Impulse response artifacts

[Michaelgraf]

Hi there. I have a few questions about my measured impulse response. Has anybody already seen impulses similar to the one in the picture below?
The second pulse at 1.8 seconds changes with different sweep lengths. Thus the error is not time invariant and therefore not an artifact of the measured space. This must be due to my measuring system.
I used: RME Fireface UC, 96kHz sampling rate, Sweeplength 2M.

I need the measurements for evaluations of the RT60, so this is no problem for me. But what it is exactly, I would wonder
Thanks a lot!

 

[John Mulcahy]

Could you post an mdat file? Ideally with 2M and with 256k sweeps.

 

[Michaelgraf]

Unfortunately I only kept the 2M measurements, because the second "impulse" was within the usable range for the shorter sweeps. Attach you will find 3 measurements. 3 different speaker / speaker positions at the same mic position in the room as in the drawing.

 

[John Mulcahy]

The spectrum of the later peaks in Q2S2 and Q3S2 is suspiciously like the spectrum of Q1S2, aside from the 7 kHz peak. Odd that it should change with sweep length though.

 

[Michaelgraf]

Hm interesting. How did you separate the impulse response to view the frequency response of the second IR?
For me it seems like a artifact of distortion. I have read some articles today and an sub harmonic distortion may looks like the same and is positioned in the positive time of the IR. If it is a sub harmonic distortion, the shift in the time domain for sweeps of different lengths is explained.
The big question is now: Is it a subharmonic distortion, where is it from? If it's distortion products of the speakers, it would be strange that the ones from Q1 and Q2 look similar since they are different types of speakers.
The second possibility: Is it conceivable that this distortion consists of resonating components or cover plates?
I have not yet studied distortion products enough to be able to estimate this.

 

[John Mulcahy]

I moved the IR window reference to the later peak and adjusted the left window a little. Easily done by dragging the markers at the top of the Impulse graph. It would be unusual for distortion to have such a broad spectrum.

 

[Michaelgraf]

Ah great Idea! Thanks!
I think, the spectrum of this artifacts is between 6 and 12kHz. The remaining frequency components seem to be rather background noise and parts of the reverberation time of the main impulse, cause the level of the spectrum is 50dB below the Q1 courve.

 

[John Mulcahy]

The noise floor of the IR is around -89 dB, windowing the second R image to include only the portions above that still leave it looking like Q1S2 with additional 7/9/10 kHz peaks.

I see the soundcard cal file appears to contain a room measurement rather than a soundcard response, was that deliberate? If not it should be removed, though it doesn't affect RT60 results.

 

[Michaelgraf]

Alright! Thats possible. These were my first measurements with rew, and therefore thinkable
But this does not explain the peaks in the high frequencies, right?

 

[John Mulcahy]

No, it doesn't. The second IR is a curious feature, I'm not sure how it comes about. 2 seconds would be an unusually long delay to find in any replay chain or venue and late reflections generally don't have such a broad spectrum.

 

[Michaelgraf]

That's what I think.
This week I have more measurements in another venue. So it will be determined if the artifacts are depending on my measurement system or venue specific.
Thanks a lot for your help!