First Laser Interferometer Measurements
Now that the mill has some basic functionality and is making parts, I wanted to try out the laser interferometer to see what information I can gather with it. First, and most obvious, is distance measurement. I set up the interferometer on the Y axis and commanded some moves between 0 and 20".
The accuracy was not bad. Instead of 20", the table had traveled about 19.996". And when moving back, it got to about 0.0006"
That is pretty good, and the numbers hold to withing 0.001" over several iterations. This is probably as good as I can get with open loop stepper motors because of their torque ripple. It may be a little worse when subjected to more weight on the table or with cutting load.
There is a lot of audible vibration. Especially in the 10-20 inch/minute range. More at some travel speeds than others, as is common with stepper motor systems. Here is frequency plot during a 20in/min move:
And here during a 10in/min move:
While they do 'sound' different the fundamental frequencies at which the machine vibrates are not any different.
What really interested me, however, is the magnitude of these vibrations. How do they compare the the cutting chip thickness? To do this, I logged distance data over the 20" move. I then cut off the beginning and end where the axis was accelerating. Then I performed a linear fit, and subtracted it from the data since this linear move is what it's vibrating around.
This is about 1 second of data and is representative. The x axis is sample number, of which there are about 600 per second. The y axis is in nanometers. Thus, there is a 2.5-3 Hz fundamental (off the scale in the frequency plot) with significant vibration at higher frequencies as well. The fundamental may, perhaps, be caused by uneven pulleys or binding in the screws. This accounts for positioning errors of up to 15 microns (0.0006"), but these are happening rather slowly from the point of view of the cutter. More relevant are the larger single-sample data swings. The largest of these seem to be about 5 microns (0.0002"). This can be a problem for smaller cutting bits. A 1/8" end mill may only take a 0.0005" cut in some materials.
So, in my mind there remained the question a to how much of the vibration measurement was real and how much was the machine moving, and so I went ahead and mounted the interferometer directly to the machine.
And it turned out that this exacerbated the vibration numbers. So clearly the steel cross member to which I mounted the interferometer vibrates more in relation to the moving table than the machine as a whole. In retrospect that makes a lot of sense since that member is only screwed to another member that itself is only screwed in a few places to the motor mounts. On the scale of microns, it's really easy to flex. The machine as a whole is far heavier than the moving table and will move a small fraction the table itself will move...
This would usually be the point to add some tuning in the motor drives or some filtering in the pulse source. However, as far as the drives are concerned, I am stuck at what I have. The geckodrives have only got a single potentiometer for tuning, and I tuned that as well as I could. The Leadshine drive I have is supposed to have some software tuning, but I have given up on trying to find the right software for it.
So, the straight forward answer is servos, which run essentially vibration free in my experience. But I am very curious what can be done with digital filtering. So I think I will try to switch from the smoothstepper to a LinuxCNC setup and see what can be done in that way before I move to servos.
I could have done much of this investigation by using a 1 micron glass scale (and I plan to do that for comparison purposes). But the laser interferometer can do measurements that the glass scale can't, such as angle, straightness, squareness and flatness. I just have to get the right optics....