I made a switchable white/pink noise calibrator back in college (professor's project, so he gave it to me
). It was ridiculously hard to do, not because the generation of the voltage signal was hard (that was easy though a little tricky at very low frequencies), but because creating and measuring the response of a broadband emitter was a royal PITA. I cannot remember exactly what I ended up using, one was some kind of array of small piezo or compression drivers I think in a small box for large mics, and a single compression driver EQ'd (all analog back then, long before DSP chips were common) in a tube(no horn) for pencil mics.
Virtually the only mic calibrators I have seen in recent days are almost all 1 kHz, with some having a 250 Hz option. The only broadband system I found some years back was way out of my price range (thousands of dollars USD). You can get a cheap one for around $100 and a NIST-trace version for around $500 IIRC.
Edit: Same prof had me create a light box the year before, another project that seemed easy but turned out to be incredibly difficult. I learned about Fresnel lenses and a whole lot about how hard it was to get perfectly even, perfectly white light across a 24" x 24" area. You'd think I would have learned to say "no" after that, but he was paying the bills, and it got me an office and lab space in the EE building.