Beta-test: DeltaWave Null Comparison software

[slartibart]

Sounds like you may have some clock drift. When using separate DAC and ADC, the clocks in the two are not running at exactly the same rate. DeltaWave corrects for this, but the result can be a timing offset between the capture and the original file. Use RMS difference (null) instead of correlation null -- correlation null is an overly sensitive metric designed to indicate timing errors.

 

[slartibart]

After reading more of the thread I turned off dither in Audacity and now I get a PK Metric of -80.1dBFS (RMS) when comparing the two players and a Difference of -66.31dB and Correlated Null of 69.26dB.
I still get the weird 16 sample offset between the DAC output and the original FLAC file and PK Metric is -53.9dBFS (RMS) with a Difference of -54.48dB and Correlated Null of 54.5dB.
This was using a cheap Behringer UFO 202 at 44.1kHz. I set both players volume to 90% to avoid any input issues. How do those results sound to you?

 

[slartibart]

Is there a minimum amplitude for the Reference and Compare files? When I compare two files with peak -8dB and RMS -27dB I get a difference of -24dB but if I compare either of those files with a file with peak -4.3dB and RMS -23.6dB I get a difference of -61.3dB.

 

[pkane]

Is there a minimum amplitude for the Reference and Compare files? When I compare two files with peak -8dB and RMS -27dB I get a difference of -24dB but if I compare either of those files with a file with peak -4.3dB and RMS -23.6dB I get a difference of -61.3dB.

It sounds like you’re doing something wrong or maybe the settings are not set correctly. DeltaWave has hundreds of settings that control how and what it does. To test audio players with DAC and ADC you should try to compare the two captures to each other rather than to the original FLAC. Drift correction, sub sample correction and dc and gain corrections should be all turned on, as well as the auto-trim option. All nonlinear EQ settings should be turned off.

 

[slartibart]

It sounds like you’re doing something wrong or maybe the settings are not set correctly. DeltaWave has hundreds of settings that control how and what it does. To test audio players with DAC and ADC you should try to compare the two captures to each other rather than to the original FLAC. Drift correction, sub sample correction and dc and gain corrections should be all turned on, as well as the auto-trim option. All nonlinear EQ settings should be turned off.

I probably am doing something wrong. Turning on auto trim has improved the difference to -74.85 dB compared to -24 dB without it.

 

[pkane]

I probably am doing something wrong. Turning on auto trim has improved the difference to -74.85 dB compared to -24 dB without it.

There must be differences at the start of the two recordings. Auto-trim removes initial samples that seem out of “character” with the rest of the recording. If not removed, these differences contribute to the large error.

 

[slartibart]

There must be differences at the start of the two recordings. Auto-trim removes initial samples that seem out of “character” with the rest of the recording. If not removed, these differences contribute to the large error.

Would you like to try the files yourself? I could put them in Dropbox.

 

[pkane]

Would you like to try the files yourself? I could put them in Dropbox.

Sure, post them. Let me know what’s what.

 

[slartibart]

Sure, post them. Let me know what’s what.

Here they are, two captures of the mpd player at 80% volume
https://www.dropbox.com/sh/1v6mc9ctj4klk5g/AABLlq3UjQWnpq8D0rl8N1zga?dl=0

 

[pkane]

Here they are, two captures of the mpd player at 80% volume
https://www.dropbox.com/sh/1v6mc9ctj4klk5g/AABLlq3UjQWnpq8D0rl8N1zga?dl=0

So there are two captures with nothing changed in the chain, including same player?

 

[pkane]

Here they are, two captures of the mpd player at 80% volume
https://www.dropbox.com/sh/1v6mc9ctj4klk5g/AABLlq3UjQWnpq8D0rl8N1zga?dl=0

The result you got, -74.85dBFS RMS null is what I'm getting. Looking through different plots, there are a couple that show why.

The largest component of the null file appears to be a 50Hz frequency. I assume mains? Reduce that and you'll get a deeper null, closer to the theoretical minimum:

There also appears to be a small difference below 20Hz between the two captures that also accounts for some difference:

By the way, this is why you needed to fix the offset and use auto-trim . This is the two files without trimming, at the start:

 

[slartibart]

The result you got, -74.85dBFS RMS null is what I'm getting. Looking through different plots, there are a couple that show why.

