Beta Test: Multitone Loopback Analyzer software

[Guy.Cohen]

There was a time I wanted to check the HUMP of the ES9822PRO. My conclusion that it has hump. I have another ADC by Nihtila which is based on AK5572 that gives the same performance without scaler as EIDA with scaler. More or less.
The Nihtila has a problem of rise of THD when one reaches close to it's 0dbfs (not as E1DA).
The reduction of the hump using the scaler is by 'holding' ES9822PRO input more or less fixed.

 

[MC_RME]

There was a time I wanted to check the HUMP of the ES9822PRO. My conclusion that it has hump. I have another ADC by Nihtila which is based on AK5572 that gives the same performance without scaler as EIDA with scaler. More or less.
The Nihtila has a problem of rise of THD when one reaches close to it's 0dbfs (not as E1DA).

The 'Nihtila' does not have this usse, the (all) AKM ADC has. That is standard behaviour, very different to the ESS ADC.

 

[Guy.Cohen]

The 'Nihtila' does not have this usse, the (all) AKM ADC has. That is standard behaviour, very different to the ESS ADC.

Correct

 

[pkane]

@Guy.Cohen , @pkane
IME, THD vs level is best suited to expose ESS hump as it allows to use a simple notch, helping dynamic range and distortion of the measurement chain. Preferably at a low frequency, like 200Hz or so.
"Too much action" at higher frequencies, be it from the single tone or the second component of a twin-tone, tends to blur the hump as the distortion is then transformed to broadband noise. We want the DAC to very slowly move through its output voltage range.

Actually, the best representation would be absolute level of the distortion residual (with much of the noise factored out) vs generator level.
A characteristic of the ESS hump is that after the onset of the specific distortion it remains almost constant in level... up to the point where the more regular distortion (H2, H3) finally takes over in absolute level.

THD vs Level looks like this, shape is very similar to IMD vs Level (but not IMD+N vs Level):

 

[MC_RME]

IMHO THD deeply buried into the noise (means in the FFT plot you can't identify the harmonics visually) should not give 'valid' THD results. For example lower level parts of the yellow curve could be dashed, so it becomes clear the shown value is whatever, but not clearly identified THD.

 

[pkane]

IMHO THD deeply buried into the noise (means in the FFT plot you can't identify the harmonics visually) should not give 'valid' THD results. For example lower level parts of the yellow curve could be dashed, so it becomes clear the shown value is whatever, but not clearly identified THD.

Good point, Matthias. But we know how to dig deeper into the noise floor, so I reran the same THD sweep with 1M FFT and 64 averages. The result is substantially the same, with maybe slight differences just below -55dB.

I should probably rerun the sweep with the APU notch filter to eliminate any influence of the ADC.

 

[Guy.Cohen]

Good point, Matthias. But we know how to dig deeper into the noise floor, so I reran the same THD sweep with 1M FFT and 64 averages. The result is substantially the same, with maybe slight differences just below -55dB.

View attachment 358889

I should probably rerun the sweep with the APU notch filter to eliminate any influence of the ADC.

My AK4499 shows that APU+SCALER into EIDA COSMOS almost the same as SCALER into EIDA COSMOS. I know the AK4499 has no HUMP but I think you will get the same results with adding the APU.
Once you get it maybe you overlay the two cases.
Thanks,
Guy

 

[Sokel]

In the "retry" feature,I think it would be better not to ask if it's set at zero.

Measuring my x-over for example the threshold is way higher that 10db when entering the filter territory and I have to confirm in every step that I don't want it to retry (0) times.

 

[pkane]

In the "retry" feature,I think it would be better not to ask if it's set at zero.

Measuring my x-over for example the threshold is way higher that 10db when entering the filter territory and I have to confirm in every step that I don't want it to retry (0) times.


