Using Cross Corelation to lower influence of ADC for DAC measurements

[KSTR]

Connecting the shorted inputs to GND or not doesn't make a difference. I haven't checked what happens when I short the inputs directly at the jack's solder pins but would assume there will be some influence as the layout is not fully hum-cancelling and the alumium box provides little magnetic field attenuation at mains frequencies.

 

[DonH56]

Connecting the shorted inputs to GND or not doesn't make a difference. I haven't checked what happens when I short the inputs directly at the jack's solder pins but would assume there will be some influence as the layout is not fully hum-cancelling and the alumium box provides little magnetic field attenuation at mains frequencies.

OK. I just know in the past shorting + to - (no ground tie) usually worked better for me than shorting them to ground, but it's been ages since I did any serious audio testing, and I do not own an E1DA. It is no-doubt dependent upon the circuit and implementation, and most of my recent experience has been with much higher-speed signals. Ground is too often an enemy to differential signals...

 

[AnalogSteph]

https://www.edn.com/measuring-small-signals-accurately-a-practical-guide/

Before I read the rest of the thread, the idea of vector averaging immediately made me think of
foo_dsp_centercut, which I think does exactly that and IME is very effective in cleaning up crackly stereo rips of mono LPs. A modern recompile suitable for Foobar2000 v2.x is available. (They don't call Foobar the Swiss army knife of audio players for nothing. ) This may prove handy for someone.

 

[nanook]

Nice tool if you want to primarily listen to the vocals (Center Cut - Center)
And for those who are into Karaoke - the tool removes the vocals quite impressively (Center Cut - Sides)

The download for Foobar 2.x (64-bit) is in this post:

foo_dsp_centercut - Page 2

foo_dsp_centercut - Page 2

hydrogenaud.io

 

[restorer-john]

OK. I just know in the past shorting + to - (no ground tie) usually worked better for me than shorting them to ground, but it's been ages since I did any serious audio testing, and I do not own an E1DA. It is no-doubt dependent upon the circuit and implementation, and most of my recent experience has been with much higher-speed signals. Ground is too often an enemy to differential signals...

The Cosmos is an odd bird with its super low and wildy variable input impedance.

My Quantasylum QA403 analyzer is 100k/200k (standard) for unbalanced/differential and I have 50R 'shorting' BNC plugs on all the 4 inputs unless I am using them. The QA has switched gain and switched attenuators up front of the A/D whereas I have no idea what's up front of the Cosmos A/D converter- probably not much other than resistor based attenuators.

Shorting + to - is worse for noise for me, than proper short to ground/chassis. But the QA403 is also isolated from supply/usb ground.

 

[FrenchFan]

Not sure if that's really required, since cross alignment is, first, used to identify time/phase differences between signal, isn't it ?

You may just change the ADC scale.
I'm using 10V as a target for Scaler, personnaly.

Hello Rja4000

As promised I tried to go to
10V on the ADC.

If the D50III is in 5V position and
the output is at 0 dB I can only work
at 50% of a full bit signal on the ADC, the 2nd gain
of the scaler is 7db.

With this level I cannot with a correlation
of 1000 reduce the noise sufficiently, and in addition the noise
of the ADC at 10v is a little higher than at 4.5v.

I arrived at -116.2 dBFSA without correlation
and at -124.4dBFSA with correlation.

The calculated noise of the signal being -134.5 dBFSA.

I tried to push the Nb of correlation with overlap to 8000
but it is still not enough, the noise of the signal with this input level
must be even lower.

So in these conditions I arrive at the limit of the possibilities
of the correlation in these conditions.

I went back to 4.5 and I saw that at -0.5db on the D50III
I did not saturate.

the measurement in this condition gives -126.62 dBFSA for the THD+N

 

[FrenchFan]

Question for All

Do you think we can ask
John Mulcahy (REW) to look at what was
said in the discussion to get his opinion.

There is his email in Help-> about Rew of (REW)

Thanks

 

[restorer-john]

Question for All

Do you think we can ask
John Mulcahy (REW) to look at what was
said in the discussion to get his opinion.

There is his email in Help-> about Rew of (REW)

Thanks

Can't you just ask him? Why does poor John need to be bombed with the same question from multiple ASR members?

 

[Rja4000]

Simplest is probably to name @JohnPM here.

 

[nanook]

I think it would be great to have the cross-correlation built into the REW.
This would allow to lower the noise (yet not suppressing the DUT noise) without using an extremely large FFT.

 

[KSTR]

I would expect a 2-channel cross-correlation aquisition mode will require the multichannel professional version of REW but I'd be happy to pay the very reasonable fee for the pro version.

 

[FrenchFan]

I would expect a 2-channel cross-correlation aquisition mode will require the multichannel professional version of REW but I'd be happy to pay the very reasonable fee for the pro version.

Left and Right it enought , exist today No ???
Is possible to work with these 2 channels in same time or no ??

 

[FrenchFan]

I think it would be great to have the cross-correlation built into the REW.
This would allow to lower the noise (yet not suppressing the DUT noise) without using an extremely large FFT.

I think it is not a problem , REW read wave today , i work more on wav record
is better to save measure .

 

[FrenchFan]

Simplest is probably to name @JohnPM here.

Excuses me , but I don't understand

 

[nanook]

Excuses me , but I don't understand

I think Rja4000 proposed that you contact John (the author and owner of REW) via this forum.
You can do this by getting him notified about this thread just like my post #25, where I asked KSTR to join the discussion. The "@" initiates the notification.

Since you introduced this cross-correlation method and did all the work, I think you are the right person to contact John, explain the idea and ask if he can build cross-correlation option into REW.
There's already the option for "coherent averaging" in the menu of the RTA window; since cross-correlation needs two channels, the implementation is probably not fully straight forward.

 

[Rja4000]

I think Rja4000 proposed that you contact John (the author and owner of REW) via this forum.
You can do this by getting him notified about this thread just like my post #25, where I asked KSTR to join the discussion. The "@" initiates the notification.

Since you introduced this cross-correlation method and did all the work, I think you are the right person to contact John, explain the idea and ask if he can build cross-correlation option into REW.
There's already the option for "coherent averaging" in the menu of the RTA window; since cross-correlation needs two channels, the implementation is probably not fully straight forward.

It would be even more efficient if John was part of the discussion here.
He was visiting ASR no later than yesterday, it seems.

 

[FrenchFan]

@Rja4000
@nanook

What do I do??
I contact him via (at)johnPM and summarize
the situation for him so he can look at the discussion??

 

[Rja4000]

@Rja4000
@nanook

What do I do??
I contact him via (at)johnPM and summarize
the situation for him so he can look at the discussion??

Why not ?

 

[KSTR]

Left and Right it enought , exist today No ???
Is possible to work with these 2 channels in same time or no ??

No, please see REW help here : https://www.roomeqwizard.com/help/help_en-GB/html/multichannelcapture.html#top

Multi-input capture is a Pro upgrade feature that allows REW to capture multiple inputs simultaneously for measurement or for the RTA

[bold mine]

You can only chose between L or R channels as the input channel used for the RTA.

But let's wait what John will have to say, he sure is reading the thread already. It might be possible to use the other channel in a similar way like it is already used for 2-channel impedance measurements even with the basic REW version:

Impedance Measurement

 

[Blumlein 88]

@Rja4000
@nanook

What do I do??
I contact him via (at)johnPM and summarize
the situation for him so he can look at the discussion??

This should get his attention or you could PM private message his account here.

PS they are called conversations on ASR.