E1DA Cosmos ADCiso Review

[capslock]

Wait. Both graphs are 9039S into Cosmos ADC (non iso)? If so, why do you get -144 dB HD3 in the first and -133 dB in the second graph? Die only difference I can see is that you went from -1 db fs to -0.5. If that is really the only difference, one should certainly not go that close to full scale.

 

[nick_l44.1]

Wait. Both graphs are 9039S into Cosmos ADC (non iso)? If so, why do you get -144 dB HD3 in the first and -133 dB in the second graph? Die only difference I can see is that you went from -1 db fs to -0.5. If that is really the only difference, one should certainly not go that close to full scale.

The first one is the precise measurement of 9039s using notch. In the second there is only dac and adc and nothing else. This difference is in the second measurement. This dac output is 3.42 Vrms and the adc 0 dbfs input voltage is 3.62 Vrms. 20*lg(3.42/3.62)=0.49 dB.

 

[PortalKeeper]

Don’t know if this on topic but thought I’d try asking anyway. Are there any dynamic mics that could be plugged into this in mono mode?

 

[nick_l44.1]

Don’t know if this on topic but thought I’d try asking anyway. Are there any dynamic mics that could be plugged into this in mono mode?

It's minimal input voltage is 1.7 Vrms. Perhaps you'll need a preamp. The "mono mode" means just that the same signal plugged into both inputs and averaged. It helps to increase DR by ~2.5 dB.

 

[capslock]

The first one is the precise measurement of 9039s using notch. In the second there is only dac and adc and nothing else. This difference is in the second measurement. This dac output is 3.42 Vrms and the adc 0 dbfs input voltage is 3.62 Vrms. 20*lg(3.42/3.62)=0.49 dB.

Ok thanks, didn't realize that. So the noise with the notch in place is considerably lower. Would this be signal dependent noise modulation, i.e. the ADC is noisier with a near-fs signal present? If the ADC displays in fact noise modulation, is your calculation still valid?

It might be worthwhile installing the alternative firmware on the 9039S that runs everything off a 24 MHz clock. According to Ivan's data, all harmonics go down to -150 dB or below.

 

[nick_l44.1]

Ok thanks, didn't realize that. So the noise with the notch in place is considerably lower. Would this be signal dependent noise modulation, i.e. the ADC is noisier with a near-fs signal present? If the ADC displays in fact noise modulation, is your calculation still valid?

It might be worthwhile installing the alternative firmware on the 9039S that runs everything off a 24 MHz clock. According to Ivan's data, all harmonics go down to -150 dB or below.

The notch filter allows you to significantly reduce the influence of noise and ADC distortion on the measurement result. This is true for any ADC. Of course, there are all kinds of noise present, but since the signal level is maximum in all cases, I believe that the measured noise levels correspond to the maximum and the cause of the noise does not matter.
The THD+N Cosmos ADC is mostly determined by noise, so a cleaner sine is not critical. Therefore, I did not take into account source distortion in the calculation. The 9039S does not require alternative firmware to switch MCLK. My sample shows THD -147.8 dB on one of the channels, but the noise increases slightly. This was measured on a Cosmos ADC + Cosmos APU (notch). Since this measurement is at the limit of the APU accuracy, it is performed at the expense of the noise measurement accuracy.

 

[morillon]

The notch filter allows you to significantly reduce the influence of noise and ADC distortion on the measurement result. This is true for any ADC. Of course, there are all kinds of noise present, but since the signal level is maximum in all cases, I believe that the measured noise levels correspond to the maximum and the cause of the noise does not matter.
The THD+N Cosmos ADC is mostly determined by noise, so a cleaner sine is not critical. Therefore, I did not take into account source distortion in the calculation. The 9039S does not require alternative firmware to switch MCLK. My sample shows THD -147.8 dB on one of the channels, but the noise increases slightly. This was measured on a Cosmos ADC + Cosmos APU (notch). Since this measurement is at the limit of the APU accuracy, it is performed at the expense of the noise measurement accuracy.
View attachment 367669

why average exp? (and fft 512k? That’s a lot...)
;-)

 

[capslock]

Wow, this is the first time I see this kind of harmonics performance of the 9038S in 24 MHz mode confirmed. Who needs a high performance analog generator any longer? Was this 9038S selected for low THD?

"The THD+N Cosmos ADC is mostly determined by noise, so a cleaner sine is not critical." What is your reason or use case for chasing THD+N over THD?

