Beta-test: DeltaWave Null Comparison software

[pkane]

Hi Paul,
I have created an account here just to say that I love your software. DeltaWave is amazing. Finally, I'm able to ditch DiffMaker.

Many thanks for this program and for making it better and better with each new release!

Is there any way to support the development?

Thank you, and appreciate the kind words! For now, testing and finding new ways to use DeltaWave is the main support I need

 

[pkane]

I'm not sure whether you've seen this one, from the "sound of DSD" thread:

By the looks of it the offending samples differ by time delay (non-integer number of samples) only. I tried matching them up in Audacity manually and had to resample to a 1 MHz sample rate to even get a -60 dB null. Still, I would imagine that dealing with arbitrary time delays is kind of basic and important.

I did see it, and responded in that thread. The main problem with that test was that the two files were too short and contained fully periodic material (a single sine-wave). DeltaWave tries to compute clock drift, variable group delay, jitter, etc., all of which requires quite a few more samples than a few seconds. I never measure with less than 30 seconds, but with lower sampling rates, such as 44.1k, I'd say use at least a minute.

Generally, music or noise test signals are preferred to sine-waves, since two sine waves can align perfectly at thousands if not millions of places, and DW has no way to pick one over the other. When using periodic signal, there is a setting that should help and needs to be turned on: Measure Simple periodic waveforms. This attempts to deal with exactly this issue.

Speaking of time delays, DeltaWave not only measures and corrects delay to much less than 1/1000 of a sample accuracy, but also measures and corrects non-linear phase effects and clock speed differences.

 

[Blumlein 88]

@pkane
Probably a foolish question, but could Deltawave figure out what the noise floor of a file is? Obviously if I had a null of nothing, but noise then the noise level is what is there. My guess is this might only work with test signals rather than music.

I see REW lists that in some of its parameters and wonder how it does that and if it is accurate.

 

[pkane]

@pkane
Probably a foolish question, but could Deltawave figure out what the noise floor of a file is? Obviously if I had a null of nothing, but noise then the noise level is what is there. My guess is this might only work with test signals rather than music.

I see REW lists that in some of its parameters and wonder how it does that and if it is accurate.

I don't think I can easily extricate noise floor from signal when the original test signal is unknown, such as with music. But let me think about it. Maybe there's a way to do this with an approximation.

 

[Coach_Kaarlo]

Not sure if you have a thread for your Earful app, but will post here.

One comment or feedback point - left and right channel muting would really improve the test.

Rare that hearing does not differ slightly in one ear from the other.

Also helps make eq. for headphones more accurate.

Edit: [awesome app, very very functional - have done most of my family with interesting results - appears my boy is not deaf but has excellent source selectivity ; ) ]

Cheers.

 

[MC_RME]

When using periodic signal, there is a setting that should help and needs to be turned on: Measure Simple periodic waveforms. This attempts to deal with exactly this issue.

Isn't it easy for DeltaWave to differ between 'music' and periodic signals? Every added option makes the program more complicated and less easy to use. IMHO it would be nice to have this differentiation done automatically.

 

[pkane]

Isn't it easy for DeltaWave to differ between 'music' and periodic signals? Every added option makes the program more complicated and less easy to use. IMHO it would be nice to have this differentiation done automatically.

DeltaWave can detect when there are too many possible alignment points easily, the cross-correlation step will tell it that. And usually it's an indication of a periodic signal. But sometimes this could be because the two files are really not matching well or are completely different.

I'd prefer that DW doesn't automatically switch to the simpler algorithm just because the files seem to match poorly. Sometimes it's because they really don't match and not because the signal is periodic I may just leave this option as a setting, but when detecting a potentially periodic signal, ask the user if this should be processed as such.

 

[I-Love-Null-Tests]

How does DeltaWave treat stereo files? Does it take all channels into account?

Sorry if this was already answered somewhere, I wasn't able to find the answer.

 

[pkane]

How does DeltaWave treat stereo files? Does it take all channels into account?

Sorry if this was already answered somewhere, I wasn't able to find the answer.

You can choose to measure/correct only one channel, left or right, sum the two channels to mono and use that as the new comparison channel, or keep left and right separate (stereo) and correct both. In this case, the same correction will be applied to both channels. You can also compare left to right or right to left, or mono to L or R channels, if you'd like.

 

[I-Love-Null-Tests]

You can choose to measure/correct only one channel, left or right, sum the two channels to mono and use that as the new comparison channel, or keep left and right separate (stereo) and correct both. In this case, the same correction will be applied to both channels. You can also compare left to right or right to left, or mono to L or R channels, if you'd like.

View attachment 90051

Ah, have missed that drop-down menu. That's awesome, thank you very much!

 

[I-Love-Null-Tests]

When I choose the Stereo option, processing is much faster, than when I choose just left channel. Shouldn't it be the other way?

 

[pkane]

When I choose the Stereo option, processing is much faster, than when I choose just left channel. Shouldn't it be the other way?

Don't see any reason for it to be much faster, unless it fails due to lack of memory. Look at the Results tab and see if the processing completed.

 

[Blumlein 88]

I don't think I can easily extricate noise floor from signal when the original test signal is unknown, such as with music. But let me think about it. Maybe there's a way to do this with an approximation.

Here is a screenshot that got me to thinking about noise. It is the delta of the waveform.

This was from a large multi-channel recording mixing console said to have a noise floor of -80 db once you notch out some idle noise from the slider automation system at 26.7 khz. You see the near solid band of level running around -80 db. So that looks like a graphical indicator of a noise floor at least with a relatively noisy device. I suppose it will do for a rough guide to noise levels.

BTW, I was rather appalled to find this level of noise is typical of recording mixing consoles. The one in this instance is some $70k for 16 channels of automated console.

 

[pkane]

Here is a screenshot that got me to thinking about noise. It is the delta of the waveform.

This was from a large multi-channel recording mixing console said to have a noise floor of -80 db once you notch out some idle noise from the slider automation system at 26.7 khz. You see the near solid band of level running around -80 db. So that looks like a graphical indicator of a noise floor at least with a relatively noisy device. I suppose it will do for a rough guide to noise levels.

BTW, I was rather appalled to find this level of noise is typical of recording mixing consoles. The one in this instance is some $70k for 16 channels of automated console.

View attachment 90273

If you turn on non-linear EQ and phase correction, you get this:

There's literally no hint of music left in the delta file when playing, all but noise. Zooming in, some of it appears to be 60Hz along with some higher frequency noise:

 

[Blumlein 88]

@pkane
In 1.053b filter 1 doesn't work at all now. Not for HP or LP, nor at the start or the end. Works in filter 2.

 

[pkane]

@pkane
In 1.053b filter 1 doesn't work at all now. Not for HP or LP, nor at the start or the end. Works in filter 2.

What exact combination did you try? Everything I've tried with Filter 1 seemed to work. Here's an example:

 

[Blumlein 88]

What exact combination did you try? Everything I've tried with Filter 1 seemed to work. Here's an example:

View attachment 90396

Same thing you have there with an HP or LP at 14,000 hz.

The 2nd filter had a value in it though it was set to off.

 

[pkane]

Same thing you have there with an HP or LP at 14,000 hz.

The 2nd filter had a value in it though it was set to off.

Like this? Seems to be working:

Check that you have valid FIR filter settings, although that would affect both filters if there's something wrong here:

 

[Blumlein 88]

That is it. If (no limit) is in filter two, even if filter two is off it works. If any other value is in filter two, though set to off, then filter one doesn't work.

 

[Blumlein 88]

So is IIR not an appropriate filter?