Frequency response and test methods

[davidm]

From the archives (thanks, Harvey), I gather that a "good" FR is more about aesthetically matching a mic to the source/room/etc than finding a “holy grail” frequency response shape.

I'm wondering how much the FR graph changes when a mic is fed a complex signal like music, compared to a test tone signal. Somewhere I read that FR tests may use a single varying fundamental, white noise, or... -- how much do these choices affect the graph? I imagine the single fundamental method would blow up all the resonances, while white noise would smooth them out. Complexity-wise, music is in between a single fundamental and white noise, so not sure if either is an appropriate test signal. Is there a best practice?

I'm guessing the answers start with "It depends..."

Thanks, David

 

[Harvey Gerst]

Pretty much the standard way to test mics was to aim them at a sound source, along with a high quality reference mic, like a B&K measurement mic. How much the test mic deviated from the B&K reference mic was the test mic's response curve. To see all the resonances, sine or warble tones were used, at a slow chart/ fast pen speed. To get a more averaged response, you used a faster chart speed and a slower pen speed.

 

[mshilarious]

Somewhere I read that FR tests may use a single varying fundamental, white noise, or... -- how much do these choices affect the graph? I imagine the single fundamental method would blow up all the resonances, while white noise would smooth them out.

They really don't, you get about the same response either way. I find that it's much more efficient to use white noise, because the frequency content is a constant, so you see more detail in a shorter amount of time versus using a frequency sweep. Of course that won't work if you are using a plotter, but with FFT there is little difference.

Either way, the actual response of the mic is much more detailed than what you see in the average specification.

 

[davidm]

Thanks for the helpful responses -- very interesting, esp. that the mic response is ~independent of source. Thanks again, David

 

[ofajen]

You should probably also be aware that there is more to mike characterization than steady state frequency response curves and on-axis response. Response to transients and off-axis coloration are also critical to the sound.

Cheers,

Otto

 

[The Axis]

-- very interesting, esp. that the mic response is ~independent of source.

The steady state frequency response graph will (theoretically) look the same whether measured with swept sine waves or random (white/pink) noise.

Two mikes with apparently similar steady-state responses can have significantly different TRANSIENT responses though.

...and the above comment about on/off axis response is very true.

Science really does work, mikes can be compared by looking at response plots, and they will sound pretty similar to what you would expect from looking at the curves. But you have to know what the curves really mean and have complete information.

As a practical matter, we usually don't have complete information, or access to an anechoic test chamber and instruments, so it is easier to listen and make a judgement call.