RTA Distortion misidentify harmonics

[bgilsdorf]

When using the RTA for distortion analysis I find that the distortion components are not correctly Identified with some sources. If I use a battery powered wien bridge osc the identified harmonics are correct. If I use my dac (Cambridge audio magic) they are not) I can use a cursor and identify the correct value. but it reduces the efficiency of the tool. Has someone else observed this behavior. Is there a simple fix. It almost seems like a frequency discrimination problem although I am not sure how that would work

 

[John Mulcahy]

If you are playing the REW sine wave generator the frequency of the fundamental will be taken from the generator setting, otherwise it will be extracted from the captured RTA data. Difficult to comment beyond that without seeing what you are referring to, a screenshot from the RTA or (better) a saved RTA measurement attached as an mdat would be helpful.

 

[bgilsdorf]

 

[bgilsdorf]

The problem appears to be that the dac output is slightly shifted in frequency. 1khz is 1.01khz
Is this a problem with the length of the sample, too small frequency bins. here is the ctual dac measurement.

 

[John Mulcahy]

Are you playing the REW sine wave generator?

 

[bgilsdorf]

yes if I use the sine wave generator I see the problem. I have not tried another dac, Only an external osc

 

[bgilsdorf]

The frequency error is less than I said. Closer to 1005 vs 1000

 

[John Mulcahy]

That suggests your DAC and ADC clocks are very far apart (turning 1 kHz into 1.005 kHz requires a 5,000 ppm clock rate error, a poor DAC or ADC clock might be off by 100 ppm). You could save the test signal to a file and play that file back through the DAC, REW would then just use the input to determine the frequency. The file duration would limit your test time though. I could look at adding an option to not use the generator frequency, though I haven't previously come across a clock rate difference large enough to cause such an issue. Probably worth trying a different DAC to see whether the problem is with the DAC or the ADC.

You should also switch to the WASAPI exclusive device entries (names start EXCL) if you aren't already using that, that noise floor looks quite like the behaviour of the Windows 16-bit conversion that happens when using the non-exclusive Java drivers.

 

[bgilsdorf]

It is actually 999 vs 1000. Changing the length of the fft seems to mask the problem

 

[keantoken]

Most of my equipment is over 100PPM in variation, that seems a little optimistic. The automatic clock difference compensation is very helpful.