The largest component of the null file appears to be a 50Hz frequency. I assume mains? Reduce that and you'll get a deeper null, closer to the theoretical minimum:

View attachment 231833


There also appears to be a small difference below 20Hz between the two captures that also accounts for some difference:
View attachment 231834


By the way, this is why you needed to use the auto-trim option. This is the two files without trimming, at the start:

View attachment 231836

Interesting, how are you zooming the waveforms? Maybe if I power the DAC or the Raspberry Pi with a power pack I can reduce the 50Hz issue.

 

[pkane]

Interesting, how are you zooming the waveforms? Maybe if I power the DAC or the Raspberry Pi with a power pack I can reduce the 50Hz issue.

Use the mouse scroll wheel, while pointing at the point where you want to zoom in. You can also point the mouse at one axis (near the numbers) to zoom in using scroll wheel just on that axis. With Ctrl+right mouse button drag, you can also select the area to zoom in on using a rectangle.

 

[slartibart]

Use the mouse scroll wheel, while pointing at the point where you want to zoom in. You can also point the mouse at one axis (near the numbers) to zoom in using scroll wheel just on that axis. With Ctrl+right mouse button drag, you can also select the area to zoom in on using a rectangle.

Right, I don't have a mouse but the trackpad works in the same way. Any idea why one waveform.appears to start before the other? I aligned them by choosing a matching sample in the original FLAC and the captures then trimming the capture so that sample appeared in the same place. I see the same in Audacity but there are around 5500 samples between the waveforms.

 

[pkane]

Right, I don't have a mouse but the trackpad works in the same way. Any idea why one waveform.appears to start before the other? I aligned them by choosing a matching sample in the original FLAC and the captures then trimming the capture so that sample appeared in the same place. I see the same in Audacity but there are around 5500 samples between the waveforms.

No idea. The original waveform plot in DW shows the samples exactly as read from the two files, and they are not perfectly aligned.

 

[slartibart]

Right, I don't have a mouse but the trackpad works in the same way. Any idea why one waveform.appears to start before the other? I aligned them by choosing a matching sample in the original FLAC and the captures then trimming the capture so that sample appeared in the same place. I see the same in Audacity but there are around 5500 samples between the waveforms.

The original FLAC file starts earlier than both captures, I think it is caused by the DAC coming out of standby so I should be able to avoid that issue by turning off standby

 

[Grooved]

The original FLAC file starts earlier than both captures, I think it is caused by the DAC coming out of standby so I should be able to avoid that issue by turning off standby

I have doubt about it because I also got the 5500 samples added when I tested with several devices and using Deltawave with a auto play-record feature.
I thought @pkane set it like that to be sure to have the recording starting a bit before the the file played, and finishing a bit after the end, which is safer and should not create a problem as long as Deltawave can cut the excess before doing measurement

 

[pkane]

I have doubt about it because I also got the 5500 samples added when I tested with several devices and using Deltawave with a auto play-record feature.
I thought @pkane set it like that to be sure to have the recording starting a bit before the the file played, and finishing a bit after the end, which is safer and should not create a problem as long as Deltawave can cut the excess before doing measurement

I did, but these are two recordings made the same way, so should both have the same lead-in.

 

[Grooved]

I did, but these are two recordings made the same way, so should both have the same lead-in.

OK, I didn't understand that. Strange, it looks more like one 1st record was done, and that this 1st file was then used to play/recording the 2nd file (instead of the original)
It's the only explanation I see to get two recorded file with 5 500 samples difference

Original = x samples
1st recording = x + 5 500 samples
2nd recording (from 1st recording) = x + 5 500 + 5 500 samples

 

[slartibart]

OK, I didn't understand that. Strange, it looks more like one 1st record was done, and that this 1st file was then used to play/recording the 2nd file (instead of the original)
It's the only explanation I see to get two recorded file with 5 500 samples difference

Original = x samples
1st recording = x + 5 500 samples
2nd recording (from 1st recording) = x + 5 500 + 5 500 samples

It was definitely the DAC coming out of standby causing the very start of the waveforms to be missing. I have since made more captures ensuring the DAC was ready before starting the track and now the captures and the FLAC file start simultaneously.