E1DA Cosmos ADC

[IVX]

don't forget to warm up the ADC, even with a hairdryer, ES9822 has a minimal H3 at about 50C.

 

[BKDad]

don't forget to warm up the ADC, even with a hairdryer, ES9822 has a minimal H3 at about 50C.

I have your temperature control board glued to the ADC chip, so I think I'm OK after a half hour warmup. But, that's a good point.

I wonder if the Topping DACs actually may be hurting their performance by the way they are tweaked.

This is why we need a Cosmos DAC for measurements...

 

[IVX]

I thinking about it last week..

 

[BKDad]

I thinking about it last week..

To complete the set, a fixed or variable LPF or BPF for the DAC output to give an almost distortion-free test signal would be great, too. Could be a separate box, like the APU is.

 

[Mike123]

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.

It is not always so. I have very average audio system with inexpensive china DAC, DIY D class amplifier and 3-way Speakers 150 Watt power and 91 dB sensitivity. I am 56 years old. But i can hear the 3 kHz 96/24 signal, which I turn on/off using the play/pause button on the player. From a listening position of 2m with an attenuation of -105dBFS. And of 1 cm from the speaker with a level of -128 dBFS. Although, of course, this has no effect on my emotions from listening to old noisy music .

 

[BKDad]

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!

I spent some more time replicating your results. I think I've got the same settings:

-0.5 dBFS from D10s
384 KHz sampling rate
22000 Hz calculation window for REW, set with high and low pass filters
256K FFT Length
8 Averages
Blackman-Harris 7 Window
87.5% Max Overlap

So, here's the results. First is 4.5 Vrms ADC sensitivity:

This is 3.5 Vrms ADC sensitivity:

Finally, this is 2.7 Vrms ADC sensitivity:

So, which one is right???

 

[BKDad]

So, which one is right?

I think this one shows the correct D10s performance:

This the D10s through the Cosmos APU to the Cosmos ADC, with the Cosmos AC sensitivity set to 2.7 Vrms. Since the fundamental is notched, the ADC shouldn't be adding much, if any, distortion itself.

All this distortion cancellation stuff can be tricky at these low levels.

BTW, no matter what, the E50 DAC still has worse 3rd harmonic performance.

Edit: My results from the D10s now are pretty close to the review measurements presented here at ASR as well as those at L7 Audio Lab. But, both those guys got far better distortion results with their E50 samples. Either my E50 is not representative of normal production, or theirs aren't. No way to tell.

 

[IVX]

When you go with Scaler, better to adjust levels so that ADC signal will be close to 0dbfs, for instance, ADC set to 10V of the sensitivity, and Scaler +7db. Of course, it is better to use ADC in the mono modem Scaler will provide it if you plug only the left channel.
If go with APU for THD+N, and the DUT has <5V output, better to set +6db gain of the notch.

 

[staticV3]

I think this one shows the correct D10s performance:

I would repeat the measurement with A: the APU's notch filter set to +6dB and B: the ADC set to 1.7V

With the D10s set to 0dBFS, that should give you an amplitude of roughly -3.16dBFS in REW.

The closer you can get the input to -.5dBFS, the further away the ADC's noise floor is from your point of interest, the more accurate your measurements will be.

 

[BKDad]

With a 20 dB null from the APU, isn't it true that the ADC will pretty much not add any distortion? That's why I say that the distortion measurement with the APU is most accurate. The differences are admittedly small, though. I just wanted to see the results and post them. They are pretty close to what ASR and L7 have published, as well as what PMA showed in his own thread. Certainly within device variations. Good enough for me, and I suspect most people.

For noise measurement, yeah you need to get levels just right to minimize the effects of the test equipment internal noise. I'm not so concerned about that.

I still can't rationalize the bad E50 results, though. I wonder if Topping adjusts the ESS registers by looking at the distortion through an AP test system. The AP is really, really good but the ADC in it isn't distortion free. Unless they use a tracking filter before the ADC, it will distort. So, the ESS registers could be cancelling the AP distortion for the 3rd harmonic. So, when I test mine in a different system with different distortion characteristics, this might be what you get. It could even be worse than no ESS correction at all. It just seems odd that both Amir and L7 have gotten similar results, presumably with different samples of the E50. Either they got custom versions, which is possible, or the common denominator is their test gear. Even if they got custom versions, the results are too similar to be coincidence. If Topping just used the same register coefficients for customer production, you'd think that I'd get similar results when using the APU, which pretty much eliminates ADC distortion. Right? When you try to cancel distortions, think of how close the "pre-distortion" has to be to get cancellation at this level.

