Behringer UMC 202HD for measurements

[MagicBus]

EDIT: Sep 15 2019. It is advised to read the whole thread before start doing anything with your soundcard. There are revisions that make previous posts invalid!


Honestly, this Test & Measurement interface for Soundcard together with ARTA was more than enough for my humble tube projects. But it couldn't do impedance measurements with ARTA limp. It has been explained that those cheap soundcards don't have adequate input channel separation for this job. Understood... So, I spent the double and got UMC202. I was pleased to see that as is from the box it does excellent impedance measurements without even requiring the headphone output for reasonable loads! And with an amp -Rod Elliott's workbench amp fitted with ARTA limp probe- it can measure coils with very low DCR. If anything else, it would be a keeper just for this. But I wanted to explore possibilities for FFT as well. The short answer is not as it comes from the factory... Bellow is the best I could get.

It makes sense. It has high gain input preamps and perhaps tuned for some musical distortion. I decided to tweak it based on some empirical experience from previous similar projects. What I thought it would be a weekend's project, it took more than a month -OK, half an hour per day. So I managed to isolate the audio codec converter http://www.mouser.com/ds/2/76/cs4272_f1-43250.pdf -ADC/DAC- and its input buffer and immediately there was a significant improvement in performance.

Still, two problems. It was clear only up to -17,6dB and there was a high frequency peak - here at 9kHz but from day to day I've seen it anywhere between 8 and 11kHz. It took a lot of sniffing around, mostly because of my ignorance. You can read here a small part of it SE to BAL conversion with active filter The conversation and some more reading helped to find my way at least to some point. This paper was very helpful https://d3uzseaevmutz1.cloudfront.ne...te/an241-1.pdf. At page #5 there is what I found inside the UMC 202 more or less. There were some differences -red in the schematic bellow

Too high output resistors for the op amps and another one across the differential inputs. I've seen that in some designs marked as "optional". Perhaps an attempt to avoid hard clipping as this soundcard is primarily meant for musicians? I can't tell. But I know that replacing the 1k resistors with 91R and removing completely the 10k resistors was another good step.
Unfortunately, the high frequency peak is still there together with its k2 harmonic and +/-1kHz IMD. The graph would be much cleaner without it. I made the thought that the capacitor circled in red in the previous schematic could be responsible for this -very low? I tried to add 1nF polystyrene caps in parallel with them. I think I saw a tiny improvement but I can't tell for sure. Any ideas about this would be very much appreciated. I can provide some more info about that.



So, at the end the feelings are mixed... I could spend a little more, or maybe a little much more and get something ready that does better, but I started from very little money and stayed there, plus DIY satisfaction and knowledge gained. I was even tempted to design a soundcard from scratch but it is the software that repeals me. The driver's functionality is very important and I've seen manufacturers failing in this domain. Apparently, Behringer does this right.


I talk about Behringer, ARTA and Pete Millett's PC interface. I do not have any relationship with these other than I use them as a low cost measurement rig.


In case anybody would be interested on the modifications, I can provide all the details.


Thanks for reading!

[MagicBus]

Almost good... I can't find what's causing this high pitched noise. I'm pretty much clueless about it. I'm posting three observations that might tell you something...
First is that although the frequency will be stable at each session, it is moving from day to day. Before I start modifying it this was between 8 and 11kHz. Now I see it happening between 6 and 8kHz. Second is that the ADC input alone -chip pins floating- is silent (see first pic). If it is connected to anything, the peak appears. By anything I mean buffers, the one on board as well as a number of external circuits, all silent and with low THD. However, everything was sharing the same ground for what matters. The third observation has to do with what you see in the second pic. It is the same channel recorded two times -note the orange line under the green- and above this is the difference of the two measurements. Zero for everything except for the peak! Does this give you any ideas about the nature of this noise?

[trobbins]

In the EMU0404 USB, internally generated noise was coupled in on one channel a lot more than the other - and the fix for the noisy channel was to divert signal from a long trace that went past a noisy circuit section by going via a shielded cable. So that issue was very much board layout and trace layout dependent.

E-MU 0404 usb 2.0 and some "artifacts peaks" on left channel

[MagicBus]

Thanks for reply!
Although I've seen this thread in the past, I forgot about it. I'll read it again. I'm sure it is a layout problem. Perhaps I was looking in the wrong place. The board is multilayer with ground planes on top and bottom so hard to see a ground problem. The ADC/input and the USB interface sections also seem very straightforward. But the DAC/output section is a mess. I thought that it could not affect the input but you never know with digital circuits. After all, this is the only thing left to check. I'll be back.

[mbrennwa]

Does the choice of the sampling frequency affect the "high frequency peak" in any way?

[MagicBus]

Hi,
About software settings: I tried Behringer ASIO, WDM and ASIO4ALL with all possible sampling frequencies but nothing affects the peak. In particular, with WDM it is possible to set sampling rates as low as 8kHz and what I noticed is that when the sampling rate is not a multiply of 8kHz (11kHz - 22kHz) there are more artifacts. With 8, 16 and 32kHz I get only the "peak". And with 44,1kHz too, even though it's not relevant to 8.

[MagicBus]

Bingo! I had the feeling that it had to do with Vcom. The ADC inputs need DC bias at Vcom level but apparently not the actual Vcom, not even buffered. There is a buffer for this on board. It's the half of a RC4580 dual op amp. The other half is also a DC buffer but referenced to a resistance voltage divider from power rail. I couldn't trace where this is going but I simply lifted the other from Vcom and fed it from the voltage divider too. Seems to work! The high frequency peak is still there but it doesn't dominate the spectrum any more.


While working on this, I tried the ADC without DC bias at all. Distortion rises dramatically and it is not about clipping. It seems that the DC bias at half the rail voltage keeps the balance internally in the digital domain. At least this is how I understand it, maybe it's something else. Anyway, all this gave me the idea to replace the fixed voltage divider with a pot to allow trimming "on the fly". I'll check this and also optimize the input capacitors of the SE to BAL converter because this has to do with 50Hz peak.

[mbrennwa]

Good stuff! Why did I switch from the 202HD to an RTX6001?

[MagicBus]

I guess because RTX6001 is better! OTOH, this one is much cheaper...

[mbrennwa]

And the RTX has proper switches to set the signal levels, not pots. And yes, it is better. But your tweaks make the Behringer really a lot more useful!