix-down channel
• Added an optional, more stable drift correction method if the normal method does not converge
• Various minor spelling/terminology fixes

As always, all comments, issues, suggestions are more than welcome!

 

[RayDunzl]

Since, as far as I can tell, the Phase window has no limit on the number of degrees of difference, can you provide an "unwrap" option there?

Looking at a wrapped display, can't tell if the difference goes on and on or just exceeds 180 degrees in both directions ending up with small overall change.

Simple Example:

Wrapped (in REW):

Unwrapped:

Reasoning:

What am I to make of this segment of a DelataWave phase display, of the same track played twice?

Does it go way out, or wobble around 0, or what?

Is there some problem in some frequency range?

Zooming in, it could be manually unwrapped and decoded (excuse the squiggle at 3.7k, should keep going down):

But to just glance at it for an overview leaves me underwhelmed:

 

[pkane]

Since, as far as I can tell, the Phase window has no limit on the number of degrees of difference, can you provide an "unwrap" option there?

Looking at a wrapped display, can't tell if the difference goes on and on or just exceeds 180 degrees in both directions ending up with small overall change.

Simple Example:

Wrapped (in REW):

View attachment 24109

Unwrapped:

View attachment 24110

Reasoning:

What am I to make of this segment of a DelataWave phase display, of the same track played twice?

View attachment 24111

Does it go way out, or wobble around 0, or what?

Is there some problem in some frequency range?

Zooming in, it could be manually unwrapped and decoded (excuse the squiggle at 3.7k, should keep going down):

View attachment 24112

But to just glance at it for an overview leaves me underwhelmed:

View attachment 24113

Yes, thanks for reminding me! Phase unwrapping could definitely be very helpful in cases like this.

 

[Pluto]

1.0.21
Looking good so far!

Concur about the phase-unwrapping option.

 

[pkane]

Re: phase unwrap

I added phase unwrap option, but need confirmation that this is what it should do. Here's an example.

Phase wrapped:

Phase unwrapped:

Considering the huge differences reaching to the tens of thousands of degrees, is the unwrapped chart at all useful? Obviously you can zoom in on parts of it, but... Is this what it should look like? @RayDunzl @Pluto

 

[RayDunzl]

Considering the huge differences reaching to the tens of thousands of degrees, is the unwrapped chart at all useful?

Given that measurement, I could ask, is the wrapped version useful?

Do the highs lag (or lead) by 30,000 degrees?

Whats the deal?

 

[pkane]

Given that measurement, I could ask, is the wrapped version useful?

Do the highs lag (or lead) by 30,000 degrees?

Whats the deal?

No idea! That’s why I‘m asking. Unwrap operation is really just a guess, trying to keep the next phase value within 180 degrees from the previous one. For this particular plot it appears to grow like crazy when unwrapped...

 

[RayDunzl]

In REW, the usual unwrapped plot for speakers has slope from top right to bottom left.

My interpretation puts higher (on the graph) as "early", below "late".

Measuring close to a speaker can give a flat curve, reflections drive it nuts, and generally down and right..

But DeltaWave is making a comparison, so, ???

 

[pkane]

In REW, the usual unwrapped plot for speakers has slope from top right to bottom left.

My interpretation puts higher (on the graph) as "early", below "late".

Measuring close to a speaker can give a flat curve, reflections drive it nuts, and generally down and right..

But DeltaWave is making a comparison, so, ???

Right. DW shows the difference in phase between two signals. Considering the phase / drift correction that's part of the process, I can't imagine that the two signals are so much out of phase, unless there are significant differences (for example, a filter that's applied to one but not to the other). Let me think about a better way to do the unwrapping, but it may just be how it is...

 

[Pluto]

Is this what it should look like?

I thought that the wrap option would, in effect, have ‘zoomed out’ the unwrapped Y axis (scaled from 0° to ±180°) to whatever phase angle was necessary to accommodate the angle you need to show. Your phase unwrapped plot shows the absolute value rather than the signed value with respect to the 0° reference.

I think this was what @RayDunzl was expecting

 

[pkane]

I thought that the wrap option would, in effect, have ‘zoomed out’ the unwrapped Y axis (scaled from 0° to ±180°) to whatever phase angle was necessary to accommodate the angle you need to show. Your phase unwrapped plot shows the absolute value rather than the signed value with respect to the 0° reference.

I think this was what @RayDunzl was expecting

No, it's not an absolute value chart. It just happens to work out that for that set of data it was monotonically increasing. The algorithm simply adds or subtracts 360 degrees to each point, so as to reduce any jumps of 180 degrees or more. Here's another unwrapped phase chart:

And this is the 'wrapped' version:

Don't know of any better method of unwrapping phase. If anyone has better suggestions, I'd be very interested.

 

[JohnPM]

It can be handy for users to pick where the unwrapped plot passes through the -180..180 range, so the range is narrower near that region of interest. Also need to remove anything that looks like a pure time delay from the data before unwrapping it (or an underlying linear trend, which amounts to the same thing), otherwise the phase steps at high frequency can be too large for unwrapping to work. Not sure how meaningful that is for this kind of comparative data though.