E1DA Cosmos ADC

[Music1969]

Hooked up a PC to measure at 384k. What is it that you see at 88k? Different than below?
View attachment 196382

Thanks for this. I'll PM you

 

[MC_RME]

Guys, for about a month I wasn't able to do my work and APU/Scaler designs were delayed as well. The reason is the same that makes many of you wake up and every morning check the news about Ukraine where innocent people suffering under putin's aggression.
Anyhow, I am close to the final version of both these products and want to ask your opinion regarding the APU notch performance. At the moment I have the choice to make APU 1kHz notch with -145db/-160db or -150db/-150db 2nd/3rd harmonics resolution. I did ask that in our discord and as I see people prefer the -150/-150 version.
This is about the opamp choice, OPA2156 makes -145db/-160db, Chinese opamp RT6863/SGM8261 -150db/-150db(less OLG and worse linearity but better input stage symmetry vs OPA2156) and a lot cheaper. Of course, I tried OPA1612 as well, some of them gave -148db/-155db but 2nd harmonic is noticeable varies from chip to chip, and I don't like the idea to test at such nanovolts levels each unit. 10kHz notch -130/-130 at least both variants. Please, let me know what do you think about that choice.

The green plot is a passive notch with corrections -50db for the 2nd and -54db for 3rd harmonic i.e. -158/-161db.
The yellow plot is APU result, the correction is -30db for all harmonics(its notch is active) i.e. -150.53/-151db.

And APU with OPA1612.

How did the OPA1656 perform in your tests?

 

[IVX]

OPA2156 makes -145db/-160db but -160db is about noise floor area so only multiple AVG may show that.

 

[MC_RME]

????
Was that a reply to my question, or someone else?

 

[IVX]

Oh, sorry, I mixed up 1656/2156, they perform nearly the same -145db for 2nd harmonic. As I know, it is nearly the same opamp with JFET-1656/CMOS-2156 input stage.

 

[MC_RME]

If I were you I would try the OPA1656 to be sure. The different input stage can make it perform differently in specific circuits. Or did I misunderstood and you tried the 1656 already?

 

[IVX]

as I said - they perform nearly the same -145db for 2nd harmonic. I tried about 3pcs of 1656, and I see no reason to try that more.

 

[Bamboszek]

Guys, for about a month I wasn't able to do my work and APU/Scaler designs were delayed as well. The reason is the same that makes many of you wake up and every morning check the news about Ukraine where innocent people suffering under putin's aggression.
Anyhow, I am close to the final version of both these products and want to ask your opinion regarding the APU notch performance. At the moment I have the choice to make APU 1kHz notch with -145db/-160db or -150db/-150db 2nd/3rd harmonics resolution. I did ask that in our discord and as I see people prefer the -150/-150 version.
This is about the opamp choice, OPA2156 makes -145db/-160db, Chinese opamp RT6863/SGM8261 -150db/-150db(less OLG and worse linearity but better input stage symmetry vs OPA2156) and a lot cheaper. Of course, I tried OPA1612 as well, some of them gave -148db/-155db but 2nd harmonic is noticeable varies from chip to chip, and I don't like the idea to test at such nanovolts levels each unit. 10kHz notch -130/-130 at least both variants. Please, let me know what do you think about that choice.
View attachment 196162
The green plot is a passive notch with corrections -50db for the 2nd and -54db for 3rd harmonic i.e. -158/-161db.
The yellow plot is APU result, the correction is -30db for all harmonics(its notch is active) i.e. -150.53/-151db.

And APU with OPA1612.
View attachment 196177

What about trying this composite combination?
Surely it would be really expensive but it is extremely clean and -180dB is achievable.

[PDF] Low-Distortion, Low-Noise Composite Operational Amplifier | Semantic Scholar

A composite operational amplifier is presented that is optimized for building low frequency ( 100 kHz) signal paths with very low distortion and noise. It consists of a cascade of two monolithic operational amplifiers and additional RC components for frequency compensation. The open-loop...

www.semanticscholar.org

 

[IVX]

the most my concern is the 2nd harmonic caused by the input stage symmetry lack. The driver stage to drive the notch is composite already.

 

[boXem]

Guys, for about a month I wasn't able to do my work and APU/Scaler designs were delayed as well. The reason is the same that makes many of you wake up and every morning check the news about Ukraine where innocent people suffering under putin's aggression.
Anyhow, I am close to the final version of both these products and want to ask your opinion regarding the APU notch performance. At the moment I have the choice to make APU 1kHz notch with -145db/-160db or -150db/-150db 2nd/3rd harmonics resolution. I did ask that in our discord and as I see people prefer the -150/-150 version.
This is about the opamp choice, OPA2156 makes -145db/-160db, Chinese opamp RT6863/SGM8261 -150db/-150db(less OLG and worse linearity but better input stage symmetry vs OPA2156) and a lot cheaper. Of course, I tried OPA1612 as well, some of them gave -148db/-155db but 2nd harmonic is noticeable varies from chip to chip, and I don't like the idea to test at such nanovolts levels each unit. 10kHz notch -130/-130 at least both variants. Please, let me know what do you think about that choice.
View attachment 196162
The green plot is a passive notch with corrections -50db for the 2nd and -54db for 3rd harmonic i.e. -158/-161db.
The yellow plot is APU result, the correction is -30db for all harmonics(its notch is active) i.e. -150.53/-151db.

