Beta Test: Multitone Loopback Analyzer software

[pkane]

It is pretty impressive it gets things this clean, this close together at higher frequencies.

View attachment 224254

Please try the updated version (uninstall the old one, first):

MultitoneSetup.zip | Powered by Box

app.box.com

In my testing, this should not exceed 0dBFS for full bandwidth up to 192k when Play Gain is set to 0. If it still clips, take a look at the waveform plot to see if it's caused by intersample-overs, since the peaks of the saw-teeth come very close to the 0dBFS and of course also exhibit Gibbs phenomenon due to limited bandwidth. Here's 44.1k with 22.05k as the high harmonic limit.

And zoomed in on the peak:

 

[pkane]

Please try the updated version (uninstall the old one, first):

MultitoneSetup.zip | Powered by Box

app.box.com

In my testing, this should not exceed 0dBFS for full bandwidth up to 192k when Play Gain is set to 0. If it still clips, take a look at the waveform plot to see if it's caused by intersample-overs, since the peaks of the saw-teeth come very close to the 0dBFS and of course also exhibit Gibbs phenomenon due to limited bandwidth. Here's 44.1k with 22.05k as the high harmonic limit.

View attachment 224256

And zoomed in on the peak:
View attachment 224257

Forgot to mention, there's now also a reverse-sawtooth waveform basically the same, but flipped around the 0 line:

 

[Rantapossu]

Wanted to ask your help to test MT output to a multi-channel device. This should work with WASAPI and possibly ASIO drivers, outputting to channels other than just left and right.

This was the most important feature I asked before. Thank you so much!

I didn't have my DAC or optical cable available, so I couldn't test properly. But there were 4 ASIO output channels available, the analog channels 1 and 2 worked like L and R before (Like it should be) and when I changed to channels 3 or 4, there was no output on the analog outputs (Like it should be). Most likely the SPDIF channels 3 and 4 were active then, but as I said, I didn't have all my test equipment available to verify that. Brilliant job!

I was like a kid on a Christmas morning when testing this!

 

[Blumlein 88]

Ok it did not fix it for me on the sawtooth. It shows 0.0 db, but still just clips the negative part of the waveform. I have to reduce it again to -2 db to get clean results. It could be intersample overs as I'm not running the two devices with locked clocks.

I also stumbled across this strange issue. I ran with both devices at 44 khz and output at 44 khz and 88 khz. Look at the wave form for the latter when the recording device is running at 88 khz. It is like it is actually half the frequency. They do more or less match when both are set to 44 khz.

And then this if I use 44 khz output and 176 khz input. A 4 to 1 ratio. Did I break it good enough this time? Sorry.

 

[Sokel]

I didn't manage to break it
(only a little)
At 1 Khz all working fine:

At 20Hz it takes 2 or three times to starts,the first time only says "Generate",the second time starts but no signal and the third time works (and the preview has that glitch which is probably me)

But when it finishes it's ok (best result ever made by E-MU!)

 

[Sokel]

Ok it did not fix it for me on the sawtooth. It shows 0.0 db, but still just clips the negative part of the waveform. I have to reduce it again to -2 db to get clean results. It could be intersample overs as I'm not running the two devices with locked clocks.

I also stumbled across this strange issue. I ran with both devices at 44 khz and output at 44 khz and 88 khz. Look at the wave form for the latter when the recording device is running at 88 khz. It is like it is actually half the frequency. They do more or less match when both are set to 44 khz.

View attachment 224269

And then this if I use 44 khz output and 176 khz input. A 4 to 1 ratio. Did I break it good enough this time? Sorry.

View attachment 224270

That's my waveform for comparison

 

[pkane]

Ok it did not fix it for me on the sawtooth. It shows 0.0 db, but still just clips the negative part of the waveform. I have to reduce it again to -2 db to get clean results. It could be intersample overs as I'm not running the two devices with locked clocks.

I also stumbled across this strange issue. I ran with both devices at 44 khz and output at 44 khz and 88 khz. Look at the wave form for the latter when the recording device is running at 88 khz. It is like it is actually half the frequency. They do more or less match when both are set to 44 khz.

