Beta Test: Multitone Loopback Analyzer software

[pkane]

I updated,yes.
I think I pinned down and has to do with scaling.
For example if the existing measurement is silence and I add a 1 Khz measurement just doesn't do it.
If the scale is similar is ok.

If you add a measurement that's at a different FFT size, it will not add it. That's because it will not fit on the same FFT display with differently sized results.

 

[Sokel]

If you add a measurement that's at a different FFT size, it will not add it. That's because it will not fit on the same FFT display with differently sized results.

Oh,that explains it,a lot of different FFT sizes used (I think,I'll test again)

 

[Blumlein 88]

Here is a post from 3 years ago explaining why I thought using 1/4 the sample rate was good along with some results for March Audio 1 DAC.

On DAC Linearity Measurement

Actually, isn't the xu208 processor doing the same thing when volume is decreased? XU208 is doing the same thing as DAC digital volume control. It's all well explained here.

www.audiosciencereview.com

 

[pkane]

Here is a post from 3 years ago explaining why I thought using 1/4 the sample rate was good along with some results for March Audio 1 DAC.

On DAC Linearity Measurement

Actually, isn't the xu208 processor doing the same thing when volume is decreased? XU208 is doing the same thing as DAC digital volume control. It's all well explained here.

www.audiosciencereview.com

The fun thing with MT is that you can set your own measurement signal, including 12k or 1k or anything else. In fact, you can even use multiple tones, although the largest amplitude one will be used to measure nonlinearity. In fact, you could even use a square wave or sawtooth!

 

[Sokel]

Yep,that was it,I didn't know it before!

 

[pkane]

The fun thing with MT is that you can set your own measurement signal, including 12k or 1k or anything else. In fact, you can even use multiple tones, although the largest amplitude one will be used to measure nonlinearity. In fact, you could even use a square wave or sawtooth!

Here you go: a linearity sweep using 1k square wave

 

[Blumlein 88]

Here you go: a linearity sweep using 1k square wave

View attachment 229483

Looks good, but are those levels below -125 db accurate to what the DAC under test is doing?

When I've tested DACs like in the post I linked I don't get those swings at the very lowest bits. I think the issue is with line level inputs to ADCs the EIN is not good enough.

 

[pkane]

Looks good, but are those levels below -125 db accurate to what the DAC under test is doing?

When I've tested DACs like in the post I linked I don't get those swings at the very lowest bits.

It is, down to the noise floor, of course. But at -144dB, here's what I get. You can see that the fundamental amplitude is very, very measurable:

 

[Blumlein 88]

It is, down to the noise floor, of course. But at -144dB, here's what I get. You can see that the fundamental amplitude is very, very measurable:

View attachment 229486

What size FFT did you use on that?

 

[pkane]

What size FFT did you use on that?

128k at 96k sampling rate

 

[Blumlein 88]

128k at 96k sampling rate

So I wonder why doing it the other way shows better linearity? I do of course need to redo such measurements of the gear I have sitting around now to confirm. This was one of the reasons I was looking for a process gain adjustment. I can do it with calibration procedure just a little tedious. I'll mess with it some when I get the chance and see what happens. I'll also do duplicates of your version with the linearity option in it when you are ready to release it.

 

[Sokel]

when you are ready to release it.

Now please.

 

[pkane]

So I wonder why doing it the other way shows better linearity? I do of course need to redo such measurements of the gear I have sitting around now to confirm. This was one of the reasons I was looking for a process gain adjustment. I can do it with calibration procedure just a little tedious. I'll mess with it some when I get the chance and see what happens. I'll also do duplicates of your version with the linearity option in it when you are ready to release it.

Not sure, but a lower resolution ADC may not show errors in the last bit or two. Not that ADI-2 Pro is perfect as a measurement device but it seems to produce a consistent result each time I measure it, so it's not purely noise causing the jumps. Amir usually measures down to -120dB, and I do see ADI-2 (this is the DAC, not the Pro version) linearity starting to deviate from straight line by -120dB. This mostly agrees with my measurements:

 

[Blumlein 88]

Not sure, but a lower resolution ADC may not show errors in the last bit or two. Not that ADI-2 Pro is perfect as a measurement device but it seems to produce a consistent result each time I measure it, so it's not purely noise causing the jumps. Amir usually measures down to -120dB, and I do see ADI-2 (this is the DAC, not the Pro version) linearity starting to deviate from straight line by -120dB. This mostly agrees with my measurements:

How would a lower resolution ADC show the correct values below the 20th bit? I don't follow the thinking there. I've pondered this for some time, and one difference is I am only turning on each bit level at any time. Not any combinations. So each step is 6.02 db. The difference in discrete and integrated levels.

 

[Blumlein 88]

Linearity errors are done in two ways, INL and DNL. INL is what your test is doing (and is the one you see done like here on ASR and Stereophile), and maybe it needs doing both ways.

 

[pkane]

How would a lower resolution ADC show the correct values below the 20th bit? I don't follow the thinking there. I've pondered this for some time, and one difference is I am only turning on each bit level at any time. Not any combinations. So each step is 6.02 db. The difference in discrete and integrated levels.

Here's the linearity result for Khadas board (I think that's what March DAC used internally, right?). There's deviation from linearity starting at about -100dB and slowly increasing. Not sure about your way of measurement, I've not had time to review it yet. With MT, you can set the step to 6.02 dB and start at -144dB if you want to repeat your measurements at bit level.

 

[Sokel]

Here's the linearity result for Khadas board (I think that's what March DAC used internally, right?). There's deviation from linearity starting at about -100dB and slowly increasing. Not sure about your way of measurement, I've not had time to review it yet. With MT, you can set the step to 6.02 dB and start at -144dB if you want to repeat your measurements at bit level.

I don't think I'll get close to that measuring it with E-MU but I will sure try.

 

[pkane]

Linearity errors are done in two ways, INL and DNL. INL is what your test is doing (and is the one you see done like here on ASR and Stereophile), and maybe it needs doing both ways.

As far as I know, INL vs DNL is just the integrated error (sum of all errors) vs individual 'bit' error. Your method of testing at a 6.02dB interval is fairly coarse. DNL requires a test at each single bit (LSB) increment, which would require 2^24 tests to be run, which is probably a bit of an overkill

 

[DonH56]

IEEE Standards 1057 (waveform recorders, the grandfather), 1241 (ADCs), 1658 (DACs) etc. define testing methods and terms, including INL/DNL/ENOB/SINAD/SNR/SFDR/etc. I have lost track of the numbers and revs since it's been a decade or so since I was involved with them.

 

[pkane]

IEEE Standards 1057 (waveform recorders, the grandfather), 1241 (ADCs), 1658 (DACs) etc. define testing methods and terms, including INL/DNL/ENOB/SINAD/SNR/SFDR/etc. I have lost track of the numbers and revs since it's been a decade or so since I was involved with them.

Hi Don,

You might be surprised to find that AES has made their own version of many of these standards that don't always agree with IEEE AP has mostly followed AES standards, from what I can tell from their website.