[JohnR]

I'm testing some soundcards, and one of them can be persuaded to produce output like this:



That's at 48 khz sample rate. At 96 kHz (no other change) it looks like this:



One interesting thing is that the noise spikes seem to oscillate at the sample rate. They seem to occur at fairly random intervals - those are just grabs of a section not a repeatable waveform. Another is that the level of those spikes varies with input gain. Not output level; input gain. If you increase the gain, they get more numerous and start to look almost like random noise. Decrease the gain and they largely disappear. So I'm completely bamboozled on this one.

Has anyone here seen anything like this?

BTW the sine wave is 1mV RMS.

[FullRangeMan]

Came to mind brick wall by product, but I may be wrong.

[Davey]

One millivolt RMS?

Heck, I'd say that's pretty good.  For most of the soundcards I've looked at, a 1mV sine signal would probably be unrecognizable in the high frequency noise present.

Cheers,

Dave.

[JohnR]

Oh. They must be getting better as a general rule, here's a different soundcard under basically the same conditions:





With the first one, the behavior is very strange. The noise seems to change with input gain. I can't figure out what's going on, I'm wondering if there's some sort of aliasing going on...

I suppose the obvious thing to do is to say forget about the first one and keep the second one But I'm curious about it and don't have better test gear to investigate further.

[kc8apf]

What is the test setup?  Are you using the soundcard for both output and input?

[Davey]

John,

What soundcard is that?  I want to buy one. 

Dave.

[JohnR]

Hi Dave, the second one is a MOTU Microbook II. It's quite a neat gadget. I put some measurements at this link - http://johnr.hifizine.com/2012/09/20/microbook-ii-measurements-i/ - I'd certainly be interested to hear any thoughts you may have about how I've gone about it.

kc8apf, yes, this is a loopback test, with line out connected to the mic input. I had the line output set for 1V full-scale, and then turned the signal generator down to 0.1% FS. The signal to noise ratio decreases quite a lot as the mic gain is raised. I can't make head or tail of it. Perhaps its time to finally bite the bullet and get a DSO.

[Davey]

Ah,  I guess the loopback configuration explains the lack of high frequency noise contaminating the sinewave........since you must be low-pass filtering the input at 48k/96k/whatever which yields the "cleaned up" display.  (Maybe some other filtering happening within the ADC architecture as well.)

If you monitored the output with an external oscilloscope, I suspect you would something much different......owing to possible noise-shaping, and other noises induced into the output signal.

I'm wondering why you used the mic-input vice the line-input for a loopback test, but will read your webpage for your test description.
Soundcards have a variety of maximum output levels and they rarely match the maximum input level of the line-inputs.......and the Windows mixer can really add a level of confusion as well.  ASIO drivers seem to alleviate most of these issues.
Of course, you're using a Mac so everything is much simpler. 

Cheers,

Dave.

[JohnR]

If you monitored the output with an external oscilloscope, I suspect you would something much different......owing to possible noise-shaping, and other noises induced into the output signal.

Yes, quite possibly. Sorry for not mentioning the loopback configuration earlier. So I wonder what that would mean for the first graph I posted... hm, well unfortunately I have no way of knowing at present

I'm wondering why you used the mic-input vice the line-input for a loopback test,

Most of the tests were done with the line input. In this case though, I was trying to come up with a measurement that would have the mic input used at high gain. The idea was to come up with something that would show up problems with the mic input with a more typical setting used for measurements. So far all it seems I've been able to do is find problems with the outputs. I know that was not a very sophisticated test and was expecting some noise but not like that.

In the case of the Microbook, I'm using the mic input for all the loopback tests, as the line output when set at 1V won't drive the line input to full scale (the line inputs on this one have no gain). Whereas the mic input has a 20 dB pad so apart from a lowish input impedance its pretty much a line input as well (max signal is then +20 dBu).