 

[nick_l44.1]

why average exp? (and fft 512k? That’s a lot...)
;-)

To keep the noise floor lower. Otherwise, noise hides harmonics.

Wow, this is the first time I see this kind of harmonics performance of the 9038S in 24 MHz mode confirmed. Who needs a high performance analog generator any longer? Was this 9038S selected for low THD?

"The THD+N Cosmos ADC is mostly determined by noise, so a cleaner sine is not critical." What is your reason or use case for chasing THD+N over THD?

It's not 9038s but 9039s. My sample have one of the lowerest distortions but not the best by noise.
THD+N is being discussed here, so I'm writing about it.

 

[daniboun]

Another good review of the Cosmos ADC

EARLY LOOK: E1DA Cosmos ADC - affordable high performance measurements for the audio hobbyist!

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

 

[morillon]

To keep the noise floor lower. Otherwise, noise hides harmonics.

It's not 9038s but 9039s. My sample have one of the lowerest distortions but not the best by noise.
THD+N is being discussed here, so I'm writing about it.

certainly....but in case of sharing it is certainly better to use more common bases..(?)
. ;-)

 

[nick_l44.1]

certainly....but in case of sharing it is certainly better to use more common bases..(?)
. ;-)

Why? It will just be a less informative image with exactly the same results. The noise floor position depends on the FFT length and has no effect on the measurements.

 

[pma]

Correct. AP is not very good in this regard and also not in averaging possibilities. REW is better in this regard and allows for better visual resolution.

 

[morillon]

Why? It will just be a less informative image with exactly the same results. The noise floor position depends on the FFT length and has no effect on the measurements.

I clearly understood the technical interest in making it more relevant-obvious..
just, for people who are a bit novice and tempted to compare to other measures presented (like those of Armim etc.), can be clarified that they are not completely done on the same bases (fft , average chesby200vs hann etc)...
(the explanation of your method approach is precisely interesting)
just my clumsy statement
;-)

 

[KSTR]

At any rate, this is an excellent ADC. Very low noise (about 10nV/rtHz @ 1.7Vrms range) right up to 100kHz, and no sight of any spurious frequencies from all the digital stuff and power supply, even down at -170dBFS. The excellent brick-wall filter (one choice among several filters) working up to 384kHz and with basically no passband ripple rates is another bonus. I don't have the iso version but run mine bus-powered via a Topping HS01 USB Isolator, works fine so far.
However, it's a bit sort of a raw diamond, but a really sharp one at that.

 

[BKDad]

The notch filter allows you to significantly reduce the influence of noise and ADC distortion on the measurement result. This is true for any ADC. Of course, there are all kinds of noise present, but since the signal level is maximum in all cases, I believe that the measured noise levels correspond to the maximum and the cause of the noise does not matter.
The THD+N Cosmos ADC is mostly determined by noise, so a cleaner sine is not critical. Therefore, I did not take into account source distortion in the calculation. The 9039S does not require alternative firmware to switch MCLK. My sample shows THD -147.8 dB on one of the channels, but the noise increases slightly. This was measured on a Cosmos ADC + Cosmos APU (notch). Since this measurement is at the limit of the APU accuracy, it is performed at the expense of the noise measurement accuracy.
View attachment 367669

Why do you choose exponential averaging? I'm not clear on the advantages and disadvantages of each averaging method.

Nice work, by the way...

 

[nick_l44.1]

Why do you choose exponential averaging? I'm not clear on the advantages and disadvantages of each averaging method.

Nice work, by the way...

Thanks! I have already answered this question. I picked up this averaging method from Ivan.

 

[BKDad]

Thanks! I have already answered this question. I picked up this averaging method from Ivan.

I completely understand the value of averaging a measurement. I just wondered why you selected exponential 0.5 over the other averaging selections.

 

[nick_l44.1]

I completely understand the value of averaging a measurement. I just wondered why you selected exponential 0.5 over the other averaging selections.

I didn't attach much importance to it. Initially, I repeated Ivan’s measurement and did not try to change the parameters, because the results were noticeably different. It turned out that I simply had a better es9039q2m chip.

 

[BKDad]

I didn't attach much importance to it. Initially, I repeated Ivan’s measurement and did not try to change the parameters, because the results were noticeably different. It turned out that I simply had a better es9039q2m chip.

Generally, the advantage of using an exponential averaging technique is to emphasize more recent samples/sweeps. This allows for easier settings on whatever potentiometer you need to adjust, for example. I didn't know if there was any other subtle advantage.