• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Beta-test: DeltaWave Null Comparison software

PierreV

Major Contributor
Forum Donor
Joined
Nov 6, 2018
Messages
1,437
Likes
4,686
Well, my line of thought was more that, if the program ever ends up in the hands of non technically minded users or even most reviewers ;) "horrible" (but non significant) graphs would generate lots of questions/debates. I can already imagine people comparing cables and focusing on differences at -200dB ;) or on inexistent phase differences (more annoying at a spurious 5 to 10°).

And no need to add anything for me, I certainly do not wish to give you more work. :)
 
OP
pkane

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,626
Likes
10,202
Location
North-East
Well, my line of thought was more that, if the program ever ends up in the hands of non technically minded users or even most reviewers ;) "horrible" (but non significant) graphs would generate lots of questions/debates. I can already imagine people comparing cables and focusing on differences at -200dB ;) or on inexistent phase differences (more annoying at a spurious 5 to 10°).

And no need to add anything for me, I certainly do not wish to give you more work. :)

That's a valid point. I'll think about putting some reasonable limits in, and not let it zoom-in all the way down to one trillionth of a dB. At least not as the initial, default view of the data.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,049
That's a valid point. I'll think about putting some reasonable limits in, and not let it zoom-in all the way down to one trillionth of a dB. At least not as the initial, default view of the data.
I think a default limit is good. But do please leave an optional view down to the ridiculous limits. I made the same mistake. Looking at the graph wondering how it made sense then noticing the values were in some small negative exponents. Like 1.5 billionths of a degree phase error.
 

Pluto

Addicted to Fun and Learning
Forum Donor
Joined
Sep 2, 2018
Messages
990
Likes
1,631
Location
Harrow, UK
1.0.11 "Index was outside the bounds of the array"

Just a quick one - I haven't really much time until after the weekend.

Attempting to load two fairly hefty files works when the two channel selectors are set to "L" but causes the above error whenever the reference is set to "R" (regardless of whether the compare channel is L or R)
 
OP
pkane

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,626
Likes
10,202
Location
North-East
1.0.11 "Index was outside the bounds of the array"

Just a quick one - I haven't really much time until after the weekend.

Attempting to load two fairly hefty files works when the two channel selectors are set to "L" but causes the above error whenever the reference is set to "R" (regardless of whether the compare channel is L or R)

Is this with DSF/DFF files by any chance? Can't reproduce this with WAV or FLAC.
 
OP
pkane

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,626
Likes
10,202
Location
North-East
Fixes for some of the reported issues in version 1.0.12:
  • Fix for audio driver shown in the list but audio not playing (@PierreV)
  • Fix for DSD file right channel selector causing an index error (@Pluto)
  • Set limits on plots to avoid zooming in to the rounding error in floating point computations (@PierreV, @Blumlein 88)
Still looking for feedback on the Windows 10 sizing issue. Did the new font selector help at all?
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,049
Fixes for some of the reported issues in version 1.0.12:
  • Fix for audio driver shown in the list but audio not playing (@PierreV)
  • Fix for DSD file right channel selector causing an index error (@Pluto)
  • Set limits on plots to avoid zooming in to the rounding error in floating point computations (@PierreV, @Blumlein 88)
Still looking for feedback on the Windows 10 sizing issue. Did the new font selector help at all?

With version .10 my sizing issues were fixed when I went into Program properties of windows and set DPI scaling to System (Enhanced). Looks sharper than before too.

When I installed the new versions it retained that scaling info and still looks fine.
 
Last edited:
OP
pkane

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,626
Likes
10,202
Location
North-East
With version .10 my sizing issues were fixed when I went into Program properties of windows and set DPI scaling to System (Enhanced). Looks sharper than before too.

When I installed the new versions it retained that scaling info and still looks fine.

I think I fixed it so you don't have to change program properties, but if it's working well for you, there's no reason to experiment any further.
 

PierreV

Major Contributor
Forum Donor
Joined
Nov 6, 2018
Messages
1,437
Likes
4,686
Audio device now available, but going to another error message... audacity loads and plays seamlessly.
Note: if you prefer nor need more tests, we can go private on this thread.

1552083724988.png
 
OP
pkane

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,626
Likes
10,202
Location
North-East
Audio device now available, but going to another error message... audacity loads and plays seamlessly.
Note: if you prefer nor need more tests, we can go private on this thread.

View attachment 23271

Thanks, this is actually a more reasonable error: it means the device driver doesn’t support the format that DW tries to play. Either sampling rate or bit number is not supported.

DW plays from an internal floating point format, which will be 32 bits by default. Let me think about how to support drivers with lower bitness.
 

PierreV

Major Contributor
Forum Donor
Joined
Nov 6, 2018
Messages
1,437
Likes
4,686
That's what I thought at first, but Audacity also handles it as 32-bit float (that's the reason why I included it in the screenshot). Quickly tried to check but since it goes through the video card, it is not as obvious as on dedicated audio drivers. I suspect it gets transparently resampled to 48KHz by either Audacity, Windows or the Nvidia driver. I wouldn't worry too much about it as that particular driver is a mess anyway and probably wouldn't be used by anyone doing comparisons or listening tests.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,049
Well if you want a new feature I have a suggestion. @pkane

You already compute a delta of spectrum which shows frequency response. Would it be a big hassle to use that for FR compensation? Diffmaker had that though I couldn't get it to work reliably enough to really try it. It might be a nice option. Since things go a bit off between 20khz and 22.05 khz you do not want to try EQ on that or the user needs to set the cutoff point (which you already could with the LP filter). So is this something that could be added without tons of work?

