Problem with amplifier stepped sine testing

[dch53]

Hi,

I've done stepped sine testing of various amps before without a problem. Now however I'm getting plots like those attached.

If I do a single frequency RTA the result is fine.

Any suggestions as to what I'm doing wrong please?

Thanks!

 

[John Mulcahy]

Are you using ASIO drivers? What is the ASIO buffer size?

 

[dch53]

Hi John,

Yes, I am using ASIO drivers. The only setting in my Juli@ control panel is "Latency" and I currently have it set to the maximum of 2048 samples.

Thing is, up to a couple of weeks ago I wasn't having this problem. I've checked all the tasks that are running and closed down as much as possible but I'm not able to get the results I used to. CPU usage is 3% or less, most of the time down around 1%.

I tried my soundcard interface in loopback mode, which used to be excellent and the results were like this:



So the problem isn't the amp I'm testing.

Again, the RTA at individual frequencies is fine:



REW version v5.2 Beta 8 but I suspect the problem isn't with the software.

Regards, Dave.

 

[John Mulcahy]

When averaging starts is the spectrum normal (looking like the RTA plot) or is it already bumped up? If it goes wrong during the averaging something may be getting on to the output, if it is already wrong when the averaging starts there may need to be additional settling time before starting to average.

If you can listen to the test tone while carrying out the test check for any pops or clicks. Also watch out for the interface being assigned as the Windows default for playback, which might allow Windows-generated sounds to get onto the output.

 

[dch53]

When averaging starts is the spectrum normal (looking like the RTA plot) or is it already bumped up? If it goes wrong during the averaging something may be getting on to the output, if it is already wrong when the averaging starts there may need to be additional settling time before starting to average.

If you can listen to the test tone while carrying out the test check for any pops or clicks. Also watch out for the interface being assigned as the Windows default for playback, which might allow Windows-generated sounds to get onto the output.

To answer some of your questions:

• the interface is not assigned as the Windows default for playback

• The 19.8Hz response is normal. All subsequent steps starting from the next one, 24.9hz, bump up from the first sample.

• I listened to a test tone and there does seem to be noise superimposed on the tone. If I reduce the amplitude to -75dBFS and press Start, the tone is just audible but the noise is plainly audible. It looks like this on the scope:

​

-75dB is a pretty low level though and perhaps that's just the residual noise in the computer/soundcard/soundcard interface system.

With a lot more testing I have found that the problem is present with the soundcard in loopback mode.



It's puzzling that the soundcard would suddenly develop this problem.

Any suggestions before I try the card in another computer (which would be a bit of a pain) or start to look for a new soundcard?

Some questions:

• How do I specify additional settling time before starting to average?

• Is there a setting I could have changed inadvertently to cause this problem?

• Why is the problem apparent on the stepped sine test but not with RTA of a single tone?

Thanks again for your help.

 

[John Mulcahy]

I wouldn't blame the soundcard, it looks fine on a normal RTA plot after all. More likely my fault. Try increasing the overlap, there may be an issue with the settling time I allow when there is no overlap for long FFTs.

 

[dch53]

...Try increasing the overlap...

Ahhhhhh! That's my problem! I've been running with an overlap of 0%!

As soon as I changed it 50% my plots were as before.

The first "average" looks like this:



while subsequent averages look like this:



I knew it had to be something I'd inadvertently changed. Now I can go back to testing amplifiers.

I do have this problem quite a lot:



I think it's some sort of interference. I'm sure I've heard a noise at the same time when testing amplifiers.

Thanks John.

 

[John Mulcahy]

If you spot a bad point you can use the Back button on the stepped sine dialog to go back and redo the measurements from that point onwards.

 

[dch53]

Thanks John. That will be very helpful.

I really need to find the cause of the glitches. I suspect the fridge.