Export Impulse Response (as wav) seems inaccurate

[Yann]

When I export a measurement's impulse response as wav, the high frequencies (10kHz -- 20kHz) are quite heavily reduced (-10dB on the afore-mentioned range). This can be seen when re-importing the impulse response into REW.

It came to my attention by using the FIR filter calculator named porc. It uses the impulse response as input, and, because of the high-frequency dip, the correction is suddenly +10dB at 20kHz. Which is not desirable.

Is this a "natural" behavior of impulse responses exports that the FIR filter calculator should be aware of, or can this be corrected with respect to the REW export ?

As an example, please find attached some measurements showing this inaccuracy.

 

[John Mulcahy]

The impulse response export does not include the cal file, that is the primary difference. If you apply the cal file on the imported response you will see they align. The secondary difference is in the way the variable smoothing behaves at the highest frequencies, which is because the measurement stopped at 20 kHz whereas the imported IR extends to 22.05 kHz - the extra content in the imported response pulls down the smoothed result. It is typically best to make full range measurements across the whole span supported by the sample rate, 22.05 kHz in this case, in which case the two would be identical.

 

[Yann]

Good to know !

Thanks for the quick response. It has been said hundreds of times, but one more can not be too much : your software is excellent, and your (also excellent) support makes it even better. I wholeheartedly thank you !

Now that I know what to look for, I have found this old thread asking the same thing (link for convenience of further readers : http://www.hometheatershack.com/for...v-including-compensation-calibration-mic.html). Since the linked post is now 2 years old, is there now a way to export an impulse taking mic compensation into account ? I obtain the correct frequency response by calculating A / B with A=measurement and B=mic calibration, but the result has no impulse response.


On a side note, it would certainly be helpful for the ignorant enthusiasts such as I to have some sort of pop-up reminders of what is the recommended way among the overwhelming wealth of available controls. (like "Be aware that this export does not include calibration files", "Var smoothing is recommended prior to calculating EQ filters", "Measurements should be taken up to sample rate / 2")

 

[John Mulcahy]

I've added an item to the todo list to have an option to include the effect of cal files when exporting the IR.

 

[Yann]

Excellent