Beta Test: Multitone Loopback Analyzer software

[pkane]

I appreciate all you've done with this (and your other software). I've been figuring on replacing this old laptop next year anyway. I just looked and it is older than I thought. Was from 2012 with an Ivy Bridge CPU. So 10 years isn't bad. Still works fine for many purposes like being a Plex server for all my video streaming.

Dennis (and anyone else willing to test). I've uploaded a test version (1.0.20) that should do a little better with regards to ASIO buffer size. At least in my testing, I can lower buffer size to 64 or 128 bytes and still get consistent results, but a lot depends on sampling rate, ASIO driver and speed/floating point support of the CPU. Let me know if this helps at all:

https://app.box.com/s/u35txjg11zv74oj4wo7n0mn6q8yzw3gb

Oh, this version should also remember your driver choices across sessions.

EDIT: if you install this and run it, you will be prompted to update to the previous version, just say 'No' to the prompt.

Regards,

-Paul

 

[Blumlein 88]

Tried out version 20 on the Babyface. It keeps the drivers from the most recent session which is nice. I was able to do 192 khz at 1 million FFT with the right results.

However, there is still something wrong when you switch sampling rates. This is true switching to a higher rate or when switching from a high rate to a lower rate. It skips the warm up, goes right into record. You see the driver in the input and output windows alternate blinking out blank and back on. It runs very slowly. I see it use 19% CPU in this case. If you leave it something like 2 minutes it will eventually finish and give a result. The results look pretty much correct except the level will show around 1.5 db higher than it was for the 1 khz tone. If I then click and run it a second time, everything works right and once finished recording you get the results in a few seconds. In the case of the Babyface on 2nd or subsequent runs it only uses 5% of the cpu.

Also it acts the same way doing a level sweep or Jtest if you have changed sample rates.

Hopefully a couple other people will try version 20 and see if they have this happen when they switch sample rates. So it works better, and it does now always finish eventually.

 

[pkane]

Tried out version 20 on the Babyface. It keeps the drivers from the most recent session which is nice. I was able to do 192 khz at 1 million FFT with the right results.

However, there is still something wrong when you switch sampling rates. This is true switching to a higher rate or when switching from a high rate to a lower rate. It skips the warm up, goes right into record. You see the driver in the input and output windows alternate blinking out blank and back on. It runs very slowly. I see it use 19% CPU in this case. If you leave it something like 2 minutes it will eventually finish and give a result. The results look pretty much correct except the level will show around 1.5 db higher than it was for the 1 khz tone. If I then click and run it a second time, everything works right and once finished recording you get the results in a few seconds. In the case of the Babyface on 2nd or subsequent runs it only uses 5% of the cpu.

Also it acts the same way doing a level sweep or Jtest if you have changed sample rates.

Hopefully a couple other people will try version 20 and see if they have this happen when they switch sample rates. So it works better, and it does now always finish eventually.

Interesting. This does sound like some driver is sending continuous change notifications when no changes should be happening. I’ll put in the option to disable driver notifications, as you won’t lose much functionality if they are not processed.

 

[Blumlein 88]

Interesting. This does sound like some driver is sending continuous change notifications when no changes should be happening. I’ll put in the option to disable driver notifications, as you won’t lose much functionality if they are not processed.

Do you need me to try another interface?

 

[pkane]

Do you need me to try another interface?

Probably not yet. Let's try turning off the driver change notifications and see if that helps. The problem is it could be some other driver that is sending these out. It doesn't even need to be the driver you're currently using. Please try this version (1.0.21):

https://app.box.com/s/7iikb4fqs1ln7wr4hgsw0mhx4js21igz

In Settings, uncheck the Detect driver changes option and see if the problem still occurs:

 

[Blumlein 88]

That did the trick. Works correctly on the first and subsequent runs when changing sample rates.

 

[pkane]

That did the trick. Works correctly on the first and subsequent runs when changing sample rates.

Great! Sounds like some misbehaving driver was causing all the trouble With this setting unchecked you'll just need to restart Multitone when you connect another DAC or ADC device to the computer before the new device will show up in the list.

 

[Blumlein 88]

Hope I'm not being too bothersome. I noticed something with higher frequency tones. In this screenshot I'm running at 192 khz. It starts testing 10 khz where the 10th harmonic is above the bandwidth of that sample rate. The difference gets worse with increasing frequency. The reported THD will be quite a bit higher than the reported TD+N which it shouldn't be. I'm assuming those missing harmonics at -300 db get calculated in as 0 which throws off the result shown. Using 15 khz shows what I mean.

 

[Grooved]

@Blumlein 88
Did you use another interface?
I see that you mentioned the Babyface several times but your capture is always showing "Fireface"

@pkane
I tested the 1.0.21 version and I was getting more problem (looks like buffer problems) via ASIO4ALL (EDIT: I had "hardware buffer" option checked, I unchecked it and it's OK now), but it gets better with FLEXASIO:
I'm even able to use it with my old MOTU and once input/output are set in FLEXASIO, I can switch any sample rate and start Multitone, and every test is a success at first try now (no signal jumping, no need to start/stop/restart test...)
But I have inconsistent results if I use directly the MOTU ASIO driver

I get the same the same results for FLEXASIO with HiFi Cable (internal loopback), with MOTU TOSLINK loopback and MOTU ADAT loopback, except the phase values:

HiFi Cable:

MOTU TOSLINK:

MOTU ADAT:

 

[pkane]

Hope I'm not being too bothersome. I noticed something with higher frequency tones. In this screenshot I'm running at 192 khz. It starts testing 10 khz where the 10th harmonic is above the bandwidth of that sample rate. The difference gets worse with increasing frequency. The reported THD will be quite a bit higher than the reported TD+N which it shouldn't be. I'm assuming those missing harmonics at -300 db get calculated in as 0 which throws off the result shown. Using 15 khz shows what I mean.

