• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

E1DA Cosmos ADC

half_dog

Senior Member
Joined
Jun 28, 2018
Messages
450
Likes
358
Location
Brazil
BKDad, the problem is the uncalibrated DAC unit, not a big deal to reduce H3 for a single unit and send that to review. If the manufacturer did promise to calibrate each unit you have reason to be unhappy because you paid for the calibration extra money. If not, all is fine ;)
Maybe could it be related to the band width? He's using 20 to 45600Hz. Amir limits to 22400Hz
 

half_dog

Senior Member
Joined
Jun 28, 2018
Messages
450
Likes
358
Location
Brazil
Yep, yep. But I mean the noise level contribution. Although I agree this H3 is very high. Using a L30+E50 in to a ten-year-old TASCAM US-366 I can hit lower values of THD+N (10 to 24000Hz). One day I'll buy my Cosmos ADC :)
 

BKDad

Member
Joined
Jul 3, 2023
Messages
83
Likes
34
Maybe could it be related to the band width? He's using 20 to 45600Hz. Amir limits to 22400Hz

That certainly affects the integrated noise power, and adds some distortion contribution from those high order harmonics. But, those high order harmonics are really low, so their addition is minimal. It's the 3rd that is the limit, as Ivan said.

I'd guess at that level it could just be a question of unit-to-unit variation. Or, maybe the officially tested unit was custom tweaked. The sample I have meets the specs as best I can tell, so I have no reason to complain. It does what Topping says it will do.

I also did not fiddle with the DAC output level to get optimum performance. It was run at 0 dBFS (2.02 Vrms). Nor did I completely optimize the level into the ADC. With the Scaler at 0 dB gain, the level into the ADC was low compared to the 2.7 Vrms full scale input range. That probably would improve the overall measured performance by a bit. And, maybe, the DAC might work better at a lower sample rate of 44.1 KHz, which is where Amir runs DACs for most tests.

My point was to show how well the Scaler-ADC combination works. So, maybe this is what it should be compared to. This is the APU-ADC combination using a compensated notch at 1 KHz. It shows how little distortion is added by the Scaler.

E50-APU.png


Or, this is a QA480 test oscillator into the same APU-ADC combination:

QA480-APU.png


Now, this is the same QA480 into the Scaler-ADC combo:
QA480-Scaler.png


Definitely, the ADC is limited. But, it still does what's claimed.

I also did a bunch of Multitone 32 tests, both flat (white) and pink weighted to simulate what actual music is like. The test results were excellent there as well. More than good enough for developing preamps and power amps. At some point you also have the consideration of how much harmonic cancellation you get when cascading all those devices, especially the even order products.

All in all, it's close enough to an AP test system for my hobbyist use. I'm not fooling myself with any measurements I make. Worrying about 1 KHz harmonics below -120 dB isn't a concern. There's lots more to consider.

So, all really is fine. ;)
 
Last edited:

IVX

Major Contributor
Joined
Mar 3, 2019
Messages
1,393
Likes
2,745
Location
South of China, SHZ area, - Слава Україні
Actually, in our discord channel, we had a long discussion with the user who wasn't able to get such a clean FFT floor with Scaler->ADC combo. He had plan to bring that discussion here, I'll remind him. I did try to solve his issue, and he got some improvements but still 50Hz odd harmonics(hum) look annoying.
 

misureaudio

Member
Joined
Jul 29, 2019
Messages
88
Likes
104
Location
Mantova, Italy
I'm testing a Topping D90MQA using Cosmos ADC and Multitone Analyzer software. Balanced cables connect both channels. Multitone Test goes ok, but I noticed a 6dB difference between the R and L channels (R > L). Results are the same on a different laptop. I tried to make a recording of a 1 kHz sine at identical levels on L/R chans using Audacity and the recorded level is identical. Using Multitone Analyzer instead produces the 6 dB difference in level between the chans. I dont have tested each polarity in the balanced input, but could it be suspectable that a polarity on the L chan is dead? I have no deep knowledge of the details about Multitone Analyzer, When I'm doing a test, Multitone Analyzer calls the ASIO bridge driver suggested by Archimago's review. When I try the recording of the 1 kHz sine, Audacity calls DS or WASAPI subsystem, ASIO is not available. However, with Multitone Analyzer even trying the DS Subsystem doesn'change the result, the 6dB difference is there.

