From the moment I received the celestron motor drive I was slightly disappointed. I immediately had to file out the case so that the switches could swing full from side to side, and as soon as it powered up and I realized it was a simple DC motor with a dimmer...well, I knew it would mean more field calibration. And calibration there is no way I could make perfect without spending a lot of time watching the camera drift, wasting precious battery and dark time.
I'm a motion control guy. This doesn't seem like a difficult thing. The earth turns at a very predictable rate. A sidereal year is 365 days, 6 hours, 9 minutes, and 9.54 seconds (if we're being overly accurate.) That means there are 31,558,149.54 seconds a year, therefore it takes 86,164.09 seconds for the sky to come full circle once. It would be easy and cheap to use something that is positionally accurate, gear it a bit, and make a controller... -a real controller- that tracks properly and without need of calibration for my purposes.
So I broke out a classic: the arduino clone known as the boarduino, of which I don't bother with a lot of the components but make use of their lovely printed boards (Thanks Adafruit!). In total it cost me about $7 in components to build the controller. Ebay was also full of miniature geared stepper motors (85:1 gearing, 5v native, etc, etc). If you have the patience to get one from Hong Kong it will cost just a couple of dollars. If not (I'm running out of time) it will cost you $6-12.
After some creative programming I calculated the potential accuracy of my setup. The oscillator isn't perfect, and neither is my code, but in total the maximum error seems to be around the equivalent of 5.6 seconds of missing sky per hour. This would just barely show up at my longest lens if I were taking hour long exposures. I don't honestly see myself taking single frames anywhere near that kind of time. Perhaps a couple minutes per exposure at best. Chances of me hand calibrating a DC motor to this level of accuracy? Nill. If I thought it needed to be more accurate I could adjust for the oscillator and coding errors- but who cares?
Now to make a mount for it...