E1DA Cosmos ADC

[BKDad]

Hi, is it possible to bring the level up to -5dBFS in the second test, like in the first one?
Unless I've missed something, it would be better to compare to get the same level at the ADC input.

I should be clearer, perhaps.

The output of the test generator is exactly the same. The output of the Scaler is exactly the same. Those weren't touched. But, by running the output of the Scaler through the notch filter of the Cosmos APU, the level of the 1 KHz fundamental tone into the ADC is reduced, with the harmonics mostly unaffected. That minimizes the distortion added by the ADC, so we just see the distortion coming from the generator.

There's way more here: Archimago on the APU

His results are similar, mostly. He used a slower sampling rate for the ADC, which would lower the noise a bit. He also had his ADC internal attenuator set differently. Plus, some of his measurements settings are different. I didn't even try to match what he did. I just wanted to do a with-APU and without-APU comparison. Also note that his generator (same as mine) is not in a shielded box and he also uses a somewhat different powering scheme which I found reduces performance a bit.

Later on, he optimized his generator to get similar results to mine: Archimago improves his generator performance

 

[Grooved]

I should be clearer, perhaps.

The output of the test generator is exactly the same. The output of the Scaler is exactly the same. Those weren't touched. But, by running the output of the Scaler through the notch filter of the Cosmos APU, the level of the 1 KHz fundamental tone into the ADC is reduced, with the harmonics mostly unaffected. That minimizes the distortion added by the ADC, so we just see the distortion coming from the generator.

There's way more here: Archimago on the APU

His results are similar, mostly. He used a slower sampling rate for the ADC, which would lower the noise a bit. He also had his ADC internal attenuator set differently. Plus, some of his measurements settings are different. I didn't even try to match what he did. I just wanted to do a with-APU and without-APU comparison. Also note that his generator (same as mine) is not in a shielded box and he also uses a somewhat different powering scheme which I found reduces performance a bit.

Later on, he optimized his generator to get similar results to mine: Archimago improves his generator performance

Thanks, I knew how the APU works but I've mainly misread your post and didn't understand at first that you used the notch filter of the APU, I thought you only used the preamp part. So the second capture makes sense to me now.

 

[IVX]

I should be clearer, perhaps.

The output of the test generator is exactly the same. The output of the Scaler is exactly the same. Those weren't touched. But, by running the output of the Scaler through the notch filter of the Cosmos APU, the level of the 1 KHz fundamental tone into the ADC is reduced, with the harmonics mostly unaffected. That minimizes the distortion added by the ADC, so we just see the distortion coming from the generator.

There's way more here: Archimago on the APU

His results are similar, mostly. He used a slower sampling rate for the ADC, which would lower the noise a bit. He also had his ADC internal attenuator set differently. Plus, some of his measurements settings are different. I didn't even try to match what he did. I just wanted to do a with-APU and without-APU comparison. Also note that his generator (same as mine) is not in a shielded box and he also uses a somewhat different powering scheme which I found reduces performance a bit.

Later on, he optimized his generator to get similar results to mine: Archimago improves his generator performance

if that FFT was taken with APU, I'm not surprised at all, I thought it was your ADC that has -150db of the distortions. A+ on the PCB means that the noise is close to -129db(A).

 

[BKDad]

if that FFT was taken with APU, I'm not surprised at all, I thought it was your ADC that has -150db of the distortions. A+ on the PCB means that the noise is close to -129db(A).

I must be really bad at communicating. So says my wife, anyway. Two guys who read two different things into what I posted! That must be on me.

The first plot is oscillator > Scaler > ADC. That's it. No cal file used in REW.

The second plot is oscillator > Scaler > APU (notch at 0 dB) > ADC. REW using the cal file for the APU.

 

[IVX]

Actually, from time to time, I see unusually low THD ADC units. Just 30 mins ago I calibrated one grade-b unit with THD@1k about -140db, THD+N of course not so good(-119db stereo, hence -122db mono) due to grade-B noise but harmonics are crazy low. So, if somebody is interested you can ask Sunny about that particular THD+ ADCiso, for the price of normal grade-B.

 

[misureaudio]

Poker of Instruments.

I'd have not to be lazy, power on the soldering iron and make some APU/Scaler cables.

Any chance to find an official E1DA dedicated cable set in the future?