Thank you in advance for any idea.
 

Attachments

  • D90MQA_MT32AP-4M_32VG_96kHz_R.jpg
    D90MQA_MT32AP-4M_32VG_96kHz_R.jpg
    101.4 KB · Views: 44
  • D90MQA_MT32AP-4M_32VG_96kHz_L.jpg
    D90MQA_MT32AP-4M_32VG_96kHz_L.jpg
    101.7 KB · Views: 45
Last edited:

misureaudio

Member
Joined
Jul 29, 2019
Messages
88
Likes
104
Location
Mantova, Italy
misureaudio, did you read this? https://e1dashz.wixsite.com/index/cosmos-adc
Regarding MONO/STEREO modes.

"The MONO mode improves the SNR/DR for 3db, so a grade-A unit typically has SNR 125db(A) in STEREO mode. MONO/STEREO modes are switchable by the UAC2 volume commands, for example, if your Windows volume slider position is less than 50%(0 to 49%) the device works as 2 channel ADC, if more than 50% it is a MONO ADC with summed Left+Right inputs mapped on the Left channel(Right input is always mapped to the Right channel). In the MONO mode, both channels need to be tied together L+ to R+, L- to R-."

Thank You very much.
 

arvidb

Member
Joined
Dec 1, 2022
Messages
96
Likes
114
Location
Sweden
Well, I've ordered a Cosmos ADC Grade B + the Hifime isolator now, so in a week or so we'll know if it works or not! Cost (including shipping for the isolator and 25 % VAT): €305, compared to €343 for the Cosmos ADCiso (Grade A?) only. (From Audiophonics.)
I can confirm that Cosmos ADC works with the Hifime isolator (under Linux, connected to a USB 3.something port). Running at 384 kHz here:

REW-Topping-D10s-E1DA-Cosmos-ADC-1.png


(I know these settings & levels aren't optimal for best THD+N; it's just a first sanity check that the devices work in loopback and with the isolator. Also, I need to read up on FFT settings again. :) )
 

Rja4000

Major Contributor
Forum Donor
Joined
May 31, 2019
Messages
2,660
Likes
4,198
Location
Liège, Belgium
I can confirm that Cosmos ADC works with the Hifime isolator (under Linux, connected to a USB 3.something port). Running at 384 kHz here:

View attachment 311991

(I know these settings & levels aren't optimal for best THD+N; it's just a first sanity check that the devices work in loopback and with the isolator. Also, I need to read up on FFT settings again. :) )
Looks like the frequency gets out of sync.
Your low frequencies are way too high compared to high frequencies, and the "foot" of your fundamental is way too large.

So, first, you should select Window Blackman-Harris 7.
Let's hope it will look better.

Also, your level are not matching.
I'd run it at 0dBFS on the Topping and change sensitivity of The ADC to 2.7V.

If possible, use an Y cable to feed both ADC inputs with the same signal and run the ADC in mono mode.

Your figure should then be closer to the Topping capabilities.
 

arvidb

Member
Joined
Dec 1, 2022
Messages
96
Likes
114
Location
Sweden
Yep, switching to Blackman-Harris 7 certainly improves things! (I should have remembered that Rectangular window only works when the input and output devices have a common clock. Cool that REW even detects the clock delta here!) Thanks!

REW-Topping-D10s-E1DA-Cosmos-ADC-2.png
 

Rja4000

Major Contributor
Forum Donor
Joined
May 31, 2019
Messages
2,660
Likes
4,198
Location
Liège, Belgium
Yep, switching to Blackman-Harris 7 certainly improves things! (I should have remembered that Rectangular window only works when the input and output devices have a common clock. Cool that REW even detects the clock delta here!) Thanks!

