Accurate 20Hz THD measurement?

[Jeb-D.]

Can anyone running a sound card audio analysis setup give me some pointers how to get accurate THD measurements at 20Hz?

I tried lowering the bandwidth and upping the FFT resolution settings. It helped some, but It is still far off. Waveform on the scope looks clean, but THD measurement says %60. This is a problem I have with every software package I have tried.

 

[scott wurcer]

60Hz is third of 20Hz, hum intrusion first thing I'd check just to be sure. By using exactly 20Hz there are several opportunities for ambiguity.

BTW I have had no problems with Audition/Cooledit in doing low frequency distortion.

 

[forr]

Hi Jeb-D,
As Scott implies, you should avoid second or third harmonic to be right at the mains harmonic.
What you are going to measure ?
I am interested in measuring harmonic distorsion at very low frequencies to detect non-linearities due to thermal effects in transistorised circuits.

 

[unclejed613]

what's the voltage going into the sound card? anything beyond 1V peak into a line level input, or 10mV into a mic input is going to be overdriving the input and read as distortion (or worse, let the blue smoke out).

 

[Jeb-D.]

Scott,
That was a very good idea, but it happens even with the soundcard output is plugged straight into the input. I'd think that the amount of 60Hz residual to cause a %60ish THD reading would defiantly be visible on the scope. Eventhough 20Hz is the worst, the high distortion issue seems to occur at any frequency 100Hz or below(approximately).

What you are going to measure ?

Once I figure out what is wrong with the setup, a vacuum tube amplifier. Distortion is always the worst in the low frequencies because of output transformer limitations.

It's got to be some driver, setting or other software thing with my PC. I just have to keep looking into it. I have no patients when it comes trouble shooting computer issues. Aside from issue mentioned, everything seems to be working. 1kHz THD measurements seem accurate.

 

[Jeb-D.]

what's the voltage going into the sound card?

Less than 1V at all times (I'm using an attenuation circuit). The THD measurement is about the same regardless of input voltage. I tried .1V and it was about the same THD as 1V. Measurements above the 100-200Hz range seem fine.

 

[rdf]

Same issue closed loop, soundcard output to input? Did any of your software have real time spectrum analysis and show oddities? SineGen, WaveSpectra and M-Audio here with no issues.

 

[unclejed613]

i wonder if it's a sampling window problem. there are settings for some software that depending on your sampling rate, the software can only store and process a certain number of samples. if one cycle of the waveform is longer than this processing window, you don't process a whole cycle of signal, and this is interpreted as distortion. if the THD reading goes down as you select a lower sampling rate, this is probably what is happening.

 

[Pjotr]

Often this is a software feedback issue. In the Windows mixer: For the OUTPUT only wav should be enabled and nothing else. If also mic-in/line-in is enabled for the sound out, you’ll get feedback that way that disturbs your measurements.

 

[Jeb-D.]

if one cycle of the waveform is longer than this processing window, you don't process a whole cycle of signal, and this is interpreted as distortion.

That was it! There was a setting in the oscilloscopes control panel. I thought it just adjusted the scopes view window, but it was adjusting the record time along with it. I usually ignore the SW oscilloscope, because I have a real one connected.

Often this is a software feedback issue. In the Windows mixer: For the OUTPUT only wav should be enabled and nothing else. If also mic-in/line-in is enabled for the sound out, you’ll get feedback that way that disturbs your measurements.

I will will be sure to do so as a precautionary measure.

Thanks guys!

 

[unclejed613]

glad to be of help..... been using spectrum lab for audio analysis and i've learned a few things...... BTW like the DEVO avatar..... the place i work has an equipment status called DEVO where the equipment is non-serviceable. i use one of their album cover pics as a sticky note for DEVO items.

 

[unclejed613]

i'd like to get a sound card with a higher sampling rate than 96khz, because there are some uses a sound card can have as a VLF receiver. Spectrum Lab includes some DSP based demodulators and filters and such that can be useful for such an application.... which reminds me, i should check for the latest version....


it's available here:
http://freenet-homepage.de/dl4yhf/spectra1.html

 

[Jeb-D.]

I'll check it out. I usually use Visual Analyser (yes, they spell it Analyser not Analyzer) which is freeware too. It's pretty lame, but useful and the price is right.

Right now, I'm using an evaluation of Multi-instrument 3.1. It's pretty nifty and has a lot of capabilities, but the user-interface isn't that good. I'm in the process of giving them some feedback how it can be improved. They seem eager to make it as good as they can.

It has a feature called "device test plan", which is pretty much similar to the "sweep" function with an Audio Precision (if your fimiliar). You can setup and run automated frequency and amplitude sweeps while capturing all kinds of data such as THD, THD+N, SNR, IMD, Crosstalk, individual harmonics (2nd-7th), ect. It can create graphs or text data of things like input vs. output amplitude, THD vs. Frequency, THD vs. output level, and so fourth.

I set it up to do an inverse RIAA amplitude + frequency sweep. So when tuning the curve of my phono stage, I just have to try to get the plotted line flat. Makes things much simpler, as I no longer have to compare the amplitude at each frequency to published RIAA curve data every time an adjustment is made.

I'm considering buying it once the evaluation period ends, but may not be able to afford it at this time. So I may need some good freeware till then.