E1DA Cosmos APU - Teardown and personal thoughts and measurements

[Rja4000]

I also request somebody state in a few simple sentences what this does. It's not an auxiliary power unit, nor accelerated processing unit. Analog processing unit. A microphone amp, but more?

This is a measurement device, very specialized for 1kHz (or 10kHz) THD (Sinad) measurement.

Basically, the issue with measuring SINAD of a device (DAC, Analog,...) is the limitations of your ADC.
You need the ADC to have a performance at least 10dB better, for peace of mind, than the device under test (DUT).

And this is where the problem is: with current DAC or headphones amp performance, you can't find a pure ADC that has this level of performance. The RME ADI-2 Pro ADC, a market reference, remains around 117-119dB SINAD. The E1DA ADC, using a new ESS ADC chip, is a bit better.
But none of them can reliably measure a DAC/amp with over 120dB SINAD.
You can't say if what you measure is the DAC/amp or the ADC.

THD and SINAD are usually measured by comparing the noise+ distortion to the signal. The signal being at 1kHz and with a level near the maximum capacity of the DUT.

So, what the traditional analyzers do is to notch out the 1kHz (remove the 1kHz peak, using a very steep and narrow analog band-cut filter), and amplify what remains BEFORE the ADC, bringing the DUT's remaining noise and distortion within the ADC measurement capability range.
And then the software computes the real values.

This is what Amir's AudioPrecision analyzer does.
The APU does the same.
And, since it's using some more recent chips, and also since it's only doing this for 2 fixed frequencies, the APU actualy outperforms somehow the AP for a signal level around 1V-5V.


EDIT: To make it clear:
In my opinion, the APU is the most useful innovation for amateur audio measurements in the last 3 years.
Or, maybe, even since the release of the RME ADI-2 Pro.

 

[Rja4000]

As a complement, one may want to read the comments I made with this measurement, performed with the APU

Topping D90LE - SINAD measurement (with E1DA Cosmos APU Notch)

In my previous measurements, I was peaking around 122dB SINAD measurement, with a direct ADC connection Using E1DA Cosmos APU Notch, (and some lessons learned from last REW editions), I was finally able to achieve a more relevant SINAD measurement of the D90LE Measurement is done with Topping...

www.audiosciencereview.com

 

[IVX]

Rja4000, right but 1kHz and 10kHz frequencies are there, two notches. The LNA preamp of course ready for a full audio-range and wider.

 

[Rja4000]

The LNA preamp

Sure.
You provide the mono 34dB or 60dB low noise preamp as a bonus

Usefull for measurements but also allowing other uses.

 

[IVX]

no, it was the initial idea of notch + LNA, the bonus is phantom power Actually, I used the same but ugly proto-tool for my bench years before and finally made that more convenient(nice and compact housing, USB-C power).

 

[Rja4000]

the first in its kind device on the market

Not really.
But it ups the performance.

 

[AudioSceptic]

maybe, however, Cosmos ADC has a gain range 1.7-10Vrms so you have an additional 15db to adjust it 34db and 60db is compromising between 3 different tasks:
a) DACs, amps, LDOs noise measurement.
Example:
AES17 Dynamic Range test performs at -60dbfs level by measuring the THD+N A-weighted. That plot shows the residual THD+N (A) of the LNA at 1-10mV level sweep. 2mV corresponds 2V 0dbfs DAC like CS43131, and -82db needs to subtract 60db to get AES17 DR = 142db(BTW, 10x better than APx555b). A typical high-performance DAC with 5Vrms 0dbfs could be measured down to 150db and so on.
View attachment 232918


b) vinyl ripping with 34db of the gain for a MM cartridge, and 60db for the MC. Cosmos ADC has internal DSP where is implemented RIAA+subsonic filter.
Example:
Cosmos APU+Cosmos ADC+RIAA.
MM case, 5mV 10ohm source APU gain 34db Cosmos ADC 1.7V sensitivity, THD+N -89db or -97.7db(A). Due to H2 and H3 being very low, SNR is also 97db(A).View attachment 232916
MC case, 250uV 10ohm source APU gain 60db Cosmos ADC 1.7V sensitivity, THD+N -66db or -78db(A). Due to H2 and H3 being lower than the noise floor, SNR is also 78db(A).
View attachment 232917


c) microphones usage.
Example:
The gain of Cosmos APU preamp is +60db(1000x times), hence the scale units are nVrms(A). After the APU unit turns On, you can see 5s of settling,
from 6s the input noise reaches 130nV, about 9s 48V power was turned On, and after 1s reached 142nS. Hence, APU's phantom power adds just 12nV to the preamp noise, that's a nearly ideal result as I think The test was performed with a shorted preamp's inputs(In+ to In-) and biased to the GND by 2x6.8kOhm resistor as a dummy condenser mic.
View attachment 232915

I apologise for these stupid questions. The Cosmos ADC can apply RIAA for vinyl but needs the APU to provide enough gain (34 dB for MM, 60 dB for MC), so I could use the 2 together to feed into a DAC for playing instead of ripping? Did you consider adding an RIAA option to the APU so it could act as a phono preamp on its own, or would this not make any sense?

