CNC conversion

Can you convert your milling machine for CNC without replacing the leadscrew with ballscrew? Can the CNC compensate for the backlash? My understanding tells me no, but I see people on the web doing it. Can someone explain.

Albert

Two methods exist for reading the travel.

In one, the turns of the screw are read. Obviously, screw inaccuracy and backlash must be allowed for in this method.

In the other, some means to positively and directly read the movement is used. In that case, backlash is inside the loop, and is automatically compensated. The slide is moved until the reading is right.

There is a series in HSM from some time ago on conversions. It has started up again.
Check the website for village press for the index. (I know you have the URL, as you post on the BBS there)

Rotate,
It is done quite often, but my opinion is that you are much better off retrofitting a CNC with an outdated control than trying to start from scratch from a manual machine.
For what 70's and 80's CNC go for these days, you will probably approach the same dollar amount by spending many hours making brackets and enclosures and drive belts and such, and buying your steppers/servos..etc.
Older CNC machines are already set up with all of that, and are tried and tested in their original configurations, not to mention that in many cases the machine iteself is specially designed for production CNC work (box ways, larger knee, ballscrews, etc.)
There are alot less steps, parts to buy, and pitfalls in this route and the end result can often be much better and still cost less than retrofitting a manual machine to CNC.
It is almost certain that either way, you will not being creating a machine with linear encoders for the positional feedback so where the control thinks the machine's position is, is dependant solely on where it *told* the steppers/servos to be.
The backlash compensation is done by the simple math of adding the known backlash to one direction of travel in the control and is done automatically by most software, however, the least true backlash will provide a better result especially when interpolating axes for contouring (imagine the motions of X and Y to make a circle in real time and how they would be effected by the softare telling one axis to "wait" while the other takes up the backlash when it reverses direction and then apply that to a more complex contouring job like making figure 8's or carving a human face).
You could also equate it to trying to sign your name with a pencil that is half broken and doesnt follow your fingers until you've traveled a bit more then normal.

Sean

During the time that the direction of one of axis is being changed, the other is made to wait. I can see how the backlash can be compensated during this time, however, while the backlash is being taken in, the work in that direction is no longer being held rididly along that axis. I don't see how any type of motor control, even with DRO feedback can deal with this "loss" of control. May be if the cuts are extremely fine this is possible, as there wouldn't be too much force trying to move the work.

Let's say the backlash on the leadscrew is 0.002". I can see how even with this backlash you can cut with dimensional accurate far greater than that, however this machine would not be able to cut a sinusoidal curved who's amplitude let say is 0.001" (or at least not reliably), since during the directional change, the cutter can grab the work and pull it in the other direction by 0.002". Even if you had a closed looped control, I doubt it whether it would be fast enough to respond to such a suddent action.

Is my understanding correct?

Albert

That seems correct to me.
My reference to linear scales had to do with a control losing it's position due to just the forces you described. Once lost, I would think that this inaccuracy could carry throughout the program. At least with a linear scale, it could regain it's position.
So it's really two issues it seems, the tendancy to not be able to recover from say...a climb cut pulling the table to the other side of the screw and then having the control compensate in addition, and also the smoothness of curves while an axis waits for the other.
I'm just learning all this myself at the present time, but I can certainly see how getting close to zero backlash could effect CNC performance.

I quizzed some folks a while back on that, and ended up where you are too.

Theoretically, I suppose if the drive were fast-slewing, it could take out the backlash before the mass of the tool carrier moved.

Our CNC wood router is run just on commanded position, obviously for wood a couple thous is not a big issue.