E1DA Cosmos APU - Teardown and personal thoughts and measurements

[pkane]

right, it happens due to the high impedance of the notch at the rejection frequency. But it's not a reason to worry about, APU's notch limit for THD+N is around -135db i.e. something like 3x times lower vs APx555.

This looks good to me, although noise appears to be close to that -135dB line and I can still see the tilt of the noise floor

 

[IVX]

I did explain my result here https://www.diyaudio.com/community/...perform-apx555b-for-30000.386001/post-7063707
In fact, I didn't see the THD+N <-132db@1kHz@10Vrms with my Cosmos DAC proto(DAC+LPF output).

 

[Michael1]

I'm in the process of gathering some test equipment.... I already have a QA403, but I was thinking about the possibility of adding a Cosmos APU so that I can accurately measure THD+N of very high performance DACs (e.g. ES9039). I was wondering if anyone uses a similar set up, or if there are any good resources on such a configuration?

 

[IVX]

ES9039(especially Q2M!) is also underrated in its datasheet, and for the same reason, AP is just a blind to see any performance difference. You can see 9039pro ASR/L7 results never go over -124-125db because it is AP's limitation when APU+pretty much any ADC shows THD+N@1k down to -131db. I won't say if anyone is able to hear such a difference but it is 2x times.

 

[Michael1]

ES9039(especially Q2M!) is also underrated in its datasheet, and for the same reason, AP is just a blind to see any performance difference. You can see 9039pro ASR/L7 results never go over -124-125db because it is AP's limitation when APU+pretty much any ADC shows THD+N@1k down to -131db. I won't say if anyone is able to hear such a difference but it is 2x times.

And there's no inherent reason why the APU and QA403 wouldn't work well together?

 

[IVX]

Oh, sorry, my bad, I got -130db with 9039Pro, not -131db yet.
APU is ok with any ADC, even with laptop built-in audio input.

 

[Rja4000]

I'm in the process of gathering some test equipment.... I already have a QA403, but I was thinking about the possibility of adding a Cosmos APU so that I can accurately measure THD+N of very high performance DACs (e.g. ES9039). I was wondering if anyone uses a similar set up, or if there are any good resources on such a configuration?

My measurements are using an APU.
As an example this or this.
As you may see, I reached -128dB THD+N.

(Currently, my best THD+N DAC is the D50 III, with outputs summed to mono. As it uses two 9039Q2M 2 channels chips, that's the equivalent of four 9039 channels summed. In theory, a 9039Pro, which has 8 channels, should be 3dB better when all channels are summed, so -131dB seems realistic.)

 

[AudioSceptic]

Oh, sorry, my bad, I got -130db with 9039Pro, not -131db yet.
APU is ok with any ADC, even with laptop built-in audio input.

Yeah: that 1 extra dB is all-important! ;-)

 

[IVX]

the truth is matters I mixed up the result I got with DAC+LPF. ES9039Q2Mx2 -> 1kHz LPF gave me THD+N -131db. Actually, it was better than -131db because the theoretical limit of Cosmos APU is -135db@10Vrms, that's too close to not consider that. If the considered value is caused clearly by the noise only, then ES9039Q2Mx2 -> 1kHz LPF has THD+N -133db.

 

[Rja4000]

I measured
EIN for E1DA APU 60dB preamp

RME ADI-2/4 Pro SE is used as ADC (Mono, 1dBu range)
Software is Virtins Multi Instrument 3.9.9.6
Voltage measured with Keithley DMM6500
BW 20Hz-20kHz - 28.4°C

Measured Gain is 59.47dB

EIN with 150 Ohm resistor

Preamp noise is around 180nVrms
(for 150nV announced, which would give -129.20dB EIN under the same conditions)


EIN with input shorted

Very similar result, with 178 nV rms noise.

In any case, this is excellent.


(The That 1510 chip preamp is advertised for 1 nV rms / SQRT(Hz) or 141 nV rms 20Hz-20kHz.
So, in therory, that should give us -129.35dBu EIN with 150 ohm load 20Hz-20kHz.
My measure is 0.65dB from that.)

 

[IVX]

I measured
EIN for E1DA APU 60dB preamp

RME ADI-2/4 Pro SE is used as ADC (Mono, 1dBu range)
Software is Virtins Multi Instrument 3.9.9.6
Voltage measured with Keithley DMM6500
BW 20Hz-20kHz - 28.4°C

Measured Gain is 59.47dB

EIN with 150 Ohm resistor

View attachment 383294

Preamp noise is around 180nVrms
(for 150nV announced, which would give -129.20dB EIN under the same conditions)


EIN with input shorted

View attachment 383298

Very similar result, with 178 nV rms noise.

In any case, this is excellent.


(The That 1510 chip preamp is advertised for 1 nV rms / SQRT(Hz) or 141 nV rms 20Hz-20kHz.
So, in therory, that should give us -129.35dBu EIN with 150 ohm load 20Hz-20kHz.
My measure is 0.65dB from that.)

If nobody uses the phantom power 48Vdc, 1/f noise is easy to improve by replacing 2 alcaps NP 10uF/50V for NP 47uF/10V or higher. I need to measure some voltage regulator noise, and I replaced 10uF with 220uF, check this out.

 

[trl]

@IVX, which caps, please? The two from the top 10uF/50V?

If so, then what exact capacitance should we choose for the "upgrade": 47uF/50V or 220uF/50V (if will fit in there)?

Also, what happens with the 48V phantom power after performing this modification?

Thank you!

 

[IVX]

Well, I believe you can't fit even NP(Non-Polar aluminum cap) 47uF/50V over there. NP 47uF/10V probably is ok.

 

[IVX]

And one more practical trick, to prove a VERY low distortion results I use 3 long(to minimize VCR you can use multiple pcs in serial) resistors in serial as a differential voltage divider. 1kOhm of each, the first and last are inputs, and the middle one is output Vin/3. In that case, APU's notch Zinput affects the distortions too much(0402 10kOhm has too much VCR). So, if you gonna do that proofing work, you better replace 0603 10k with 0402 100k resistors.

 

[johny_2000]

Well, I believe you can't fit even NP(Non-Polar aluminum cap) 47uF/50V over there. NP 47uF/10V probably is ok.

And the other two 10uF 50V caps (C37 and C48) for the notch filter inputs?
Can they be replaced with higher value capacitors?

 

[IVX]

sure but APU notch section is intended to measure 1k or 10k distortions and 10-20Hz distortions don't matter.