View attachment 224269

And then this if I use 44 khz output and 176 khz input. A 4 to 1 ratio. Did I break it good enough this time? Sorry.

View attachment 224270

Clipping is almost certainly intersample-overs. Check that the blue waveform on the Wave plot doesn't exceed the bounds from -1 to 1. If it doesn't, but the recording clips anyway, it's due to the intersample overs. This is possible even with the same in/out rates or synch'ed clocks, due to internal oversampling by the DAC, but even more likely when two independent clocks are used (DAC and ADC).

You broke it good in that one I'll check, but I'm pretty sure the sawtooth test signal is fine. The problem is with the horizontal (time) scale on the waveform plot -- it appears to scale everything to the recording sampling rate, even though the blue waveform is generated for the output sampling rate. Only an issue with the plot itself, and only when the input/output sampling rates differ. I'll fix it!

 

[Sokel]

At last,I succeeded to totally break it,first by freezing it and get an BSOD as a bonus when I killed it in task manager
I t had to do with changing FFt length,probably my machine can't handle over 262144 in that but it didn't happened there,it happened when I changed it back to that length after I was at 524288.

 

[Sokel]

Oh,and in one occasion repeated the measurement by itself,probably didn't like the 103 SINAD and gave me 104 the second time.

 

[pkane]

At last,I succeeded to totally break it,first by freezing it and get an BSOD as a bonus when I killed it in task manager
I t had to do with changing FFt length,probably my machine can't handle over 262144 in that but it didn't happened there,it happened when I changed it back to that length after I was at 524288.

Large FFTs on slower computers without a floating point coprocessor/accelerator a the CPU level can take a long time the first time you use that FFT size. It should run faster the next time if you let it finish that first run, as it's doing hardware testing to try to find a faster way to compute that size FFT in the future.

As before, BSODs aren't caused by the app. You probably killed it while the audio driver was doing something important and didn't like to be interrupted

 

[Sokel]

Large FFTs on slower computers without a floating point coprocessor/accelerator a the CPU level can take a long time the first time you use that FFT size. It should run faster the next time if you let it finish that first run, as it's doing hardware testing to try to find a faster way to compute that size FFT in the future.

As before, BSODs aren't caused by the app. You probably killed it while the audio driver was doing something important and didn't like to be interrupted

Yes,the message was telling something of a "stopcode"

 

[pkane]

Oh,and in one occasion repeated the measurement by itself,probably didn't like the 103 SINAD and gave me 104 the second time.

MT prefers higher SINAD numbers, as that's what I train it on

 

[Sokel]

And yes,I left it alone at 524288 and it eventually finished fine.The problem is when I stop it and start again.

 

[Sokel]

Finished the million too (took almost a minute)
CPU was at 23-30% the most and memory 4.2/8.

 

[pkane]

Finished the million too (took almost a minute)
CPU was at 23-30% the most and memory 4.2/8.

View attachment 224304

Make sure you exit MT cleanly (close it instead of killing it) after you do this for the first time -- it'll save the results of FFT/hardware testing and use them the next time you ask for 1M sized FFT.

 

[Sokel]

Make sure you exit MT cleanly (close it instead of killing it) after you do this for the first time -- it'll save the results of FFT/hardware testing and use them the next time you ask for 1M sized FFT.

I just realized I had the random amplitude on all the above,if that matters in any way

 

[pkane]

I just realized I had the random amplitude on in all the above,if that matters in any way

That's only used to generate multitone signals. For sawtooth the amplitude for each harmonic is defined by the waveform itself. With sawtooth, square wave or triangle test signals you can only adjust the overall level of the signal, but not the individual harmonics.

 

[Sokel]

(it's now easy to see why I'm a sucker for stuff advertised like Gigabytes "ultra durable",etc.,that's the place electronics experience violent deaths )

 

[Sokel]

The FR sweep is interesting

 

[Sokel]

Is it the nature of the signal not to work over 7K? (tested it)