View attachment 312005
Good
Now your result is ADC-noise-limited.
If you increase the sensitivity of the ADC, you'll gain 4dB or so.
In mono mode, you'll gain another dB or more (max 3dB).
So you should be able to get or approach Amir's 113dB SINAD figure. Or whatever value is the actual SINAD of the Topping at 384kHz.

(I know that was not your point. Obviously, the isolator works ;) )
 
Last edited:

BKDad

Member
Joined
Jul 3, 2023
Messages
83
Likes
34
Yep, switching to Blackman-Harris 7 certainly improves things! (I should have remembered that Rectangular window only works when the input and output devices have a common clock. Cool that REW even detects the clock delta here!) Thanks!

That's really great performance. Better than I can get out of my D10s or E50! Are you running balanced mode?
 

arvidb

Member
Joined
Dec 1, 2022
Messages
96
Likes
114
Location
Sweden
It's the unbalanced D10s. Just normal RCA outputs. I use a proper "balanced" RCA->XLR cable though with the shield tied to inverted signal at the RCA connector.
 

BKDad

Member
Joined
Jul 3, 2023
Messages
83
Likes
34
It's the unbalanced D10s. Just normal RCA outputs. I use a proper "balanced" RCA->XLR cable though with the shield tied to inverted signal at the RCA connector.
Well, those are certainly excellent results! The distortion is better than what Amir measured.

I guess I'm jealous...
 

arvidb

Member
Joined
Dec 1, 2022
Messages
96
Likes
114
Location
Sweden
Cool. :) Have you tried running at 384 kHz? I seem to get 4-5 dB lower 2nd and 3rd harmonics at 384 kHz compared to 96 kHz. It doesn't affect total figures much though; the higher harmonics get higher in amplitude.
 

BKDad

Member
Joined
Jul 3, 2023
Messages
83
Likes
34
Cool. :) Have you tried running at 384 kHz? I seem to get 4-5 dB lower 2nd and 3rd harmonics at 384 kHz compared to 96 kHz. It doesn't affect total figures much though; the higher harmonics get higher in amplitude.
I will try duplicating your setup tomorrow. -0.5 dBFS, 4.5 Vrms sensitivity for the input pad, and both 96 and 384 KHz. Easy enough once it's all up and running.

The 3rd and 5th harmonics are the ones that stand out, to me.

Thanks for pushing me to try this.
 

BKDad

Member
Joined
Jul 3, 2023
Messages
83
Likes
34
Well, that certainly was interesting.

With -0.5 dBFS out of the D10s into the ADC set for 4.5 Vrms sensitivity at 384 KHz sampling rate, I get much better results than with 96 KHZ. Still not close to what you're getting, though. This is with a Scaler between the DAC and ADC set to 0 dB gain , BTW.

I get very similar results at 192 KHz sampling rate.

When I change the input attenuator on the ADC between 2.7 Vrms and 8.5 Vrms, the 3rd harmonic varies by over 6 dB. This is with no other changes.

I suspect that when all these converters using ESS chips tweak for best distortion performance, it's at a certain level out of the DAC and into the ADC. I'm not sure just where the distortion of the DAC is cancelling the distortion of the ADC. And how much of the cancellation actually affects the DAC itself and how much affects the analog stages following. Or, even the ADC. There are certain combinations that are better than others. That makes me question any results below maybe -120 dBc with any test system and certainly makes me wonder about ranking products in order of goodness based on tenths of a dB in SINAD performance.

But, for building your own gear on your own test bench and measuring the results, -120 or -130 dBc is really pretty great.

BTW, the E50 has about 10 dB better noise floor than the D10s, but no matter what I do, the 3rd harmonic out of both channels is not as good as with the D10s. It stays right at about -120 dBc. That's consistent across sample rates, ADC attenuator settings, and any DAC output level within a dB or two of dBFS. I'm certain that level can't be heard in the very best of audio systems by the most sensitive of ears, especially with actual music, but it sure can be measured.
 
Top Bottom