Frequency response testing using PC and AI.

[Diabolical Artificer]

I've recently been experimenting with various free software audio testing programs and ran a few tests yesterday on a valve amp I've designed and built. I ran some  frequency response tests on it using first audiotester.de, then a simple frequency response program and lastly with REW.

Every test showed a less than flat response with it varying by as much as 10dB in places. I then ran a manual test using a very good vintage wein bridge oscillator (THD less than 0.01%) and a vintage B&K AC valve voltmeter. the response was flat from 2hz to 20khz with a slight dip at 4hz (1/2dB) and a slight rise of another 1/2dB from 10khz up to 20khz. I monitored the IP with a Fluke bench DMM, the OP with a Fluke 25 as well as the B&K AC voltmeter as well as both on a Tektronix scope.

So what's the craic? The Soundcard was calibrated on all three bits of software and the old test gear has also been recently calibrated,  I tend to believe the old test gear. One thing I have noticed on a few of the software programs is that the sweep is modulated.

The only difference in the two types of measurements is that the PC software goes through an attenuator, both use the same dummy load. thought's?

Andy.

 

[ruffrecords]

Try the white noise source in REW for direct measurement of frequency response.

Cheers

Ian

 

[trobbins]

In REW, if you are using the MEASURE icon sweep, then try using a higher LENGTH value (up to 4M for latest rev) as that slows down the sweep time.

Can you describe if you did that sweep with the same output and input probe and settings but with the output and input probe tips joined together (ie. a loopback test using all the test hardware, but not with the amp directly in the loop). 

Can you describe the test setup from soundcard output through to soundcard input (with amp in the middle somewhere).

 

[Diabolical Artificer]

Thanks Ian, will do.

Loopback test - I thought of that after posting, IE testing a straight bit of wire, will do so and report back.

Test set up is SC line OP to amp IP using  2ft long screened lead. OP of amp goes into 8 ohm dummy load that has 3 BNC's per CH to monitor OP with scope etc. Also inside the same box is a 100:1/10:1 variable attenuator this goes out to the SC line in.

Testing a loopback/straight bit of wire type test is problematic as the SC will have trouble driving an 8 ohm load probably. I will try monitoring the sweep OP of the SC card with an analogue meter to see if the OP V of the SC is changing and will do some more tests and report back.

Andy.

 

[trobbins]

If you can disconnect the 8 ohm from the attenuators and do a loop back to the top of the attenuator, then that would include any loop response shape from the  attenuator and your line in.

 

[abbey road d enfer]

Every test showed a less than flat response with it varying by as much as 10dB in places.

That is quite disturbing. I have tested a few soundcards and I noticed some have sugnificant deviation but not as much as what you measure.

  I tend to believe the old test gear.

I do too.

One thing I have noticed on a few of the software programs is that the sweep is modulated.

I think you mean warbled. It's not unusual for acoustic sweep measurements.

The only difference in the two types of measurements is that the PC software goes through an attenuator, both use the same dummy load. thought's?

That's what nags me with PC+souncard measurements; it's supposed to be simple and compact, but you end up with all sorts of bits and pieces that could all go wrong (including software).