Problem machining accurate size part on my Bridgeport CNC conversion

[fastturbovet]

I bought a Boss8 R2E3 that already had servos and ballscrews and we have upgraded this machine from the original control to Geckodrives and the Linux EMC software.

We ran the first parts today and I noticed when measuring the finished part it is off from the size listed in the CAD drawing. Its an egg shape that is about 1.5" long and 1" wide at its widest part - just a simple 2D part. The length is off by about .05" and the width is off about .04". The part is oversized by this amount. We were using a speed of 2500rpm and 6ipm feedrate with a 2-flute 1/2" endmill with 1/8" depth of cut per pass. We tried machining the part out of PVC plastic first, and initially thought the error could be caused by shrinkage of the plastic. But then we tried it out of 7075 Aluminum with the same results.

Another interesting thing was that the path the tool followed was measured to be .525 and not the .500 expected. Measuring the tool shows .499 which is pretty dead on, so how come the cut path is off?

Any ideas about what could be causing my parts to be off? I was cutting the Aluminum dry and squirting on WD-40 occasionally to help lube it a little. I dont have the coolant hooked up yet (will do that soon).

Help!

[DieGuy]

I am going to be watching this thread with baited breath. I am picking up my R2E3 on Friday and plan the same conversion. Were you able to reuse the power supplys for the Gecko?

What kind of motors does it have? the one marked 48vdc?

[ger21]

Is the gcode correct? Are you using G41/G42?

[Karl_T]

How much backlash have you got? This sounds like the problem. (I hope EMC has backlash comp) I'd suggest a very simple manual program to machine a square block. For example it you can find a piece of stock 1x1, put it in the vise and machine to 0.75x0.75. You'll be oversize by a certain amount (all machines have some backlash) Increase you backlash comp and try again on another 1x1 part. When you got it right try more complex shapes.

Karl

[ViperTX]

I would try just machining in one direction (x)...checking it....reversing direction and checking it again...if there is backlash it will show up there. Do the same for the Y. Of course you could just lock the table, using a dial indicator to measure table movement and manually attempt to move the table in each direction....that will also give you the backlash number, plus then you'll know where the problem is.

[fastturbovet]

FYI problem solved. It was a combination of things.


1) Backlash setting on the machine. The machine has .001 of measured backlash... we had the compensation set to 0. This was only a very small component of the error

2) We were using Visual Mill to generate the G-code. Some options were set wrong in software and it was using line-segments instead or arc g-codes

3) There somehow ended up some sort of rounding error in the g-code, so that a dimension spec'ed as 1.5000 would end up as 1.5003 or so... these errors seemed to add up a little and cause a bit of trouble.

4) The biggest issue was we had made an interface board between the EMC software and the Geckodrives (basically a parellel breakout box). There is/was a microcontroller in there which controls the coolant, spindle on/off, and the signals to the servos. Well, the BP has a LOT more signals going to it but we had used the code from our CNC router project, so there was a timing issue on this board causing it to occasionally lose its position and it would "drift" in the Y during the machining operation

5) The tool holder was not secured 100% every time and apparently was wiggling a little throwing off part dimensions.

After fixing the above, we machined the same part to the spec'ed size, and measuring with a vernier it is dead-nuts accurate to within 1/2 thou consistently. I am pretty happy with that considering its a 20 year old maching, with a home-made CNC conversion and we have little/no experience with machining/CNC.

Thanks!
Mike