 

[IVX]

The Cosmos ADC can apply RIAA for vinyl but needs the APU to provide enough gain (34 dB for MM, 60 dB for MC), so I could use the 2 together to feed into a DAC for playing instead of ripping?

100% but please prepare a decent DAC for that.

 

[IVX]

Not really.
But it ups the performance.

any example?

 

[AudioSceptic]

Guys, this is tech stuff, you don't need that if have no plan to measure distortions, noise, or a dynamic range at finer levels than $30K Audio Precision analyzers do. Also, this is the tool(preamp) for ripping vinyl by the most "digital" approach(I may miss something but Cosmos ADC+APU combo is the unique offer on the market today), or this preamp may work as a good studio mic preamp with built-in high-quality 48VDC phantom power.
That's it.

But if the APU is mono, we would need 2 to rip vinyl in stereo?

 

[pma]

I use my openamp phono pre (open project)

Preamplifier for MM cartridge

with Cosmos ADC to rip vinyls. Excellent results and I believe the best way to do so.

 

[staticV3]

But if the APU is mono, we would need 2 to rip vinyl in stereo?

If you want to do it in one go, then yes. Alternatively you can record L and R separately and realign in post.

 

[AudioSceptic]

This is a measurement device, very specialized for 1kHz (or 10kHz) THD (Sinad) measurement.

Basically, the issue with measuring SINAD of a device (DAC, Analog,...) is the limitations of your ADC.
You need the ADC to have a performance at least 10dB better, for peace of mind, than the device under test (DUT).

And this is where the problem is: with current DAC or headphones amp performance, you can't find a pure ADC that has this level of performance. The RME ADI-2 Pro ADC, a market reference, remains around 117-119dB SINAD. The E1DA ADC, using a new ESS ADC chip, is a bit better.
But none of them can reliably measure a DAC/amp with over 120dB SINAD.
You can't say if what you measure is the DAC/amp or the ADC.

THD and SINAD are usually measured by comparing the noise+ distortion to the signal. The signal being at 1kHz and with a level near the maximum capacity of the DUT.

So, what the traditional analyzers do is to notch out the 1kHz (remove the 1kHz peak, using a very steep and narrow analog band-cut filter), and amplify what remains BEFORE the ADC, bringing the DUT's remaining noise and distortion within the ADC measurement capability range.
And then the software computes the real values.

This is was Amir's AudioPrecision analyzer does.
The APU does the same.
And, since it's using some more recent chips, and also since it's only doing this for 2 fixed frequencies, the APU actualy outperforms somehow the AP for a signal level around 1V-5V.


EDIT: To make it clear:
In my opinion, the APU is the most useful innovation for amateur audio measurements in the last 3 years.
Or, maybe, even since the release of the RME ADI-2 Pro.

And when doing a sweep, the Audio Precision has a sliding notch filter to match?

 

[pma]

If you want to do it in one go, then yes. Alternatively you can record L and R separately and realign in post.

From vinyl?? You will get plenty of random errors between 2 separate vinyl runs and rips. This suggestion is not good, from technical point of view.

 

[Rja4000]

And when doing a sweep, the Audio Precision has a sliding notch filter to match?

The AP, as far as I understood, is using various methods, according to the measurement you do.
As an example, it, of course, doesn't use an analog notch to perform multitone measurements... Only the ADC.

So I don't know for the "sweep" (what exactly do you refer by that ?),
but I expect that the THD vs frequency uses, indeed, a variable frequency notch filter.
I might be wrong though.

It also uses a mix of analog generator and DAC generator.

About AP, you may ask Amir or our fellow @SIY, who wrote a review of the APx555B in AudioXPress some time ago

 

[IVX]

From vinyl?? You will get plenty of random errors between 2 separate vinyl runs and rips. This suggestion is not good, from technical point of view.

I did try that with Quartz PLL turntable, at least at one track it is Ok. But, of course, I did that because I had only one single APU on my hands at those time.

 

[AudioSceptic]

If you want to do it in one go, then yes. Alternatively you can record L and R separately and realign in post.

I tried ripping vinyl, but it was so tedious that I decided to not bother again and just get the albums on CD (or just play the vinyl if there was no CD version). Doing L & R separately would multiply the tedium beyond the limits of my sanity.

 

[IVX]

And when doing a sweep, the Audio Precision has a sliding notch filter to match?

correct, for that reason AP uses a variable freq notch.

 

[AudioSceptic]

I use my openamp phono pre (open project)

Preamplifier for MM cartridge

with Cosmos ADC to rip vinyls. Excellent results and I believe the best way to do so.

Looks good. Is there an MC version?

 

[AudioSceptic]

The AP, as far as I understood, is using various methods, according to the measurement you do.
As an example, it, of course, doesn't use an analog notch to perform multitone measurements... Only the ADC.

So I don't know for the "sweep" (what exactly do you refer by that ?),
but I expect that the THD vs frequency uses, indeed, a variable frequency notch filter.
I might be wrong though.

It also uses a mix of analog generator and DAC generator.

About AP, you may ask Amir or our fellow @SIY, who wrote a review of the APx555B in AudioXPress some time ago

I meant a frequency sweep, such as the usual 20 Hz - 20 kHz.