HP5508A interferometer replacement hardware - first test
Ok, so it is time to use the interferometer for some real world test. I have a heavily modified Roland PNC-300 CNC machine. The parts it made were not 100% accurate in the x and y dimension and I wanted to know if the CAM program was to blame, or the machine, or what.
The way this machine works is that the computer sends its data via the serial port in a proprietary format called CAMM-GL III. Then the microprocessor in the PNC-300 computes trajectory data and cutting instructions from that.
Originally, stepper motors were driven via MTD2001 ICs. But I removed the steppers, intercepted the signals and fed them to a STM32 based board. This board processes both limit switch and stepper signals and feeds them to the microcontroller and servo driver boards, respectively.
For the Z axis, I added two Yaskawa driven servo motor ball screws
The servo boards are Granite Devices IONI drives controlling two linear servo motors with 0.5 micron glass scale feedback encoders.
So, technically, the positioning should be very,very accurate, but that assumes that the factory supplied microcontroller makes accurate calculations and puts the right number of pulses to the (now nonexistent) steppers.
So to test this, I put a mirror on the y-axis and moved it. Then I compared the amount measured by the interferometer to the amount the axis was supposed to move. I had to do this more than once. It turned out that
- the mirror absolutely has to be bolted down, as the vibration in the axis is enough to cause slippage
- it takes a few minutes for everything to stabilize (I suppose thermally, motors warming up etc...)
anyway, after a while the measurements became very consistent. The units are in mm.
So about 5 micron deviation over 8 mm distance. I should mention that I had to sit very still because the CNC and Laser were not connected via a stiff mechanism (Wood floor/carpet). If I leaned to the side, the floor flexed enough to mess up the measurement.
The original CNC electronics claim to step in 0.01 mm steps, and I was able to verify that to my satisfaction. Here is a test cut:
That tail fin is 0.2 mm wide! The model was cut with a 2mm square cutter. No ball mill finishing. The stock spindle in the PNC-300 uses preloaded deep groove ball bearings with the preload supplied via a coil spring! This causes chatter in aluminum cuts and really precludes cutting steel (unless using a 0.5 mm cutter), so a spindle upgrade is probably the next item on the list...