 

[MC_RME]

There is a difference between ESS and AKM. If you lower the input level on an AKM ADC to -20 dBFS, all harmonics are gone. If you do this on an ESS, harmonics and spurious tones stay, and you need to go down to at least -60 dB to get rid of them. The ESS advantage is that it works at highest level and therefore can measure THD+N very good, while AKM will add THD then - or add noise at lowered level.

The Audio Precision AD section is absolutely perfectly designed, uses a 100 dB notch filter, and can show THD (not THD+N) below -150 dB (via FFT, not as value), no matter the input level - if you ever find a source that is that good.

 

[IVX]

nothing is perfectly designed but big bang. AP's variable notch is full of unavoidable compromises. MDAC(actually 2 stages in models with -130db notch, APx555b is better because doesn't notch more than -25db) limits harmonic distortion performance worse than -150db(and very level dependant, if you need to see more realistic H2 better to go overvoltage, if you looking for the realistic H3 undervoltage).

 

[AnalogSteph]

There is a difference between ESS and AKM. If you lower the input level on an AKM ADC to -20 dBFS, all harmonics are gone. If you do this on an ESS, harmonics and spurious tones stay, and you need to go down to at least -60 dB to get rid of them.

Although it is possible to make them go away with DSD output, so I reckon a bit of shaped dither may be worth a shot...

 

[MC_RME]

Although it is possible to make them go away with DSD output, so I reckon a bit of shaped dither may be worth a shot...

I don't see how this relates to using an ADC. Or how to add shaped dither to a DSD signal.

 

[Music1969]

I thinking about it last week..

Cosmos DAC.

Then little monoblock Cosmos AMP Maybe 50W ?

 

[arvidb]

Is there a schematic of the frontend of Cosmos ADC somewhere, or some other way to see what kind of input protection there is, if any? I'm wondering how careful I have to be to avoid overvoltage. I don't want to break this thing!

Perhaps it would make sense to add an absolute maximum input current spec on the homepage?

 

[AnalogSteph]

I don't see how this relates to using an ADC. Or how to add shaped dither to a DSD signal.

Well, yes, I was thinking more of the DAC side... what you were describing is nothing less than the venerable "ESS hump". Here's an example:

PCM vs DSD

Did a little comparison and I was wondering if any friend with better gear here has tested this. Results you're about to see is probably narrowed to the specific DAC but it's interesting to see how it handles things differently. I will not try to decipher the results other that DSD seems much...

www.audiosciencereview.com

DSD (a bunch of shaped dither) clearly reduces the "fuzz".

Now that being said, I don't see why much the same shouldn't apply to their ADCs as well... one would basically have to pull a PCM1630 and explicitly generate some random noise (Sony literally employed a digital pseudorandom noise generator, probably for level consistency), preferably of the shaped variety (N=2-ish?) so in-band SNR does not suffer too much. It definitely is something worth playing with for sure, and in fact you may already have all the equipment at hand.

 

[arvidb]

Is there a schematic of the frontend of Cosmos ADC somewhere, or some other way to see what kind of input protection there is, if any? I'm wondering how careful I have to be to avoid overvoltage. I don't want to break this thing!

Doing some more research this is what I have been able to find:

Cosmos ADC is able to withstand 10 Vrms even at 1.7 V gain setting.
Ivan did a test with 20 Vrms 1 kHz sine wave applied for 10 s at a gain setting of 10 V, which the Cosmos ADC survived.

So that's good to know, that it doesn't seem to be super-sensitive to over voltage at least.

 

[BKDad]

I would repeat the measurement with A: the APU's notch filter set to +6dB and B: the ADC set to 1.7V

With the D10s set to 0dBFS, that should give you an amplitude of roughly -3.16dBFS in REW.

The closer you can get the input to -.5dBFS, the further away the ADC's noise floor is from your point of interest, the more accurate your measurements will be.

OK. Here is the same Topping D10s DAC into the same APU into the same Cosmos ADC as before. The APU output is applied to both the left and right inputs of the ADC and the ADC run in mono mode. The system has been warmed up for a half hour. (Reminder: The ADC has the optional thermostat board.) 384 KHz sampling rate,

The APU gain has been set to +6 dB.

First, the ADC sensitivity at 3.5 Vrms:

Now, ADC sensitivity at 1.7 Vrms:

I can't match arvidb's distortion nor can I get to Amir or L7's noise level. Bad karma, probably.

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

 

[staticV3]

Now, ADC sensitivity at 1.7 Vrms:

I can't match arvidb's distortion nor can I get to Amir or L7's noise level. Bad karma, probably.

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.