I've been using the current version, and everything seems to be working just great. I've been doing things like EQ and noise addition to clean files to learn what different things sound like in the difference file. Low end and high end boosts and dips are easy enough to hear for what they are. Ripples in response like from my ADC leave a fairly clean residual of the music at a low level. Various noise leaves noise in the residuals. So Deltawave is very useful like it is. :)
 
OP
pkane

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,626
Likes
10,202
Location
North-East
Well if you want a new feature I have a suggestion. @pkane

You already compute a delta of spectrum which shows frequency response. Would it be a big hassle to use that for FR compensation? Diffmaker had that though I couldn't get it to work reliably enough to really try it. It might be a nice option. Since things go a bit off between 20khz and 22.05 khz you do not want to try EQ on that or the user needs to set the cutoff point (which you already could with the LP filter). So is this something that could be added without tons of work?

I've been using the current version, and everything seems to be working just great. I've been doing things like EQ and noise addition to clean files to learn what different things sound like in the difference file. Low end and high end boosts and dips are easy enough to hear for what they are. Ripples in response like from my ADC leave a fairly clean residual of the music at a low level. Various noise leaves noise in the residuals. So Deltawave is very useful like it is. :)

Good suggestion. The non-linear level matching does pretty much what you describe, but using a fitted curve instead of the data to match FR of the comparison file to the reference. That generally works better with longer files, as this allows noise to be reduced through averaging.

I've actually had frequency compensation using a calibration file in DW for quite a while. I didn't find that it helped improve the null so it's been hidden all this time :)

1552159353537.png


What I tried before was to use a white-noise capture to compute the frequency compensation, an impulse response WAV file, such as produced by REW or rePhase, an actual frequency response as measured by REW from a sweep, and even REW loopback sound card measurements. So far, I've found that such compensation can make the spectrum curve look better, but the null is usually made worse. I suspect it has to do with noise in the measurements. Maybe a longer white-noise capture, say 5-10 minutes, would produce a better result as it will average a lot more samples.
 
OP
pkane

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,626
Likes
10,202
Location
North-East
[I tried to post this already, but apparently the message never showed up...]

Update version 1.0.14 of DeltaWave is available.

Changes in this version:
  • Added manual and automatic resampling for audio playback on audio cards where native rates are not supported (thank you @PierreV for reporting and helping to troubleshoot the problem!)

  • Changed non-linear drift correction to be turned off by default
 
OP
pkane

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,626
Likes
10,202
Location
North-East
Version 1.0.15b is out!

Changes in this version:
  • Fixed a problem that caused a larger phase difference in the first few seconds of a track compared to the rest of the track
  • Removed non-linear polynomial gain correction option
  • Removed non-linear drift correction option
  • Added non-linear EQ/frequency matching correction (use with caution, only if there are gross differences in spectra of the two tracks!)
1552450438817.png


Here's what EQ does on a track. Before EQ:
1552450711023.png


After EQ:
1552450645357.png
 
Last edited:

Pluto

Addicted to Fun and Learning
Forum Donor
Joined
Sep 2, 2018
Messages
990
Likes
1,631
Location
Harrow, UK
1.0.15

All good so far. What exactly does the EQ symbol (bottom right) designate?

A few suggestions –

Is it time to implement drag and drop for entry into the 'reference' and 'compare' fields?

Could you please make the 'computing' spinner far more prominent. The same comment applies to the 'drawing' spinner but this may be beyond your control. Likewise, the 'drawing' spinner penetrates other windows on the desktop.

The 'results' text tab could use a scroll bar when its contents so demand.

A big red flag indicating that the user has invalidated the current computation, by changing the operating parameters, would be handy.

When pressing 'match', some of the functions such as loading the files appear to be repetition of work done on the same files if 'show' had been pressed earlier. For instance, unless the input files had been changed, presumably the 'load' phase need not be repeated. Would it be worth making a general effort to not repeat unnecessary work?

Keep up the great work!!!
 
OP
pkane

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,626
Likes
10,202
Location
North-East
1.0.15

All good so far. What exactly does the EQ symbol (bottom right) designate?

A few suggestions –

Is it time to implement drag and drop for entry into the 'reference' and 'compare' fields?

Could you please make the 'computing' spinner far more prominent. The same comment applies to the 'drawing' spinner but this may be beyond your control. Likewise, the 'drawing' spinner penetrates other windows on the desktop.

The 'results' text tab could use a scroll bar when its contents so demand.

A big red flag indicating that the user has invalidated the current computation, by changing the operating parameters, would be handy.

When pressing 'match', some of the functions such as loading the files appear to be repetition of work done on the same files if 'show' had been pressed earlier. For instance, unless the input files had been changed, presumably the 'load' phase need not be repeated. Would it be worth making a general effort to not repeat unnecessary work?

Keep up the great work!!!



EQ indicates the non-linear equalizer is engaged. Look in Settings under non-linear correction.
1552569851216.png

This is a somewhat dangerous operation as it tries to equalize the spectrum of the comparison file to match the spectrum of the reference, so generally should be turned off unless you specifically want the same frequency response.

Is it time to implement drag and drop for entry into the 'reference' and 'compare' fields?

That's there. Go to Windows Explorer, pick one or two audio files, and drop them onto DeltaWave.

The 'results' text tab could use a scroll bar when its contents so demand.
1552570122526.png

When pressing 'match', some of the functions such as loading the files appear to be repetition of work done on the same files if 'show' had been pressed earlier. For instance, unless the input files had been changed, presumably the 'load' phase need not be repeated. Would it be worth making a general effort to not repeat unnecessary work?

There are definitely optimizations possible there. Thanks a lot for testing and your feedback. Keep it coming!
 
Last edited:
Top Bottom