View attachment 219243

TD+N is calculated for the frequency range you've selected (20-20k), while THD is calculated over the full 0-96k range. I'll fix this.

 

[Grooved]

Hope I'm not being too bothersome. I noticed something with higher frequency tones. In this screenshot I'm running at 192 khz. It starts testing 10 khz where the 10th harmonic is above the bandwidth of that sample rate. The difference gets worse with increasing frequency. The reported THD will be quite a bit higher than the reported TD+N which it shouldn't be. I'm assuming those missing harmonics at -300 db get calculated in as 0 which throws off the result shown. Using 15 khz shows what I mean.

View attachment 219243

I get the opposite with the old MOTU (better THD and distortion on two H only, but higher noise than with your interface):
EDIT: just saw the THD result explained by @pkane

 

[pkane]

Rolling up all the changes in the last few days into a whole new version 1.0.22:

Changes in 1.0.22​

• Change: improved detection of the start of the playback waveform
• Add: option to turn off audio new/removed driver notifications to alleviate too many updates caused by some drivers
• Fix: THD is now calculated within the selected frequency range (was over the whole bandwidth)
• Change: optimizations to allow for smaller ASIO buffer sizes, faster sampling rates, and better performance on slower computers

 

[Blumlein 88]

@Blumlein 88
Did you use another interface?
I see that you mentioned the Babyface several times but your capture is always showing "Fireface"

It is still the Babyface. RME uses a unified driver for several products. The driver shows up as Fireface. The Fireface was originally a Firewire interface so a bit of a legacy name. Current Fireface devices use either USB or Thunderbolt.

So Paul is rolling out the improvements today. Already up by 3 versions from yesterday. Good work!

 

[pkane]

So Paul is rolling out the improvements today. Already up by 3 versions from yesterday. Good work!

Good thing I'm not charging by version numbers!

 

[Grooved]

It is still the Babyface. RME uses a unified driver for several products. The driver shows up as Fireface. The Fireface was originally a Firewire interface so a bit of a legacy name. Current Fireface devices use either USB or Thunderbolt.

So Paul is rolling out the improvements today. Already up by 3 versions from yesterday. Good work!

Thanks! I used Fireface, from the Firewire one, but I didn't used a Babyface and didn't know it was showing up like this. I understand know
Which one is yours? First one, Pro or Pro FS?

And thanks @pkane for the work!

 

[Blumlein 88]

Thanks! I used Fireface, from the Firewire one, but I didn't used a Babyface and didn't know it was showing up like this. I understand know
Which one is yours? First one, Pro or Pro FS?

And thanks @pkane for the work!

Pro FS.

 

[Blumlein 88]

Good thing I'm not charging by version numbers!

Yes, I almost added to the end of my post that we should get that guy a raise.

 

[Sokel]

Ok,I have a little summary of the things that i can or cannot do.
First the best I could do.

It seems that wasapi works better for me.

What I could not do is mix -n- match between wasapi and ASIO between e-mu and my other dacs,a message that "invalid number of input channels 2 must be in the range [0.0].Every time I tried it,I restarted and once I restarted the PC too.
In one occasion it froze when I changed the color of a measurement in history.

Other than that I really enjoy it,the only thing I haven't measured are my power amps (and the power sockets on the wall),I'm too scared for that and E-MU clips like crazy with anything above 4 volts,i have to make voltage dividers I think.

Thanks Again!

 

[pkane]

Ok,I have a little summary of the things that i can or cannot do.
First the best I could do.

View attachment 219391

It seems that wasapi works better for me.

What I could not do is mix -n- match between wasapi and ASIO between e-mu and my other dacs,a message that "invalid number of input channels 2 must be in the range [0.0].Every time I tried it,I restarted and once I restarted the PC too.
In one occasion it froze when I changed the color of a measurement in history.

Other than that I really enjoy it,the only thing I haven't measured are my power amps (and the power sockets on the wall),I'm too scared for that and E-MU clips like crazy with anything above 4 volts,i have to make voltage dividers I think.

Thanks Again!

Mixing WASAPI and ASIO shouldn't be a problem. But, I've had someone else report using two different ASIO drivers giving the same error, so I want to see if I can track this down. What specific drivers give this error when you use them? How are each configured, what sample rate, number of bits, etc.? What Multitone settings are used (WASAPI exclusive or shared, ASIO sampling rate)?

 

[Sokel]

Mixing WASAPI and ASIO shouldn't be a problem. But, I've had someone else report using two different ASIO drivers giving the same error, so I want to see if I can track this down. What specific drivers give this error when you use them? How are each configured, what sample rate, number of bits, etc.? What Multitone settings are used (WASAPI exclusive or shared, ASIO sampling rate)?

Ok,for example the Khadas Tone 1 driver (V4.86.0) which is popular here and the E-MU driver.
Same sample rate for both (48000),WASAPI exclusive,tried both 24 and 32 bits.