Calibrating a microphone against a reference

sm52

Member
Thread Starter
Joined
Mar 14, 2019
Messages
863
I had a task to make a calibration file for another microphone. I measured the speaker with both microphones from the same position in space. Then he divided A into B and B into A. Looked at the tab all SPL how they look. The original graphs differ by about 9.6 dB. Therefore, the results of division are spaced relative to zero. Before smoothing, they are almost mirror image, but differ slightly in amplitude, up to 0.3 dB. After smoothing, they are not mirrored either in amplitude or in shape. Up to 200 Hz they are slightly different in amplitude, after 200 Hz they are very different.
1. Should they be mirrored with respect to zero?
2. What is the correct way to get a graph of the difference between two microphones to create a calibration file?
 

Attachments

  • microphone.mdat
    13.4 MB · Views: 16

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
Your starting measurements have far too many reflections. Move the source to the middle of the room and place the microphones near the source. A frequency dependent window can help eliminate HF reflections. Align the levels of the two measurements before dividing them. You don't appear to have a cal file for the reference mic, so the result will not necessarily be true, just the same as the unknown response of the reference.

Smoothing is done on squared magnitudes. Smoothing the reciprocals of a set of data is mathematically not the same as taking the reciprocal of the smoothed data.
 

sm52

Member
Thread Starter
Joined
Mar 14, 2019
Messages
863
It's clear. Is it correct to divide the reference by the unknown?

But the room is the same. The reflections are the same. Both responses must have the same set of reflections and useful signal. Is not it so?
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
No. Tiny differences in position and construction will affect the result, your A/B shows that clearly. Microphones have quite smooth responses, nothing like the result.

Divide the unknown mic by the reference.
 
Top Bottom