E1DA Cosmos ADCiso Review

[restorer-john]

measured with LCR Reader MPA (a truly terrific device, despite the cheap looks).

Very cool device, but unfortunately no good for me. I'm left handed and all "probe" type meters put the display upside down and facing my palm.

 

[Roland301]

But SINAD/THD+N is not that useful for comparison as we want to know the individual contributions of noise and harmonics. A those distortion levels noise basically always dominates.
Could you kindly post 1kHz spectra with and without HPSA so we could compare to Amir's plot from the first post?

Here you go. Note that I have a 4Vrms signal referenced against dBV here so the vertical offset is different from Amir's. I think this is with no averaging.
The upper harmonics and "grass" change a bit at different sample rates. Mostly the 2nd, 3rd, and 5th stay consistent.

 

[CleanSound]

.You do need to understand its limitations if you are going to use it for measurements so please don't jump in with both feet without learning about that. An amplifier for example can easily blow up its input stage without a scaler to bring its level down (on top of needing a dummy load).

There is a 48v, 2.5mm, 4 pole trrs input you can use for power amps. But still need to proceed with caution.

 

[Roland301]

There is a 48v, 2.5mm, 4 pole trrs input you can use for power amps. But still need to proceed with caution.

"48v" and "power amps" sounds funny next to "2.5mm 4 pole trrs" in a sentence
Seems like a good way to short an output with an accidental cable yank.
Feels like having switches to toggle the XLR over to that circuit would be much safer...

 

[KSTR]

Very cool device, but unfortunately no good for me. I'm left handed and all "probe" type meters put the display upside down and facing my palm.

This one is different. It has a G-sensor and switches display automatically the correct way.

 

[KSTR]

Here you go. Note that I have a 4Vrms signal referenced against dBV here so the vertical offset is different from Amir's. I think this is with no averaging.
The upper harmonics and "grass" change a bit at different sample rates. Mostly the 2nd, 3rd, and 5th stay consistent.

Thank you so much for the follow up.

Signal levels are comparable (-1.5dBFS for Cosmos ADC, -1.9dB for APx555). I see about -130dBc for the AP and about -134dBc for the Cosmos, eyeballing the highest component. Cosmos has very low harmonics with the 3rd dominating, APx has dominating 5th (unless that's a spurious needle) and more equal levels for the other harmonics.
Noise (-117 vs -116dB) is also very comparable (and this is a B-grade unit, and not in mono mode for the R-channel). SINAD is very comparable, therefore.

Cosmos in B-grade is definitely on par with the AP in this comparison and Ivan has done an excellent job (the low HD2 and HD3 tell his distortion compensation calibration of each individual unit is dialed in pretty well, and maybe it could even be bettered with an APx in HPSA mode as the monitor).

In HPSA mode, APx obviously improves both in distortion as in noise floor... we would have to compare that with the APU(notch) + Scaler(Autoranger) + Cosmos(ADC) combo.
But a standalone 0-Grade Cosmos in Mono (or inverted Mono) mode and with compensation dialed in with an APx would really come close already or a bit better wrt noise floor.

 

[restorer-john]

This one is different. It has a G-sensor and switches display automatically the correct way.

That's awesome! I may have to put one on my wish list. Thanks for the heads up.

 

[capslock]

Thank you so much for the follow up.

Signal levels are comparable (-1.5dBFS for Cosmos ADC, -1.9dB for APx555). I see about -130dBc for the AP and about -134dBc for the Cosmos, eyeballing the highest component. Cosmos has very low harmonics with the 3rd dominating, APx has dominating 5th (unless that's a spurious needle) and more equal levels for the other harmonics.
Noise (-117 vs -116dB) is also very comparable (and this is a B-grade unit, and not in mono mode for the R-channel). SINAD is very comparable, therefore.

Cosmos in B-grade is definitely on par with the AP in this comparison and Ivan has done an excellent job (the low HD2 and HD3 tell his distortion compensation calibration of each individual unit is dialed in pretty well, and maybe it could even be bettered with an APx in HPSA mode as the monitor).

In HPSA mode, APx obviously improves both in distortion as in noise floor... we would have to compare that with the APU(notch) + Scaler(Autoranger) + Cosmos(ADC) combo.
But a standalone 0-Grade Cosmos in Mono (or inverted Mono) mode and with compensation dialed in with an APx would really come close already or a bit better wrt noise floor.

Concur. -- what is inverted Mono mode?

 

[KSTR]

what is inverted Mono mode?

You wire the L and R inputs in opposite polarity and do a (L-R)/2 calculation in the software of the analyzer (REW has an option for that).
Even order distortion products tend to cancel out this way, as do internal interferences common to both channels. Uncorrelated noise is 3dB down just as in simple mono averaging.

