Nyquist says 1.5 Kits/sec. Need to be able to capture a high and a low to reproduce the original data.
Mic input likely expecting +/- 1V signal ("line level"), I would think take your digital signal, divide it down with a couple of resistors, and run thru a 10uF polarized cap to remove the DC component.