Normalize generator output?


I need to test microphones in the ultrasonic range (30kHz-40kHz), and besides the QA403, I have a nice microphone from GRAS that has a flat response in this range.

My plan is to get a tweeter that is capable of outputting sounds in this range, and while this is possible, the frequency response is not very flat.

So would it be possible to first measure the speaker with my “good” microphone, then alter the generator output, so the speaker has a more (apparent) flat response?

This way I could a more picture of the performance of the microphones I want to test?

Could “user provided” input to the signal generator solve this?

If that is the case, how do I use the featur?

if your aim is to see how the mic under test response differs from the reference mic response, you can save the reference response and use it as WEIGHTING USER 1; then measure the unknown mic and apply the weighting: this way you will see how the DUT response differs from the reference one.
Of course you should also consider that the 2 mics sensibility is different, but it is a solvable problem with some post processing.

User weighting functions was added in 1.197 Release Release 1.197 · QuantAsylum/QA40x · GitHub

And documentation is here User Windows · QuantAsylum/QA40x Wiki · GitHub

But I think you might need a calibrated microphone to do what you want to do. Measure with the unknown microphone, the calibrated microphone and use the calibration file to correct the measurement.

Sounds great, I will take a look at it!

I do have a calibrated microphone from GRAS, so I should be okay in that sense

With a calibrated microphone the speaker response will cancel if you divide the two measurements in the frequency domain. This way you don’t need to mess around with any advanced user feature. But it assumes you are happy doing data analysis; I think stuff can be done in Excel. I do quite like Python for this kind of stuff. Happy to give the analysis a go if you post the data.

Dan, dividing 2 equal response will give a flat 0dB line response; what I suggested was to apply the “weighting user 1” option to the DUT: it divides the 2 response, the DUT and the GRAS one. This way you get a RELATIVE response, simply and with no post processing needs.
I also suggested to do some more work to normalize the Gras and DUT response since the mics sensibility is different, and so the dB values seen in the graph. Here you need some post processing: I would scale the DUT response to match the Gras one, in the 35kHz range, since of the 30k-40k required range. Then apply the “weighting user 1” yo the scaled DUT response.

Hi Clane, I think your approach is a good idea, well worth doing; I meant to say that in my first post but when writing with a baby running around it’s easy to get distracted!

I’m quite interested in metrology so I’m coming at this from that perspective. I want to better understand the standard way of doing this measurement. I think it must involve a measurement with the calibrated mic, the uncalibrated mic and the calibration file?

It’s like doing a phono measurement: applying an anti-RIAA file in QA403, you get the phono response, that is how much the DUT response differs from the RIAA standard.
In Ehlert case, you have a Gras flat response, and a tweeter not flat response: we assume the total response as the reference one. Then you measure with the MIC DUT (instead of the Gras), being the tweeter the same: to this response you apply the weighting (the reference response). The obtained response is how much the MIC DUT differs from the reference response.

1 Like