UCA202 owners, please read!

I'm in a hurry, but real quick, if you do a loop-back cal, then run a FR sweep while still looped-back, you should get a flat line. If not, you've done something wrong. Or, you have a bad UCA202

Make sure you have the mic in your computer turned off. There are some things that need to be set up within the windows settings for it to work correctly.

Try turning the C-Weighting off and see if the phantom energy below 15 Hz goes away.

By phantom energy, I mean it's not real, it's not really there, it's being added in from somewhere, and I think it is the C-Weighting. As for why it would induce error in your traces, beginning at 15 Hertz on down, C-Weighting picks up where your Mic Calibration file ends, which I'm guessing in your case is around 15 Hz.

This. My UCA222 does the same thing if I leave C-weighting enabled.

I dont see any wierd boost at 15hz on any of these!!!! Looks like a pretty straight forward response on that last one

EDIT: I see what you are saying now, the response begins to rise again as you move further down from 15hz. Got it and I saw this too with the c weighting...The A weighting seems to have it go away, but you are still seeing some rise down where you shouldnt. If there isnt enough content that far down, the soundcard might be doing something funny. I know with my original setup it did just that below 12-15 where I didnt have enough substantial output, but now that I do, it measures properly. strange

Try it with No-Weighting.
Beast--I believe what Mark means is, the sub's output should continue its downward slope. But at 15-16 Hz, it begins rising with C-Weighting, sooner with A-Weighting (around 35 Hz).

and in your preferences tab as well?

It's your noise floor. It is difficult to get a decent SNR below 15hz. Many measurements are not showing anything of value that low. After awhile you can spot what is noise. Calibration files with a lot of boost exacerbate the issue. I'm usually skeptical of anything that low in frequency of less than 80dB with the spl calibrated and no cal file in use.

What he said...

It is noise, not signal.

In my opinion, measuring anything below 10 Hz with DIY-grade gear is pointless. Typically, I do not present data below 10 Hz for that reason, but you caught me in the F-20 thread.

I understand that this is just noise - but where is the noise coming from? Old thread - I know.

Here's an RTA with no signal - just ambient floor noise.


And this is the same window, with pink noise generated but with the receiver MUTED.


And this is a 0-500 sweep with mic and soundcard cals shown. Again, the speakers are MUTED, so there is no sound here other than the noise floor, which is a little above 40 dB.


I don't see how this could come from the soundcard calibration, as the soundcard loop measurement from which the cal was made is down 0.6 dB at 10hz and down 3 dB at 4hz. And the resulting loopback measurement after applying the calibration is flat as can be.


Could this be some sort of cross talk between the input and output within the UCA-202?

Here's another one.

These are saved from REW's RTA. Blue is just the noise floor of the soundcard, my computer, whatever. Pink is with full-band pink noise being output from the UCA-202. The inputs of the UCA are empty - no mic, nothing.


Doesn't this look like a big problem? Even at high measurement levels (which I'd prefer to avoid), at what point do you think these results are trustworthy?

Here is a combined chart of the above with some close-mic measurements of 0-500 sweeps (on top). Not exactly the roll-off I'd expect for a sealed box.

I had the same issue with the card. I since moved to Edirol UA-1EX, the problem is solved.
John M of REW said that "the output impedance/input stage interactions" are likely to blame.



green line is REW sweep with speakers disconnected. the black line is RTA snapshot of the room.


I think it is safe to rule out the noise at this point.

I'm looking into this. I do see this behavior with some combinations of hardware and software, but I am not seeing it consistently.

If you use a mic with a variable preamp, you can calibrate your mic at a higher level to run your mic signal hotter if you want to better signal to noise numbers at low measurement levels.

The issue is that the apparent (because I only see it on some systems) noise floor of the UCA202 starts to rise below about 80 Hz, and gets to a point where the headroom available for measurement drops to around 25 dB by 15 Hz, which buries a low-level signal below the noise floor.

Yes, this is definitely a limitation when it comes to measuring things at low levels below 40 Hz or so, but as I said in my previous post - I only see it with some combinations of hardware and software. I am looking into this as I can.

I confirmed that it happens with my XP laptop while running both the Java and the Behringer ASIO driver. With the same laptop and different sound cards, I see "proper" noise floor behavior. I can also confirm that I am seeing similar behavior in both REW and ARTA under XP using the Behringer ASIO driver, so this is not an REW issue. So far, I am not seeing the same behavior with Windows 7 (32-bit) and either the Java or the ASIO4All driver, but I have not tried the Behringer ASIO driver under Win 7 yet.

Anyhow - this seems like "new" behavior to me. While I don't have screengrabs to prove it, I don't recall a noise floor like this in my early use of the UCA202. One of the first things I do with any sound card I'm planning to use for acoustic or electrical measurement is a set of basic measurements: Loopback response, noise floor, and a quick check of THD at 0 dB and 1 KHz.

While it isn't as good as my Emu 0202, the Behringer measured well enough for me to work with it and recommend it for several years, and it is widely available for less than $30. Since I don't see it in all cases, I do not think it is hardware-related, so I hope this is as simple as a setup or driver issue.

Anyhow - I am looking into this, hope to have more information soon.

Thanks, lilmike. I have an m-audio firewire device that I have to use for phantom power anyway. I've taken the UCA out of the equation for now. If I get the time I'll test your idea of higher mic signal.

I didn't try any different drivers. I just used the Win764's native control of the device.

I see the same behavior across all my UCAs and in both OS's that I use. It appears to be a hardware issue.

Basically, as was explained earlier, this is "crosstalk", where the output signal makes it to the inputs even when they are not connected. It appears to be an inherent design issue with the UCA, and I see it in both the 202 and 222. The sneaky part is that it is very frequency dependent. At low frequencies, the amount of signal leakage is high, and it decreases as frequency rises, to where by 1 KHz, it is not significant at all.

Is it an issue? Depends on what you're doing.

If the UCA is used as the signal source or the signal input?
Not an issue, as the other "side" of the UCA is not in use.

If the UCA is both the source and input, as I typically use it?
Well, the noise floor rises by about 50 dB at its worst. Depending on the setup, this may cause issues with low-level signals at low frequencies, as the signal to noise ratio is diminished.
Within its limits, this can produce a perfectly acceptable measurement. In my case, my noise floor is around 30 to 35 dB below my loopback signal. While not as high as I'd like it, there is still ample room for a signal. At -30 dB, this noise is a 3% distortion or error in the signal. However, if the signal is not as high, and the noise is only -10 dB, you're at 31% error.

I run the hottest signals I can for this reason.

If people are concerned, I see two ways around this.

Use a USB Mic instead of the line in on the UCA, and fortunately, there are some great options now. This eliminates the possibility of crosstalk.

Use a second device - dedicate one to line out, and one to line in. This is even cheaper than a USB Mic if you've already built a mic or already own a calibrated mic and preamp.

I've experimented with the two-device option, lots of options using the Java drivers. Under ASIO, things are a little tougher, but using 2 UCAs at the same time works fine with the latest beta of the ASIO4All drivers and the latest Beta of REW under XP and Win 7. This approach won't work at all with the Behringer ASIO driver in either case. Though it is a little fiddly to set up, I can maintain a better than -80 dB noisefloor from the single digits to 20 KHz this way.

I'm looking at other line in and line out options, but have not found anything better to date. I've also ordered several new cards for evaluation, but they're not here yet.

Great!! I just ordered a 202 from the bay.

I wonder if there is a way to reduce the cross talk. I wonder if this is happening due to the chips they are using
or a pcb design problem , Or possibly the cables used outside of the 202. That is inadequate shielding .

Nashou