E1DA Cosmos ADC

[arvidb]

Let's hope this is useful to somebody reading...

It's good to get an understanding of the capabilities of this measurement system, and not just from those who lucked out and got the best "specimens" (or happened to take measurements under perfect conditions etc). It helps me get a better understanding of when to worry when doing actual measurements of other equipment.

My intention when buying these was to get something that's good enough that I won't have to worry about squeezing out those last decibels of fidelity, and your THD+N figure of well below -100 dB would still be good enough for that, for my use! (Which is sanity-checking/characterising fidelity of hobby audio setups and builds.)

 

[BKDad]

Are you sure that your D10s is playing at 0dBFS?

That REW is showing -5.21dBFS at the ADC's output means that your D10s is outputting about 1.56Vrms, which would be the result of a test signal at about -2dBFS amplitude.

-0.5 dBFS according to REW. Just double checked. I did not measure the level with a Voltmeter, though. In the past, I found that what REW claimed was pretty accurate.

 

[BKDad]

It's good to get an understanding of the capabilities of this measurement system, and not just from those who lucked out and got the best "specimens" (or happened to take measurements under perfect conditions etc). It helps me get a better understanding of when to worry when doing actual measurements of other equipment.

My intention when buying these was to get something that's good enough that I won't have to worry about squeezing out those last decibels of fidelity, and your THD+N figure of well below -100 dB would still be good enough for that, for my use! (Which is sanity-checking/characterising fidelity of hobby audio setups and builds.)

I hear you. Too many variables from sample to sample variations, perhaps specimen selection (I like your idea there), and so on.

 

[IVX]

BKDad, pls try to short APU notch inputs better if right on the 3.5mm plug and show me FFT from the ADC connected to the APU. This way we'll make sure that the APU has a typical noise floor, like other APUs(my screenshot below is without a calibration file). Also, it's a good idea to use 48k or 96k of sampling than 384k.

Also, you can check the noise performance of d10. Use Scaler with 26db gain and -60dbfs DAC's output to read THD+N A-weighted.

 

[BKDad]

BKDad, pls try to short APU notch inputs better if right on the 3.5mm plug and show me FFT from the ADC connected to the APU. This way we'll make sure that the APU has a typical noise floor, like other APUs(my screenshot below is without a calibration file). Also, it's a good idea to use 48k or 96k of sampling than 384k.

Also, you can check the noise performance of d10. Use Scaler with 26db gain and -60dbfs DAC's output to read THD+N A-weighted.

