About Saturday, I started feeling better, but a little dopey.

What Actually Happened:

Saturday:

        I started looking at the output from previous runs and started debugging.     The following improved the results:box
  1. Split the scan and the data plotting into two separate programs.    One reads the log from the other.
  2. Calibrate the machine -- It is 4000 counts per revolution, and about 500 counts per inch of the analog sensor
  3. Verified the linearity of the sensor (1%).    Looked at repeated scans to confirm no drift.   None found, highly repeatable.    Read error typically 5-15 counts per 1000 units.
  4. Play with step speed -- found bugs in the stepping routines that were interfering with the command parser.   Higher speed seemed like smoother output, but may be electrical noise issue.     The 5+ from the USB appeared to be wandering a bit.supply   
  5. Change scale of input, and use internal reference. 
  6. Verified graphs against raw data.
  7. box_blockbox_block_rawbox and block
  8. Still wasn't quite satisfied that the output was useful -- stuff looked a little too wrong, but measurements from the graph seemed surprisingly accurate.

Sunday:

  1. Realized that the graph tool was auto-scaling.    This made any errors and distortions look worse than they otherwise might be.
  2. Removed bad data, and understand limits.pyrmid_imagepyrmid_rawpyrmid_clean
  3. Decided to go with 3D and see how far I could get.
  4. Stripped Z slide from old semiconductor wafer robot.donor_z
  5. Researched servos, encoders and realized I would need to make another board, or swap to steppers.
  6. Swapped the drive to a stepper.
  7. Mounted Z on the Theta table.
  8. Put a foot on it.    Feet on the feet save on crushed fingers.
  9. Packaged the electronics on it.scanner
  10. Broke a driver, milled into package and reattached the connector.broke driverdriver mill
  11. Orientation of the sensor matters.
  12. Blew all of the stepper drivers at about 2:00 am.
  13. Figured out that there is something wrong with the driver board I built -- short?
  14. Debugged low torque, and programmed up the 8-fet board from before to act as a unipolar step/direction interface.   drivers
  15. Tested, found low torque problem was the test motor I was using, not the algorithm.
  16. Integrated to USB driver i/f board, remounted.
  17. Passed out.

Results thus far:

  1. The sensor and table are accurate and repeatable.
  2. The sensor does not like black and/or reflective things.
  3. Currently, point mapping beyond the center is not correct.
  4. There may be some linearity problems with straight lines.  

Will it ever be 3D.... let's see how much code gets written between 9:30 and this presentation!    (Probably not much)   Update: None, just this report!