So, I'm looking at generating a polar plot to see directional dependence of a sensor that I've got. I downloaded a basic tool in MATLAB called dirplot, which I can read my measured data over 360 degrees into. I'm confused as to what frequency of source to use to, as whenever I've looked at manufacturer quoted polar plots some of them are just 250Hz - does that mean a source of a pure 250Hz sine tone? Some of them have lines for 250, 500, 1k, 2kHz... on the same plot - does that mean a source playing 1:1 octave banded white noise 0-250, 250-500, 500-1kHz...?
When you can see various frequencies in the polar plot, it means the measurement have been made several times with pure sinusoids at the given frequencies.
Notice that to measure frequency response of a microphone you need a calibrated loudspeaker (to be sure that the same acoustic pressure is delivered whatever the frequency) and usually an anechoic room (to simulate open space/field).
In addition to what @audionuma said, the frequencies you'd use will depend on the frequencies that you're interested/recording.
If I'm mainly recording speech, and somehow need to correct for mic directionality there - then I'd consider frequencies <= 8 kHz. If you're interested in sounds with higher frequencies, then the play ack frequencies will also need to be higher.
An alternative to using a frequency-reaponse corrected speaker workflow is to use a calibrated mic to measure playback levels of the speaker. The calibrated mic will provide playback level measurements of the uncorrected speaker. You also may not really need to correct your speaker as I find nowadays most good quality consumer grade speakers do have a flattish response till 15-20 kHz in my experience.