I think this is what you want. APU notch at 1 KHz, 6 dB of APU gain, shorted APU input at the connector. 1.7 Vrms sensitivity for the ADC. 96 KHz sample rate, no calibration file. FS sine Vrms for REW measurement: 1.7 Vrms.

~~~~~

This is the D10s to the Scaler to the ADC. -60 dBFS out of D10s. 26 dB gain for the Scaler. 1.7 Vrms sensitivity for the ADC. 96 KHz sampling rate. FS sine Vrms for REW measurement: 1.7 Vrms.

Just for comparison, I made the same measurement for the D10s with 384 KHz sampling rate. FS sine Vrms for REW measurement: 1.7 Vrms.

 

[IVX]

Well, your THD+N isn't A-weighted but let's roughly add 2.5db, hence, your DR is 123db(A) i.e. 3db better than promised. As I noticed the d10s output is less than 2Vrms, that's why AP measurement shows -120db(A), these analyzers are too noisy and have -126-127db(A) its own noise at 2Vrms. log10(((10^(-123/20))^2+(10^(-126/20))^2)^.5)*20 = 121db(A) in the ideal case. I recommend amplifying d10 output with Scaler close to 7-10V(13.34db is a good choice to get 9Vrms after Scaler) and next ->APU(0db, don't need +6)->ADC 2.7-3.5V. If the result will be the best of you saw before, I'll explain why
BTW, did you set THD+N AES17-2015 filter option on REW??
PS: log10(((10^(-130/20))^2+(10^(-126/20))^2)^.5)*20 = 124db(A) This is what we can see when AP tries to measure CS43131 aka Meizu HiFi DAC and so on.

 

[BKDad]

Well, your THD+N isn't A-weighted but let's roughly add 2.5db, hence, your DR is 123db(A) i.e. 3db better than promised. As I noticed the d10s output is less than 2Vrms, that's why AP measurement shows -120db(A), these analyzers are too noisy and have -126-127db(A) its own noise at 2Vrms. log10(((10^(-123/20))^2+(10^(-126/20))^2)^.5)*20 = 121db(A) in the ideal case. I recommend amplifying d10 output with Scaler close to 7-10V(13.34db is a good choice to get 9Vrms after Scaler) and next ->APU(0db, don't need +6)->ADC 2.7-3.5V. If the result will be the best of you saw before, I'll explain why
BTW, did you set THD+N AES17-2015 filter option on REW??
PS: log10(((10^(-130/20))^2+(10^(-126/20))^2)^.5)*20 = 124db(A) This is what we can see when AP tries to measure CS43131 aka Meizu HiFi DAC and so on.

OK. These are with the D10s > Scaler > APU > Cosmos ADC. The Scaler was set to 13.38 dB of gain. The APU was set to 0 dB of gain, 1 KHz notch. DAC sensitivity set to 3.5 Vrms.

First, this is the D10s at 0 dBFS output. AES filter option in REW OFF. (Should it be on??) 96 KHz sampling rate.

Now, with the AES filter option ON. 96 KHz sampling rate.

Previously I made measurements with the D10s at -0.5 dBFS. This is that with the AES filter option OFF. 96 KHz sampling rate.

Same as above, but with AES filter option ON. (I messed up the on display dBFS late - it really is -0.5 dBFS)

Finally, for comparison, this is the D10s at -0.5 dBFS, AES filter option ON, and 384 KHz sampling rate.

So, which one is "right"???

 

[IVX]

-112.0db vs -112.2db and you asking which one is correct? Man, this is an audio measurement, we should be ready for a happy dance if our test2test results are so close as yours Actually, you can omit Scaler from the scheme and repeat your previous test but with AES filter, you'll see -112db again.

 

[BKDad]

-112.0db vs -112.2db and you asking which one is correct? Man, this is an audio measurement, we should be ready for a happy dance if our test2test results are so close as yours Actually, you can omit Scaler from the scheme and repeat your previous test but with AES filter, you'll see -112db again.

I was joking...

I didn't think about using the AES17-2015 filter option. I knew it was there and completely went stupid about using it with the APU. Thanks for the reminder.

The distortion levels of the individual harmonics change some between sample rates and all that. Look at the odd order products. Not sure which one is really right.

 

[IVX]

BKDad, now you can test the DAC's jitter to compare with that https://www.audiosciencereview.com/...topping-d10s-usb-dac-and-bridge-review.14859/

 

[BKDad]

BKDad, now you can test the DAC's jitter to compare with that https://www.audiosciencereview.com/...topping-d10s-usb-dac-and-bridge-review.14859/

You might like this. I tried to use the same settings as in the original review. Maybe I should've used a blue trace.

I guess that USB isolation might matter after all, even if you use an AP test system with all sorts of fancy interfaces that have great common mode signal rejection throughout the audio band. (No idea what it's like above the audio band.)

 

[IVX]

I like that because again and again cheap Cosmos devices overtake $30000+ monsters(if you know how to do that)

 

[trungdtmc]

My cheap thermostat )
With a 5v/180* PTC heater and ne555 delay turn off board. I set timer to take es9822 to above 75*C it will turn off heater, after cool down, the ADC could reach to grade A number

 

[BKDad]

So, when can we place our orders for Cosmos DAC?

Cosmos ADC performance

Also...

Is the performance of the new ADCiso any better than the original ADC, assuming the same grade and using an external USB isolator?

 

[IVX]

ADCiso has about 3db less H3 at 10kHz(-123db@-1dbfs).

 

[BKDad]

OK, a last sanity check.

This is a "Victor" oscillator running through a Cosmos Scaler set to zero dB of gain to the Cosmos ADC. 384 KHz sampling rate. It's a classic ADC, so the USB isolator is external.

Now, I've added a Cosmos APU at 0 dB of gain between the Scaler and the ADC. Otherwise, the same.

I think this effectively shows the limits of the Scaler plus ADC combination for distortion. I made no attempt to optimize the signal levels for noise.

BTW, when running directly into the lower input impedance of the APU instead of through the Scaler, the 3rd harmonic from the Victor oscillator increases to about -135 dBc. This version of the Victor oscillator is maybe 10 years old - I understand that the newer ones are better.

 

[Grooved]

OK, a last sanity check.

This is a "Victor" oscillator running through a Cosmos Scaler set to zero dB of gain to the Cosmos ADC. 384 KHz sampling rate. It's a classic ADC, so the USB isolator is external.

View attachment 317063

Now, I've added a Cosmos APU at 0 dB of gain between the Scaler and the ADC. Otherwise, the same.

View attachment 317065

I think this effectively shows the limits of the Scaler plus ADC combination for distortion. I made no attempt to optimize the signal levels for noise.

BTW, when running directly into the lower input impedance of the APU instead of through the Scaler, the 3rd harmonic from the Victor oscillator increases to about -135 dBc. This version of the Victor oscillator is maybe 10 years old - I understand that the newer ones are better.

Hi, is it possible to bring the level up to -5dBFS in the second test, like in the first one?
Unless I've missed something, it would be better to compare to get the same level at the ADC input.

 

[BKDad]

Hi, is it possible to bring the level up to -5dBFS in the second test, like in the first one?
Unless I've missed something, it would be better to compare to get the same level at the ADC input.

Same level coming from the signal generator. I didn't touch either it or the fixed gain of the Scaler.

The lower level of the fundamental shows the effects of the APU's notch and gain structure into the ADC. There's a cal file used in REW to compensate for that in the display and the distortion calculations.

At least I hope I did that part correctly.

 

[IVX]

OK, a last sanity check.

This is a "Victor" oscillator running through a Cosmos Scaler set to zero dB of gain to the Cosmos ADC. 384 KHz sampling rate. It's a classic ADC, so the USB isolator is external.

View attachment 317063

Now, I've added a Cosmos APU at 0 dB of gain between the Scaler and the ADC. Otherwise, the same.

View attachment 317065

I think this effectively shows the limits of the Scaler plus ADC combination for distortion. I made no attempt to optimize the signal levels for noise.

BTW, when running directly into the lower input impedance of the APU instead of through the Scaler, the 3rd harmonic from the Victor oscillator increases to about -135 dBc. This version of the Victor oscillator is maybe 10 years old - I understand that the newer ones are better.

OMG, it seems you can find on the bottom of your ADC my mark "THD+", right? H3 level is extremely low, far not everyone Cosmos ADC has it!

 

[BKDad]

OMG, it seems you can find on the bottom of your ADC my mark "THD+", right? H3 level is extremely low, far not everyone Cosmos ADC has it!

I only found this one mark: