Using Cross Corelation to lower influence of ADC for DAC measurements

[FrenchFan]

Thanks, true, KSTR and FrenchFan. Both ADC channels see a linear combination of the signal and their own imperfections. This is what the actual cc process sees.

So the signal distortion and the individual ADC distortion can add such that e.g. the digitized data of one ADC channel has the same amplitude but 180deg phase (for a harmonic) compared to that of the other ADC channel. In this case CCA magnitude would give a larger harmonic level and CCA R0+ would discard the negative correlation result and the plot would show - 237dB just like in FrenchFan's measurement.

Net:
Same issue with quantitative harmonic distortion measurement as without CCA when the distortion of ADC is in the same order of magnitude as the distortion of the signal.
Thus CCA helps us on the ADC side regarding noise and nonharmonic spurs, but not with harmonic distortion.

@KSTR , @FrenchFan , did I get it right this time?

nonook

I think that's a good summary

For me, with my ADC, I really have a
difference between the 2 channels.

Are the causes well identified,
I don't know.

In any case, the internal distortion correction
of the ADC can lead to phase rotations on H2 and H3
that's for sure.

So for me nothing guarantees the phase between the 2 channels
of the ADC.

For the measurement of the amplitudes of the Harmonics with R0+ I think
that this is not going to be the right method.

Too sensitive to the phase variation between the channels.

 

[KSTR]

My takeaway so far is that the two CCA methods are welcomed and useful additions to our toolbox and like any other method they have pros and cons. There is no magic bullet.
CCA lowers system random noise and spurs while not affecting DUT noise, something no other method can offer.

 

[nanook]

Today it is colder at my place, it may be the cause that the measurements differ between the days when they are taken.

Ivan once mentioned, that the harmonics of the E1DA ADC(iso) depend on temperature.
He adjusts every single ADC but when operating it at a different temperature a re-adjustment would be necessary.
-> You shouldn't wonder when there is a slight variation.

He once showed a thermostat for the ES9822 in the E1DA ADC:

E1DA Cosmos ADC

Sorry for that question, not native english here, but I would like the confirmation: it's not possible to use it as an ADC in audio system because the USB output has to be connected to a computer, am I right?

audiosciencereview.com

 

[KSTR]

ESS's "distortion compensation" is based on cancelling by pre-distortion with the opposite characteristic, which makes it very sensitive to even smallest variations.

 

[JohnPM]

The behaviour of "Use cross correlation magnitude" is a bit confusing.
Should I select it together with "Cross correlation averaging" ?

The cross correlation options to use magnitude or invert the reference channel only have an effect when using cross correlation averaging.

 

[Sokel]

Ok,*I think* I got the hang of it:

I used a different DAC, a split mono signal,the rest as instructed and @Rja4000 window and I got near Amir's result.
I could let it more to clean up high but it didn't change much at the result as I watched the last 500 CCs.

 

[FrenchFan]

Ok,*I think* I got the hang of it:

View attachment 387067

I used a different DAC, a split mono signal,the rest as instructed and @Rja4000 window and I got near Amir's result.
I could let it more to clean up high but it didn't change much at the result as I watched the last 500 CCs.

Use "Cross correlaion averaging" + "Use cross correlation magn." it is very more smooth and give good result

 

[FrenchFan]

The cross correlation options to use magnitude or invert the reference channel only have an effect when using cross correlation averaging.

Just one info
Correlation don't stop at "stop at" number .

 

[FrenchFan]

I wanted to thank "ASR" for the platform
he gave us to develop an
idea.

I thank you all for participating in
this debate which is still fascinating.

I thank @JohnPM for allowing the
distribution of this feature.

By the way, I give him the GOLD medal
of the Olympic Games

Discipline: "CORRELATION"

In addition, I thanked "Enrico Rubiola"
For his paper
"The cross-spectrum experimental method"
(the Bible)
Email attached.

 

[nanook]

Honor to whom honor is due.

It's really a pleasure to "work" with enthusiasts!

 

[Rja4000]

A test with Virtins Multi Instrument 3.9.11
Source DAC D50 III Mono
997Hz 0dBFS 44.1kHz
ADC RME ADI-2/4 19dBu range
FFT Size 32k, Dolph Chebyshev 200
BW 20-20k Hz
AES-17 1 octave notch

EDIT: Re-tested with 3.9.11.1 here, and I get similar results than REW, fi selecting the added "In Phase" option..

Same with REW 5._40 Beta 50

As may be seen, both measure the same SNR (0.1dB difference is not significant)

REW is measuring lower THD and therefore gets higher SINAD (more in line with Notch measurementt)
REW also converges much faster.

But it's good to see this implemented in MI still.

 

[FrenchFan]

A test with Virtins Multi Instrument 3.9.11
Source DAC D50 III Mono
997Hz 0dBFS 44.1kHz
ADC RME ADI-2/4 19dBu range
FFT Size 32k, Dolph Chebyshev 200
BW 20-20k Hz
AES-17 1 octave notch

View attachment 387115

Same with REW 5._40 Beta 50

View attachment 387113

As may be seen, both measure the same SNR (0.1dB difference is not significant)

REW is measuring lower THD and therefore gets higher SINAD (more in line with Notch measurementt)
REW also converges much faster.

But it's good to see this implemented in MI still.

It is not a DAC it is a plane .

 

[Rja4000]

It is not a DAC it is a plane .

At around 230€, it's much cheaper than a plane

By the way, I have another one under test which is even cheaper and gives me this (with Notch) in stereo mode (Left channel here).
(The D50 III is measured with both channels summed)

 

[nanook]

I have another one under test which is even cheaper and gives me this (with Notch) in stereo mode

This looks very nice indeed, especially the higher order harmonics more or less missing.
@Rja4000 Which DAC is this?

 

[Sokel]

At around 230€, it's much cheaper than a plane

By the way, I have another one under test which is even cheaper and gives me this (with Notch) in stereo mode (Left channel here).
(The D50 III is measured with both channels summed)

View attachment 387118

Smells like SMSL DL200. (?)

 

[IVX]

Hi all
Here is a measurement on the D50III Jack output,
we are identical to the manufacturer measurement.

I just have to discuss with Ivan the distortion
never goes below 140db.

bye

the AVG of H3 level of Cosmos ADC/ADCiso is around -138db, at 1kHz -.5dbfs signal. The ADC should be well warned, let it work 15mins at 768k sampling(measure also at 768k). The worst case is -134db, and the best case is about -150db(I mark such units as THD++, THD+ if all harmonics are lower than -140db, THD if some of them reach -140db). H2 typically -145..-150db.
PS: currently Sunny has Cosmos ADCiso Grade A THD+ with such THD:

 

[FrenchFan]

the AVG of H3 level of Cosmos ADC/ADCiso is around -138db, at 1kHz -.5dbfs signal. The ADC should be well warned, let it work 15mins at 768k sampling(measure also at 768k). The worst case is -134db, and the best case is about -150db(I mark such units as THD++, THD+ if all harmonics are lower than -140db, THD if some of them reach -140db). H2 typically -145..-150db.
PS: currently Sunny has Cosmos ADCiso Grade A THD+ with such THD:

Thank Ivan for these precisions

 

[FrenchFan]

In attachment new Bug with R0+ correl = 1000

 

[nanook]

@FrenchFan: You mean the discrepancy in level 30dBFS sent and 28.9dB measured?

 

[FrenchFan]

@FrenchFan: You mean the discrepancy in level 30dBFS sent and 28.9dB measured?

Approximately -30db
-28.9 is the right value .
I tried this morning with 3 lines of -140dBFS
and no problem with R0+ at the DAC output.
It's random .