FFT question with audiotester

[zigzagflux]

Wondering why my 1 kHz FFT looks the way it does. I have measured this device before and gotten the expected results (same low noise floor with typical spikes at each harmonic).

Attached the noise floor and the 0dBFS sine wave. Besides the fact I can't seem to keep AudioTester running for more than 20 seconds without locking up

Running BH, but it doesn't really matter how many data points I acquire, nor which type of window, I get the same junk.

[SY]

What's your OS? Have you checked the system sound settings (e.g., effects off)? Control panel settings for your soundcard? Select the correct recording device, disabling the others? Anything running in background?

[zigzagflux]

Can always rely on you SY for quick responses.

I suppose here is where I get smacked around. I am using a MacBook Pro running Boot Camp. Windows 7 Pro 64 bit, no SP1. I have specifically deleted the on board hardware driver, so the only device available to AudioTester is the M-Audio Profire 610.

Fresh OS install, so there is nothing else running- the only reason I have Win7 on this machine is for AudioTester.

Control panel for M-Audio is pretty straightforward, nothing out of the ordinary from previous measurements. From AudioTester I have selected ASIO; I can play around with some alternates. What is preferred?

As a side note, I took some FFT measurements back in April using an HP laptop running WinXP Pro, measuring a completely different piece of equipment. Similar weird behavior which I didn't bother to investigate at the time. Attached is what I got, so I suspect this must be either operator error or an AudioTester issue with M-Audio.

[Pano]

Is that an FFT windowing problem? Although sometimes I see similar when the USB bus is glitchy out to the sound card.

[zigzagflux]

As it turns out, ASIO is the only driver that seems to work at all. Probably a Boot Camp issue. I'll borrow a real PC from work tomorrow and see how that does.

I would use all Mac, but there is no software app that I like for under $100. When all you want is basic FFT, it is surprising you can't find something.

[SY]

Sorry, with Macs, I plead total ignorance.

As Pano said, glitches in one of the buses will cause funky spectra like that- in the case of my PC (with the card mounted internally), I have to disable wireless NIC and devices in order to get stuff clean, otherwise the spectra look like yours. Ditto my other PC with an external Firewire card.

As a first step, I would try to turn off as much running in background or un-needed tasks as possible.

[zigzagflux]

Trying an HP laptop.
Sound card disabled.
Ethernet disabled.
Wireless disabled.
All the latest M-Audio and Audiotester updates.

Still same problem.
I have used this same computer in the past, and gotten the attached results, which is what I am expecting to see.

The device I am measuring is a DCX2496, D/A connected directly to Jensen output transformer. SPDIF input. Not much that can go wrong I would think. When I mute the DCX outputs, I get a quiet noise floor. This behavior only occurs when I have output excitation.

??

[SY]

What are the recording and playback devices and parameters in AudioTester set to? Can you use ASIO?

[zigzagflux]

ASIO is about the only thing that works reliably.

I think I found the issue, which is extreme jitter on the source. WOW.

The really bad plot in post #1 was using an AppleTV, which has toslink output to Twisted Pear optical/spdif converter, then directly to the DCX2496 onboard digital transformer. No analog I/O board installed in this DCX.

For the heckuvit, I put in a 2 kHz 24 dB L/R lowpass filter, and sure enough, the high frequency junk was gone. So this is not a problem with the measurement, D/A, Jensen, or output grounding; this has to be coming in by the digital source.

Not surprised that the AppleTV had issues, I used the toslink output on my MacBook Pro to the TP converter. Much improved sine wave FFT (first plot). Note that I am using a dedicated Mac source, but using the HP laptop with the sound card and Audiotester for measurements.

Okay, so it's the source. Then I removed the TP module, and fed the SPDIF output of my M-Audio card directly to the digital transformer on the DCX. Even better (second plot).

So, is this an indirect method of quantifying relative jitter? Is optical really that bad, or is this just my source? I am surprised the MBP toslink is 20 dB worse than the M-Audio spdif.

[SY]

It's nice when your instruments help pinpoint a problem! I probably wouldn't use this as a measure of jitter (though that could potentially be one of the sources of the problem- when jitter causes issues, it can be seen in the analog doamin), but with a clear difference like this in the analog output, something bad is happening and you know what source to avoid.

It may be interesting to repeat the same two measurements at 5k and 10k.