Beta-test: DeltaWave Null Comparison software

[pkane]

Yes, a longer song recorded with
View attachment 553881
selected and automated matching active.

Thanks, there were some new features implemented in the recorder in .27. I'll take a look.

 

[pkane]

Yes, a longer song recorded with
View attachment 553881
selected and automated matching active.

Can you please try the recorder in v2.0.28?

https://app.box.com/s/udy8ujifatplvpsezegfe5vk8zyk9j1j

 

[Rantapossu]

Can you please try the recorder in v2.0.28?

https://app.box.com/s/udy8ujifatplvpsezegfe5vk8zyk9j1j

Thanks!

Now it shows:

If I answer "No", the result is:

With "Yes", it stops the matching right away and gives the same result immediately:

 

[Rantapossu]

V2.0.28 makes the recording about 0.6 seconds shorter than the working V2.0.24 (30.0 seconds vs. 30.6 seconds):

 

[Rantapossu]

When you add the 0.1 second sound card latency (Between the reference and compare) from the start...

... you only have 29.9 seconds (Instead of 30.5 seconds with v2.0.24) of recorded audio to match 30 seconds of the reference data.

 

[Rantapossu]

I did further research...

It seems that both v2.0.27 and v2.0.28 are fundamentally broken.

Of course I don't have all the unreleased versions, but v2.0.24 is the last version for me that can correctly match two completely identical bit perfect songs.

v2.0.24:

v2.0.27:

v2.0.28:

And -120.84 dB and -122.56 dBA correlates my previous bug report about the automatic matching doing it against the digital copy of the reference and not against the analog recording.

 

[pkane]

I did further research...

It seems that both v2.0.27 and v2.0.28 are fundamentally broken.

Of course I don't have all the unreleased versions, but v2.0.24 is the last version for me that can correctly match two completely identical bit perfect songs.

v2.0.24:
View attachment 553961

v2.0.27:
View attachment 553962

v2.0.28:
View attachment 553963

And -120.84 dB and -122.56 dBA correlates my previous bug report about the automatic matching doing it against the digital copy of the reference and not against the analog recording.

Please try v2.0.29 -- this appears to work as expected in all my tests: https://app.box.com/s/ls9p0cwqjbqwlofvp30nmg7a5q66fahh

 

[Rantapossu]

Please try v2.0.29 -- this appears to work as expected in all my tests: https://app.box.com/s/ls9p0cwqjbqwlofvp30nmg7a5q66fahh

Thanks again!

The good news: Bit perfect matching of two identical files works again and all the 16-bit, 24-bit, 32-bit and 64-bit savings of the Reference, Compare and Delta are still ok without 6 dB boost.

The bad news: Nothing has changed with the matching of analog recording.

I still get this (Doesn't matter if I use automatic matching during the recording or load previously recorded file and click the Match -button):

And it still records too short clips (30 second clip is still 0.6 second shorter than with v2.0.24):

 

[pkane]

Thanks again!

The good news: Bit perfect matching of two identical files works again and all the 16-bit, 24-bit, 32-bit and 64-bit savings of the Reference, Compare and Delta are still ok without 6 dB boost.

The bad news: Nothing has changed with the matching of analog recording.

I still get this (Doesn't matter if I use automatic matching during the recording or load previously recorded file and click the Match -button):

View attachment 554161

And it still records too short clips (30 second clip is still 0.6 second shorter than with v2.0.24):

View attachment 554162

Thank you for testing!

The error seems strange, I don't get the same when I test. The result of the recording seems correct for a 30 second clip in v2.0.29 -- it looks like it ends right at 30 seconds? Maybe if you could share the original and the recorded file that's too short and gives you the error, I can try to figure out what's going on.

 

[Rantapossu]

And v2.0.24 can match the recording of v2.0.29 correctly but not the other way around.

This is a 30 second clip recorded with v2.0.29 and matched with v2.0.24 and it works:

Of course, the delta can't be a full 30 seconds anymore because the recording is too short. The delta is only about 29.9 seconds long:

 

[Rantapossu]

This seems to be due to differences between ASIO and WASAPI. I just got matching working with WASAPI. I'll do a couple more tests...

<EDIT> No it wasn't, it was a measurement error </EDIT>

 

[Rantapossu]

The result of the recording seems correct for a 30 second clip in v2.0.29 -- it looks like it ends right at 30 seconds?

It has to be longer than 30 second, because of the latency of the sound card. Previously it was 30.6 seconds and it was enough, now it's exactly 30 second and this will truncate the delta, because there is no enough material to match the whole 30 seconds.

Unfortunately, my recent observation about the ASIO/WASAPI differences was a measurement error, matching doesn't actually work with WASAPI either .

Both ASIO and WASAPI recordings are exactly 30 second long:

 

[Rantapossu]

Actually WASAPI is even worse, because of the higher latency. WASAPI doesn't have anywhere near 30 seconds of actual data (Actually only about 28.8 seconds) to match because of that. This is at the start of the recordings, compared with ASIO. There's almost 1 second additional delay:

 

[Rantapossu]

Thanks Paul for everything!

I just sent a DM with links to the recordings.