Beta Test: Multitone Loopback Analyzer software

[Blumlein 88]

View attachment 229634

I get 146.7 which is -.2db.Close enough.

Looks pretty close to me.

 

[Blumlein 88]

So I was looking back at some FFT basics to see what I've forgotten. Is the Fundamental in the spectrum graph the single sided FFT without being doubled for the negative part of a double sided FFT which is usually ignored? That would make sense, and easy enough to compensate for.

 

[Blumlein 88]

So doing this manually for linearity, and compensating for the level difference in hardware etc. Below is a chart with signal levels sent to the DAC and the levels recorded by the ADC starting at -120 dbFS and going in 2 db steps to -144 dbFS. The column beside it shows the difference in what was recorded vs expected. I used 128k FFT, 4 averages, with Topping D10B feeding RME Babyface Pro FS at 48 khz with a 1 khz tone. Using more or fewer averages and larger FFT sizes causes the values to bounce around a couple tenths either way. Looks awfully good and less bouncing around than what was posted earlier.

 

[Sokel]

So doing this manually for linearity, and compensating for the level difference in hardware etc. Below is a chart with signal levels sent to the DAC and the levels recorded by the ADC starting at -120 dbFS and going in 2 db steps to -144 dbFS. The column beside it shows the difference in what was recorded vs expected. I used 128 FFT, 4 averages, with Topping D10B feeding RME Babyface Pro FS at 48 khz. Using more or fewer averages and larger FFT sizes causes the values to bounce around a couple tenths either way. Looks awfully good and less bouncing around than what was posted earlier.

View attachment 229657

It sure looks good!

 

[Blumlein 88]

Same conditions as before only using a 12 khz tone.

 

[pkane]

Same conditions as before only using a 12 khz tone.

View attachment 229660

Sorry, but did I miss something broken yet? Just trying to catch up.

 

[pkane]

I promise to practice it!

Roger with the "Level"!

But no matter how much I expand it, "SinAD" doesn't show up...

SINAD is just THD+N without the minus sign. I didn't feel it needed its own entry in the results

 

[Sokel]

Sorry, but did I miss something broken yet? Just trying to catch up.

Nope,we found out ourselfs,thank you

 

[Rantapossu]

SINAD is just THD+N without the minus sign. I didn't feel it needed its own entry in the results

Yes, I know.

I thought you wanted to have the same values both as a text and as a plottable graph. No need to add if it's designed to be so.

 

[pkane]

Yes, I know.

I thought you wanted to have the same values both as a text and as a plottable graph. No need to add if it's designed to be so.

The reason SINAD is available as a variable in the sweep is that it is in the positive dB range, which makes it easier to plot against other positive values, like SNR, for example. TD+N is negative and if you try to show TD+N vs SNR on the same plot, you'll get a very poorly scaled chart that includes a huge range of values, while SINAD vs SNR will be much easier to read.

 

[Moto]

Nope,we found out ourselfs,thank you

I know you guys figured it out but would you explain it to an ignorant guy like me?

 

[Rantapossu]

This is NOT a bug report @pkane . It's rather some kind of inquiry about background information.

So, how is the "No signal" -limit done? Does it affect only to the left channel?

When you have a strong signal on the left channel (and almost no signal at all on the right channel)

you can do this:

And when you have a strong signal on the right channel (and almost no signal at all on the left channel)

you can't do it anymore (At least with my equipment). It won't go past 40 Hz easily, here is the 7th try of the 40 Hz measurement:

When the graph is ready, there are only 5 points left and there's a lot of skips on the middle frequencies.

Some variable is likely out of range on the left channel between 40 Hz and +10 kHz and it most likely activates the "No sound" error checking.

Could you please explain how the "No sound" process is made and what is the variable and the value that activates it?

 

[pkane]

This is NOT a bug report @pkane . It's rather some kind of inquiry about background information.

So, how is the "No signal" -limit done? Does it affect only to the left channel?

When you have a strong signal on the left channel (and almost no signal at all on the right channel)

View attachment 229708

you can do this:

View attachment 229707


And when you have a strong signal on the right channel (and almost no signal at all on the left channel)

View attachment 229721

you can't do it anymore (At least with my equipment). It won't go past 40 Hz easily, here is the 7th try of the 40 Hz measurement:

View attachment 229709

When the graph is ready, there are only 5 points left and there's a lot of skips on the middle frequencies.

View attachment 229720

Some variable is likely out of range on the left channel between 40 Hz and +10 kHz and it most likely activates the "No sound" error checking.

Could you please explain how the "No sound" process is made and what is the variable and the value that activates it?

The initial recording is waiting for first channel to return something other than zeros/silence. I think the wait is 5 or 10 seconds. If nothing is detected, the recording will be stopped with an error message. If you have sound in R channel but L is silent, then you may want to record R+L instead of L+R.

 

[Rantapossu]

Thanks for the explanation!

It does indeed work:

Does the "No sound" detection wait for the actual complete silence? Or something like "better than -100 or -120 dBFS silence"? If the limit is complete 100% digital silence, then it was something else that prevented my previous "analog left channel middle frequency measurements" which obviously had some amount of background noise and distortion.

 

[pkane]

Thanks for the explanation!

It does indeed work:

View attachment 229728

View attachment 229729

Does the "No sound" detection wait for the actual complete silence? Or something like "better than -100 or -120 dBFS silence"? If the limit is complete 100% digital silence, then it was something else that prevented my previous "analog left channel middle frequency measurements" which obviously had some amount of background noise and distortion.

At the start of the recording, anything above -120dBFS is considered signal, below that -- silence.

 

[Rantapossu]

At the start of the recording, anything above -120dBFS is considered signal, below that -- silence.

Thanks, that explains everything! Now I can finally start to make my crossfeed measurements...

I think that the crossfeed measurements were previously impossible when we had only one channel measurements, because of the "No sound detection". Now with the stereo measurement ability I can use the other record channel to "cheat" the "No sound detection" and use the other channel to actually measure the crossfeed and then at the end tick the "cheat channel" to be unvisible on the plot.

 

[Blumlein 88]

I know you guys figured it out but would you explain it to an ignorant guy like me?

In my case, which I think Sokel has in mind, it is how to do very low level measurements and know what the results really mean. In the spectrum graph of a single tone, the level shown is 6.02 db lower than expected. So at a minimum all I can say is if I adjust for that 6.02 db then the results are at the levels expected vs Full Scale. I more or less confirmed it works and Sokel did a test that appears to confirm it as well. It also agrees with what is shown in the results tabulation if you open it from the square near the upper right corner.

I have a guess as to why this is, but only pkane can confirm it or explain how I am misinterpreting the results shown.

If I didn't answer what you were asking, then be a bit more specific and I'll explain it as far as I can.

 

[Rantapossu]

@pkane could you add the 5th "Legend" value called as "None" to turn off the legend completely?

It would be handy when the method of the measurement being used is irrelevant to the viewer or it could even cunfuse him/her. You could then PhotoShop the explanations later to the plot like Amir does with his measurements.

 

[Rantapossu]

Something very interesting came in the mail. Can anyone guess what it is?

 

[Blumlein 88]

Something very interesting came in the mail. Can anyone guess what it is?

View attachment 229814

Cosmos ADC?