I would carefully rethink what you're trying to do and why. This could end up costing you many times the value of your telescope.

Do you really want to digitise 30 seconds worth of 20kHz PWM signal (and to what resolution?)... and when you've got it, the BIG question is what value will it be to you? All you'll see is your motors start, speed up, slow down and stop. No big deal, you can see that happen already by watching the telescope move.

Let's say you're going to Digitise it to a pretty useless 2-bits of information over 30 seconds... let's do the math...

2 bits is actually 4 samples per 50uS PWM perid. One sample every 12.5uS... that's one full byte of data every 100uS, or 300kb of data over 30 seconds - 600kb for your two channels.

Now a PIC can grab that - no problem there, but disposing of 16 bits every 100uS is another story... 600kB of fast RAM? perhaps. DSPIC? perhaps... $1500 or more in parts, design and development time... almost certainly - and that's to gain almost no actual useful information that you can't already guess.

If you must throw money at studying motors spinning up and running down (I suppose it's the mech-techie version of watching paint dry), then an off-the-shelf data-logging card for yor PC is perhaps a better (and cheaper) way to go...