0
\$\begingroup\$

It is typical for a microphone to have a non-flat frequency response like so:

Frequency Response

Is it typical to flatten the response curve by weighting the signal in frequency domain to get the most accurate representation of measured sound?

If not, then why?

If yes, then why don't manufacturer ever provide the response curve in numerical form?

\$\endgroup\$
1
\$\begingroup\$

It is typical for a microphone to have a non-flat frequency response like so:

Yes, it is typical for microphones to not have a flat response curve. It is really hard to manufacture them in a really dead flat way. Even measurement microphones can look like this:

enter image description here

which while much better than the example you gave, is far from dead flat in the higher frequency ends.

Is it typical to flatten the response curve by weighting the signal in frequency domain to get the most accurate representation of measured sound?

That depends on the field you do your stuff in. For the most part it is all about subjective quality of sound. If you want to do real number measurements, then they do it, but they start with almost flat responses of the mic anyways.

If yes, then why don't manufacturer ever provide the response curve in numerical form?

What numerical form do you want to have? What single number represents the whole graph? Essentially none is capable of doing so. The best you get is some frequency range which says "Between 20Hz and 14kHz we are not entirely totally crap". Which is why good manufacturers provide these graphs, and when you buy mics without that graph you don't care about it anyway.

\$\endgroup\$
  • \$\begingroup\$ 'If you want to do real number measurements, then they do it, but they start with almost flat responses of the mic anyways.' Isn't it cheaper to get a mic with non-flat curve and fix it later in DSP? What's the catch here? 'What numerical form do you want to have?' Oh I just meant response curve as a table with columns for frequency and dB. \$\endgroup\$ – andrey g Apr 14 '16 at 13:19
  • \$\begingroup\$ The catch is signal to noise ratio. Equalization might do wonders in flattening the response but can not improve SNR. Think of a corner case where in a certain, narrow bandwidth your mic gains -80dB. What's gonna happen if you equalize it back to 0dB? \$\endgroup\$ – Vladimir Cravero Apr 14 '16 at 13:42
  • \$\begingroup\$ @VladimirCravero: That is quite orthogonal to the response curve issue and in case of scientific measurements a given, considering the need to make the same considerations for all the equipment that follows the microphone (amps, adc etc.). When it comes to the musical side however if after their 30 channel equalizer it sounds good, then it sounds good, no matter the objective SNR for a specific frequency. And the corner case of 80dB is pretty contrived given that even the bad microphones don't often go beyond 40dB total range, and if you care about equalization you would not have bought that. \$\endgroup\$ – PlasmaHH Apr 14 '16 at 13:49
  • \$\begingroup\$ @PlasmaHH I agree, but OP asked for the catch and I thought we were speaking of measurement equipment, not music applications... What's the catch you have in mind? \$\endgroup\$ – Vladimir Cravero Apr 14 '16 at 13:52
  • 1
    \$\begingroup\$ @VladimirCravero: this is an interesting field on its own and can be done with just microphones without any references: en.m.wikipedia.org/wiki/Measurement_microphone_calibration \$\endgroup\$ – PlasmaHH Apr 14 '16 at 19:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.