View attachment 358935

Ah, that's a bug, thanks! When the number of errors equals or exceeds retry count, MTA will ask if you want to continue to retry to give you a chance to correct whatever issue may exist in the system and to continue. Unfortunately, when retries is set to 0, any first large error will trigger this message. Not quite what I intended I'll fix that.

 

[pkane]

My AK4499 shows that APU+SCALER into EIDA COSMOS almost the same as SCALER into EIDA COSMOS. I know the AK4499 has no HUMP but I think you will get the same results with adding the APU.
Once you get it maybe you overlay the two cases.
Thanks,
Guy

THD sweep with the E1DA APU notch is looking very different, with almost no hump and much lower THD:

 

[Guy.Cohen]

THD sweep with the E1DA APU notch is looking very different, with almost no hump and much lower THD:

View attachment 359084

Notch without scaler ?
Anyway, what is the conclusion?
The hump in your setup is purely ES9822PRO ?

 

[Guy.Cohen]

Also, nice if you overlay

 

[Music1969]

Doesn't Cosmos ADC itself have ESS ADC hump?

 

[pkane]

Notch without scaler ?
Anyway, what is the conclusion?
The hump in your setup is purely ES9822PRO ?

Notch without the scaler. I think the difference is the interaction with ADC. The notch eliminates ADC THD effects.

 

[Guy.Cohen]

Notch without the scaler. I think the difference is the interaction with ADC. The notch eliminates ADC THD effects.

Also sounds reasonable. In my case the scaler alone achieves the scaler plus APU. Almost the same.

 

[pkane]

Also sounds reasonable. In my case the scaler alone achieves the scaler plus APU. Almost the same.

For me Scaler didn't make a difference, same THD sweep results with or without the scaler.

 

[Guy.Cohen]

For me Scaler didn't make a difference, same THD sweep results with or without the scaler.

Did you play with compensation 2nd/3rd harmonics compare to 'stocked' cosmos? Integration parameters?

 

[pkane]

Did you play with compensation 2nd/3rd harmonics compare to 'stocked' cosmos? Integration parameters?

No

 

[Guy.Cohen]

No

Here is the DO300EX either: APU>SCALER>ADC or Scaler>ADC. You can see in my case results are more or less the same.

 

[pkane]

Another week, another update

This time, it's the preview version 1.1.9: https://app.box.com/s/ue7ll9xmvwogst817x2l1xg09opvgy47

A few fixes, enhancements, and some new features added:

1. Add: spectrum range display shows min and max values for all frequencies when spectrum is computed from averaged data
2. Add: IMD+N variable to Level and Frequency sweeps
3. Add: color selector property for 3rd measurement in Level and Frequency sweeps
4. Add: option to position distortion panel on the left, in the center, or to the right (default)
5. Add: line color indicators next to the Show 1 and Show 2 checkboxes are now clickable so you can change colors directly there, without opening the properties sheet
6. Fix: color indicators are now shown on start up (previously were invisible until first measurement)
7. Change: added automatic driver refresh on any audio device change (was previously disabled because it caused intermittent hangs)
8. Add: various useful settings to both sweep property sheets
9. Fix: in sweeps retry setting of 0 will no longer display a prompt when a large error occurs, effectively disabling retries

Details
1. Show signal swings around the average value in the spectrum plot. Lets you see how far the amplitude of each frequency deviates from the average during the whole capture. This will not be shown if average number is set to 1.


2. IMD+N is like THD+N, except it combines intermodulation and noise. This is similar to the measurements made by APx analyzers

3. Setting to change the color of the new, third variable in sweeps:



4. Panel Position setting on Spectrum plot. Possible values are Left, Center, and Right (default)

5. You can now click on the color indicators at the top to use Windows color selection dialog for plot series 1 or 2



7. The list of drivers in both, in and out drop-down lists will automatically update when audio devices are added, removed, or their properties changed (sample rate, bit length). Please test this, especially if you have a lot of ASIO devices. I'm hoping that I found the way to avoid the hang that would occasionally result from driver updates.