Beta Test: Multitone Loopback Analyzer software

[pkane]

Like this:

View attachment 259532

Re: intersample-overs

A more interesting test for would be to use the same preset with output at 44.1k, but recorded at 192k or even 384k to see the harmonic distortion generated by these sine waves. I think that's where you might see the effect (if any) of intersample overs. Here's ADI-2 Pro FS:

The 7350Hz test, looks much better:

 

[Rja4000]

Re: intersample-overs

A more interesting test for would be to use the same preset with output at 44.1k, but recorded at 192k or even 384k to see the harmonic distortion generated by these sine waves. I think that's where you might see the effect (if any) of intersample overs. Here's ADI-2 Pro FS:

View attachment 259598

Which is exactly what I did above to measure THD evolution for intersample overloads.

That would be very helpful to automate it, IMO.

The 7350Hz test, looks much better:

View attachment 259609

Try the 11kHz with decreasing the output level by 0.5dB on the RME.
It should give the same.

 

[Sokel]

Yep,various effects!

 

[Sokel]

Where is the fundamental of the 7350???
It appears ok to the other two,on the chart above a signal appears at 14700 but it's not the fundamental,it's not white.

 

[pkane]

Where is the fundamental of the 7350???
It appears ok to the other two,on the chart above a signal appears at 14700 but it's not the fundamental,it's not white.

Try lowering the level from 1.25 to 1.24 and see if it reappears.

 

[Sokel]

Try lowering the level from 1.25 to 1.24 and see if it reappears.

Yep.

 

[pkane]

Yep.

Strange. I'll need to take a look at that preset. I have the same issue. Seems 1.25dB peak (as per paper) isn't correct for some reason. It must be just below that level.

 

[MC_RME]

Re: intersample-overs

A more interesting test for would be to use the same preset with output at 44.1k, but recorded at 192k or even 384k to see the harmonic distortion generated by these sine waves. I think that's where you might see the effect (if any) of intersample overs. Here's ADI-2 Pro FS:

View attachment 259598

Please note that this is an older unit with AK4490 and no additional digital headroom (Vol 0 dB = 0 dBFS to the DAC with full scale signals). At around Vol -1.0 the harmonics are down. Also when doing this make sure B/T and Loudness are off. The choice of DA filter also changes the result significantly. But in no case the harmonics fold back into the audible range.

I would also advise against averaging. The raised noise floor is a typical artefact of ISP overload and should be shown.

 

[MC_RME]

Strange. I'll need to take a look at that preset. I have the same issue. Seems 1.25dB peak (as per paper) isn't correct for some reason. It must be just below that level.

Play Gain should be the same as when changing the (digital) volume at the player (here the ADI-2 Pro). I can confirm that 1.25 gives wrong signal and 1.24 does give the correct one. But using 1.24 and any gain at the ADI, for example +3 dB, does not give a wrong signal (just higher harmonics), so there is something other wrong than just the 'level'. Play Gain should be able to handle +4 dB with this signal.

 

[pkane]

Play Gain should be the same as when changing the (digital) volume at the player (here the ADI-2 Pro). I can confirm that 1.25 gives wrong signal and 1.24 does give the correct one. But using 1.24 and any gain at the ADI, for example +3 dB, does not give a wrong signal (just higher harmonics), so there is something other wrong than just the 'level'. Play Gain should be able to handle +4 dB with this signal.

Turns out the samples that are generated are very slightly over 0dBFS (by a tiny fraction). From MT perspective, that’s a valid signal since everything is in double floating point, but the audio library conversion routine to integer doesn't know what to do with anything greater than 0dBFS

I've adjusted the signal generator to ensure that samples are clamped to -1/1 range and now the fundamental no longer disappears. And noise is measured OK even with averaging

 

[pkane]

Quick update to v1.0.69 to fix the issue with samples exceeding 0dBFS level causing missing fundamental, as well as making scaling adjustments to the level sweep plot to fit all the data in:

https://app.box.com/s/ue7ll9xmvwogst817x2l1xg09opvgy47

Also updated the intersample-over preset, just the 5512.5Hz one, to have the correct level of 0.69dBFS. As before, copy these to C:\ProgramData\Multitone Analyzer\Presets folder on your computer: https://app.box.com/s/w5sehf9ow20ym5inh1tm3h2l010w1mcw

 

[Sokel]

Ok,intersample overloads works fine now but...
Level sweeps are a little crazy at the start,no matter the level:

-35db

-25db

-20db

-15db

 

[Sokel]

Also in Freq sweep -3db Play Gain should translate to about -3.2db with my settings but it seems to play near the -1db.

Normal level

Freq sweep level

 

[Rantapossu]

Quick update to v1.0.69 to fix the issue with samples exceeding 0dBFS level causing missing fundamental, as well as making scaling adjustments to the level sweep plot to fit all the data in:

https://app.box.com/s/ue7ll9xmvwogst817x2l1xg09opvgy47

Also updated the intersample-over preset, just the 5512.5Hz one, to have the correct level of 0.69dBFS. As before, copy these to C:\ProgramData\Multitone Analyzer\Presets folder on your computer: https://app.box.com/s/w5sehf9ow20ym5inh1tm3h2l010w1mcw

Thanks again!

Mixing of different sampling rates and FFT sizes from history on the same plot doesn't work:

Step 1. Take a measurement, 64k FFT:

Step 2. Change FFT to for example 256k (You don't have to make a new measurement, just go to the step 3):

Step 3. Open the previous measurement (From the step 1) from the history:

The previous 1 kHz fundamental is now 4 kHz and the graph is completely weird. If you do the 256 k FFT measurement first and then change to 64 k FFT, the fundamental on the history graph is 250 Hz.

 

[Rantapossu]

If you go a step further and actually do the 256k FFT measurement too, the graph is slightly different when both measurements are visible:

The fundamental is now 1 kHz on both graphs, but the 64k FFT measurement is still odd...

When I change the FFT setting back to 64k and open only the 64k FFT measurement from the history, it's ok now:

 

[Rantapossu]

And there's a minor graphical glitch too.

The channel colors aren't visible when you freshly start MT:

They appear during the first measurement when the graph refreshes the first time:

 

[MC_RME]

Mixing of FFTs worked for me, but you have to use the biggest FFT size as starter, then add same or smaller via history. It won't work the other way round. I used that extensively with both different FFT sizes and different sample rates. But it might been broken in the latest version, will check.

 

[Rantapossu]

Mixing of FFTs worked for me, but you have to use the biggest FFT size as starter, then add same or smaller via history. It won't work the other way round. I used that extensively with both different FFT sizes and different sample rates. But it might been broken in the latest version, will check.

This is measured with an another computer and freshly installed MT (256k FFT measurement, only the FFT size is changed to 64k FFT on the settings before opening the same measurement from the history):

The fundamental is 250 Hz.


Then two different measurements, 256k FFT first and 64k FFT after:

The 64 FFT one stops at 5.5 kHz when the both graphs are visible. Fundamental 1 kHz on both now.

Can you replicate this?

 

[Sokel]

Can you replicate this?

Yes,same thing:

 

[Sokel]

...and if I clear the existing measurement and just load them from history I get also this (no matter the order I open them):