 

[capslock]

Cosmos ADC Input Impedance, AC 1kHz, measured with LCR Reader MPA (a truly terrific device, despite the cheap looks).
View attachment 367252
Perfect match of L and R channels as expected (maybe @pma's unit has issues with the micro switches?) and in line with manufacturer's specification.
The grey values are for completeness only.

Awesome! Why is SE not exactly half of balanced but rather like 2/3?

 

[KSTR]

Why is SE not exactly half of balanced but rather like 2/3?

Because the input resistance is not just simple resistors to GND, rather we have a specialized circuit here, a fully differential OpAmp and that's how these circuits behave, it depends on the gain and the feedback resistors. With gains approaching zero we get to the simple model of half of differential impedance, whereas at infinite gain the impedance is the same, balanced or SE.
With equal resistors R (unity gain)., the balanced impedance is 2*R and the single-ended impedance is 1.33*R.

 

[daniboun]

I share with you additional measurements of the Cosmos ADC via APx555.

Thanks @docjordan with whom I collaborated for my APX555 test of my Bamboo V2 project too

Cosmos E1DA APx555 measurements

 

[nagster]

apx555 and upv were measured. I wanted to do cross measurements, but I gave up because of the risk and effort of moving/restoring them.

 

[KSTR]

I share with you additional measurements of the Cosmos ADC via APx555.

Thanks @docjordan with whom I collaborated for my APX555 test of my Bamboo V2 project too

Cosmos E1DA APx555 measurements

Thanks for this.

But, did you check the unit is not running mono averaging mode on the left output channel?
This a common pitfall when using this device, notably under Windows. One needs to set volume control in TweakCosmos App to zero after the stream has been started. Windows Volume slider only works for this when the Comtrue Driver has been installed.

The 2dB lower noise on the left channel might indicate this, but of course it could be just regular inter-channel tolerances.
FWIW, with my unit the right channel is the quieter one, by 1dB, and 22Hz...22kHz unweighted noise is -122dBFS, 1.7Vrms input range setting, inputs open (as I found the noise floor doesn't change when input is shorted with a shorting XLR (Pin1,2 and 2 connected together).

 

[weme]

(LCR Reader MPA)
That's awesome! I may have to put one on my wish list. Thanks for the heads up.

If the LCR doesn't have to look so fancy, a slightly lower accuracy is sufficient and it can also cost less, the old DER EE DE-5000 could be an alternative. The TL-21/TL-22 measuring units are both built in four-wire technique. However, the measuring tips of the tweezers should be hard gold plated. (I haven't done this yet).

 

[HpW]

HpW, that looks good but is rather cryptic. Care to share more?

It is the work since several years, as looking for any better result's and different figures as as currently seen / measured by current system(s).
As it is a closed HW/SW project even no NDA's given.

 

[Roland301]

@ Roland301: what happens with the harmonics when you go down -1 dB? My experience with AKM has been that they do best at -6 to -15 dB. What level works bests depends on the ADC and the front end.

@ KSTR: wow, they must have bought a stash before they went EOL.

Ah missed your ping.. The AP's input is auto-ranging, so it's tricky. The input sensitivity was reported as 5Vrms with the 4Vrms signal. According to @nagster 's distortion sweep here it looks like you can eke out a little more performance with lower level.

 

[Roland301]

Actually, I went back and took more detailed FFTs of APX555B ADC mode at 4Vrms , 3.15Vrms (-1dB), and several common input sample rates to show what the harmonics and noise floor do. FFT setting is 32k AP-equiripple with 10 averages to clean it up.

 

[capslock]

Thanks. So at fs and -1 dB, it is a tad better than my RTX6001 which uses the same converter. Maybe that is the selected chips. How about at -6 or - 10 dB?

 

[nick_l44.1]

I share with you additional measurements of the Cosmos ADC via APx555.

Interesting measurement. For a complete picture, only the FFT of the AP generator is missing, especially its noise level.
I have a Cosmos ADC B DR 127.3 dB(A)@mono and tried to find out its real THD+N. The signal taken from E1DA 9039S has the following FFT:

THD+N measurement:


As you can see, the source is not ideal. Its noise (-126.2 dB) is very close to the result (-122.1 dB). To obtain such noise as a result, the ADC's own noise during measurement must be -124.24 dB. Adding this value to the measured THD (-129.9 dB) we get -123.2 dB. And also need to compensate for the discrepancy between the signal voltage and the 0 dBFS voltage of the ADC (-0.5 dB) since Cosmos does not have an autoranger. In total, I estimate the THD+N of this ADC to be -123.7 dB. This value cannot be much worse, otherwise, by reverse logic, it will be impossible to obtain THD+N -121 dB as a result of my measurement.