And APU with OPA1612.
View attachment 196177

Hi Ivan, I like the -145/-160 version.

 

[IVX]

Hi Ivan, I like the -145/-160 version.

me too but chinese opamp is 3-4x cheaper

 

[audiofun]

me too but chinese opamp is 3-4x cheaper

Question: are those op amps using the same circuit? If that's the case, people can just swap the op amps to the ones they preferred. You can just use the cheapest one by default and let DIY users do the rest. Leaving a socket there is even better.

 

[IVX]

DIYers can do that indeed.

 

[audiofun]

DIYers can do that indeed.

So I'd suggest just leave sockets and cheap op amps there. customers can do whatever they want.
if sockets are not possible, please use SOIC-8 package. do not use SON one as those are relatively more difficult to deal with.

 

[G-K]

@IVX Today l made again measurements with my grade A Cosmos ADC. The measurement is without notch filter. I played arround with temperature to Cosmos ADC and Victor's generator. Unfortunally l had no thermometer connected. I am waiting that your Scaler will be available to be able to measure lower level signals. By the way, the input range of Cosmos ADC was setted to 3.5V and the output resistors of Victor's generator are shorted. Output voltage of Victor's generator was 2.78 V RMS

 

[G-K]

Own noise level of my Cosmos ADC was -129 dBA

 

[Grooved]

@IVX Today l made again measurements with my grade A Cosmos ADC. The measurement is without notch filter. I played arround with temperature to Cosmos ADC and Victor's generator. Unfortunally l had no thermometer connected. I am waiting that your Scaler will be available to be able to measure lower level signals. By the way, the input range of Cosmos ADC was setted to 3.5V and the output resistors of Victor's generator are shorted. Output voltage of Victor's generator was 2.78 V RMS

Impressive, never got the 2nd and 3rd so low with any source, even while changing chip temperature.

 

[Grooved]

Those formats are similar enough that you'd still get a seemingly valid signal, though the sample values would of course be garbled.

I did more test with the scope (I only did with 44.1kHz before) and found something strange, that may be linked to the problem to locked audio interface via SPDIF: I plugged the Cosmos ADC on another computer to test with scope, checked in Windows settings it was on 44.1kHz, but once I measured, the scope was showing me 48kHz on LRCK!
I then change in Windows setting to 48kHz, scope shown 48. Change to 88.2, scope shown 88.2,... back to 44.1, and this time scope shown 44.1
So I don't know if each time on first plug, it runs at 48, but better change it, going back and click OK in Windows settings.

The last recording I tried, after changing the MCLK wire and using again a 20ohm resistor, and after changing to 48kHz in Windows then back to 44.1kHz, the audio interface was locked from the start on the digital output board and never lost the sync, until I started a software with ASIO4ALL by default with Cosmos ADC turned on in ASIO4ALL. But I still had distorsion or length/pitch change

Overall, I have around 4.2Vpp on the three I2S pins, except on MLCK pin I added where I get between around 220mVpp only

Other strange thing I see in MCLK frequency :
- started at 44.1, but LRCK shows it's 48kHz
- changed to 88.2 (and LRCK showing 88.2) : MCLK f=7142.9 Hz and 218.8mVpp
- changed to 96 (and LRCK showing 96) : MCLK f=7142.9 Hz and 218.8mVpp
- back to 44.1 (and LRCK showing 44.1 this time): MCLK f=6250 Hz and 250mVpp
- changed to 176.4 (and LRCK showing 176.4): MCLK f=7142.9 Hz and 218.8mVpp
- back to 44.1 (and LRCK showing 44.1 this time): MCLK f=6250 Hz and 250mVpp

Anyway, I will check it again as it makes no sense to have MCLK with lower frequency than LRCK (and should be different between 44.1 multiples and 48 multiples)

EDIT: I didn't check enough and I can't measure MLCK correctly with the scope I have at home (not enough bandwidth), so I will try later with another one.

@IVX is it possible for you to check it in the firmware on which you added MCLK?

...
This is about the opamp choice, OPA2156 makes -145db/-160db, Chinese opamp RT6863/SGM8261 -150db/-150db(less OLG and worse linearity but better input stage symmetry vs OPA2156) and a lot cheaper. Of course, I tried OPA1612 as well, some of them gave -148db/-155db but 2nd harmonic is noticeable varies from chip to chip, and I don't like the idea to test at such nanovolts levels each unit. 10kHz notch -130/-130 at least both variants. Please, let me know what do you think about that choice.

I would pick -145dB/-160dB too, but that's right that a solution to be able to change opamp and also used -150/-150 would be great.

 

[IVX]

this is how I did add the MCLK to the GPIO4

 

[IVX]

Own noise level of my Cosmos ADC was -129 dBA

It is grade-0 usually but sometimes, users report that SNR is lower after they got the unit. I did not realize what happens but ES9822 does that from time to time. So, I believe fluctuations may go on the opposite side too, and measured here grade-A in your hands becomes grade-0.
Regarding -150-160db harmonics I wouldn't trust that ADC in this range of the distortions -140-145db maybe with a regular recalibration + thermostat board of course(or another t-conrol of the ADC case).