 

[IVX]

Like that? https://www.aliexpress.us/item/3256...08896804!sh!US!776512775&gatewayAdapt=glo2usa

 

[misureaudio]

Like that? https://www.aliexpress.us/item/3256804203213458.html?spm=a2g0o.store_pc_allProduct.8148356.22.2bb97555DYtJ4H&pdp_npi=3@dis!USD!US $7.99!US $7.99!!!!!@2103246c16970422087282752e117d!12000029008896804!sh!US!776512775&gatewayAdapt=glo2usa

Thanks.

 

[BKDad]

ADCiso has about 3db less H3 at 10kHz(-123db@-1dbfs).

Here is the same test set up as before, except now using a new ADCiso. Everything else is the same. (See Previous Post). Power for the ACiso is from a battery power pack. A linear AC wall wart from Jameco looks the same.

The thermostat is installed and the plot shows the effect of 30 minutes of warmup. Sample rate was 384k. 96k and 192k look to be almost identical with regard to distortions.

 

[IVX]

is that measured with Scaler?

 

[BKDad]

Here's a puzzle for all of you.

Tonight, I compared the results of the above test setup (at a slightly different power voltage level) using REW in both Windows and macOS. Same computer, just rebooting into different operating systems.

Everything was set up the same. I went back and forth a couple times and the measurements were essentially perfectly repeatable. Even the cables weren't touched.

First, here is the REW graph from Windows 10:

This is the macOS REW graph:

Very close, but not the same! Look at the 2nd harmonic. Interesting, eh?

Which one is right? At this level, it probably doesn't matter. But, it still is curious.

Now, for my real second question.

Why does the output of the Cosmos ADC glitch or bounce on occasion? This happens in both the macOS and Windows versions of REW. I thought it might be the computer. So, I tried on another Mac that just runs macOS. Same thing. And, I could see the same phenomenon using a macOS only audio program named Amici. Same set-up, except there is a different USB-C cable between the ADC and the computers. Any ideas?

 

[BKDad]

is that measured with Scaler?

Yes.

The "Victor" oscillator (Victor Oscillator) - mine is not the latest - to the Scaler then to the ADCiso. Both the Scaler and ADCiso were powered by battery power packs.

 

[IVX]

As I understand, you have 2 ADC on your hands, both of them glitching? Try pls different USB ports and/or the shortest cable.

 

[BKDad]

As I understand, you have 2 ADC on your hands, both of them glitching? Try pls different USB ports and/or the shortest cable.

I did. This happens across two different computers. Two different operating systems. Tried different cables, none more than about .3M long. Obviously, different ports.

The glitches get averaged out in REW, although sometimes I have to restart the averaging manually.

In the Windows REW version I'm using FlexAsio as the driver.

I also have observed this using the Multitone software application in Windows. There I have a wider choice of drivers. They all seem to do it.

Edit: This isn't a complaint. It all is very workable. I just wondered what it could be.

 

[IVX]

As I understand, you have 2 ADCs on your hands, both of them glitching?

 

[BKDad]

As I understand, you have 2 ADCs on your hands, both of them glitching?

Yes.

Sometimes more often than others. I have much more experience with the ADC Classic, of course.

 

[IVX]

I can see such glitches only with cheap cable, the white one in the pic, it is a very flexible and nice-looking cable vs fat and hard fully-featured black cable.

 

[GXAlan]

I can see such glitches only with cheap cable, the white one in the pic, it is a very flexible and nice-looking cable vs fat and hard fully-featured black cable.

Should we just use Thunderbolt rated USB-C cables?

 

[IVX]

I dunno, the white one has 4pcs thin wires inside without any shielding, and it works but with lost packets.
I have one more modification of that cable, now it is red with braiding, and thick like the black one but still the same inside.

 

[BKDad]

Should we just use Thunderbolt rated USB-C cables?

That was one of the cables I tried with a new M2 Mac Studio. It wasn't any better.

Please note that the situation for me is quite workable. I just am curious. At first, I suspected that I had some setting wrong in the bowels of Windows or maybe FlexAsio. But, then the same thing happened with a different computer running macOS, with two different software applications.

One of my friends spends part of his workday working on USB problems like this. He says USB is quite capable, but the transmission is hardly ideal. He sees errors all the time in supposedly great USB systems and cables. You just rarely detect them because they don't affect the final transmission quality. But, since audio is a real time process, you can hear when a packet gets corrupted. It could be a pop on the playback side. For spectral measurements with an ADC and software like REW, you get one sample that's not right and the averaging function blows up. BTW, telecommunications system have the same issues. They get around it all with comprehensive error correction schemes. We don't have that with audio.

I'll play around some more.