I mounted the generator more solidly, and also made a post where I mounted the miniature bayonet base lamp (in an old military style lampholder), some binding posts to read the generator output voltage, and an inductive proximity sensor. I also connected it to the TigTac prototype and ran into some problems, as discussed in a
HSM thread, but it now works reasonably well. Here is a copy of the latest post in that thread, showing the (almost) final set-up, with some video clips of the engine running on compressed air and lighting the 12V lamp, and running as a compressor by applying 20-25 VDC to the motor:
I added a 1k resistor from the sensor input to the PIC pin being used, and a 220 nF capacitor to ground. This seems to have eliminated the noise that previously caused erratic readings. The original code used an interrupt-on-change (IOC) method to detect contact close and open events, but that was very sensitive to noise, and difficult to implement a software filter. Now I am polling the contact state every 1 mSec in an ISR and requiring 3 consecutive stable readings to determine status. Since there is equal delay on sensing proper open and closed conditions, the accuracy of the dwell and RPM (based on time between closings) is only affected by the 1 mSec uncertainty of the sample. Thus a modest 600 RPM corresponds to 100 mSec/rev and accuracy is about 1%. Here are some pictures of the setup:
Some video clips of the operation and waveforms:
http://enginuitysystems.com/files/TigTac/TigTac_1908.AVI (111 MB)
http://enginuitysystems.com/files/TigTac/TigTac_1909.AVI (147 MB)
http://enginuitysystems.com/files/TigTac/TigTac_1910.AVI (160 MB)
and a video clip of the motor/generator running on compressed air (warning - loud compressor noise!):
http://enginuitysystems.com/files/CAMM112/Air_Motor_Generator_1900.AVI (170 MB)