Accuracy of my mill

[Terry G]

I've been machining for about 5 years now with my CNC taig mill.

Recently I've noticed that when I am cutting out circular patterns, the diameter does not seem to be very accurate. For example, I just made a #35 24T sprocket out of 1/8" aluminum. I used one of the autocad scripts, and verified in detail all the dimensions to make sure they were correct. I put the proper taper on the teeth (not really needed for what I'm doing, but good learning experience), and gave it a 3/4" in bore.

When I finished, I cleaned up the part and tired a new #35 chain on it. The chain went about 2/3 of the way around, then would not mesh properly anymore. It was as if the sprocket was too big (or teeth spaced a little to far apart). I double checked all my autocad dimensions, and they were perfect.

So I started inspecting my part per my drawing, and immediately noticed a couple of things. First of all, when I measured from one tooth valley, direclty across to the opposite tooth valley, the diameter varied from about 2.676" to 2.685". The drawing called out 2.672".

I also measured the inside 3/4" bore I cut out using a 3/8" end mill with a .003" finish pass, and the inside bore varied from .738" to .742" instead of .75" per the drawing and g-code.

I have also seen this on other round holes I have cut out, where the diameter tends to be less than the drawing. I've checked my end mills, and they are all to spec, and I even went through my g-code and verified that all the numbers are correct.

As far as the mill, I took the entire thing apart about a week ago, cleaned all the ways, adjusted all the gibs perfectly, and got the backlash down to .002" in the Y direction, and .0015" in the X direction. I measured this using my test indicator on the end of the table and jogging it back and forth .001" at a time. My table flatness was .0005", and I trammed the column to around .002".

And of course, I am getting the same kind of results before and after I went though the tear down. Also to note, when I machine square/rectangular parts with pockets, cutouts, etc.. they come out to within +/- .001".

I'm at a loss, any ideas of what else I can look at to help eliminate this problem? (Besides a new mill!!)

Thanks.

[WayneHill]

Terry,

Try leaving more stock for the mill to cut on the finish pass. Slow the feed down on the finish pass. A dull end mill will bounce off the material instead of cut it.

Good luck,
Wayne

[Terry G]

I typically leave about .003 to .005 for my finish passes. But I can try leaving a bit more. I'll slow down the cutter a bit also.

My end mills are pretty sharp, and the finish looks really smooth.

I'm starting to suspect my controller software. I haven't updated it since I purchased the mill 4 or 5 years ago. I figured it is working great, so why mess with it. But there have been many new versions that have come out with the latest being released just a few months ago. I'll give that a shot and cut some circles with your recommendations and see what happens.

Thanks.

[tnik]

I normally leave (depending on the material, but not much) .01 to .015 for a finish path on a profile. For production jobs, I always use a rougher endmill and a finisher.

speeds and feeds have alot to do with it to.. also the length of the tooling.. there's alot of things that could be giving you the results that your getting.

[cam1]

Hi: Why not verify that the axes are moving to their commanded position. I'd recommend placing a pin in the spindle and an indicator on the table, and writing a simple program to advance to and retreat from the indicator. Sounds to me like you have a problem with losing steps...

cheers

[Terry G]

I have a closed loop servo setup. The axis are moving to their positions when doing simple x back and forth and y back and forth movements. Parts that are not round don't have this problem.

The problem is only with circles, and I am beginning to think it is just with arcs. I'm going to test this by making a circle with a lot of straight lines, then doing the same circle with an arc. I have some fine tipped drag engraver bits I can maybe use in some plastic.

I may be going down the wrong path, but I'll soon find out.

One more thing I just remembered, I was machining some plastic enclosures a couple of weeks ago that needed a small pocket, and 6 .125" holes. I was using a 1/16" end mill for the rectangular pocket, which came out perfect. Instead of changing my tool, I just ran the 1/16" bit for the holes (I had to do 80 of these). Well to my surprise, on the first box the holes came out .115 diameter. I had to increase my hole size to .135 to get the .125 I was after.

Anyway, I'll do some testing tonight.

Thanks for the help.

[Halfnutz]

Unfortunately, "closed loop" step and direction servo's can be capable of missing steps just like stepper motors. The drive is closed loop but the nature of a step and direction control program is normally (there are of course exceptions) to send out pulses and "assume" that the drive recognizes every pulse. If the signal is week or dropping out the drive may not see every pulse or recognize every pulse edge that it is assumed to. If there is an intermittent connection or a low voltage (like during circular interpolation when two axis are being driven simultaneously) it is possible that you could be missing steps. Just a possibility to explore.

[WayneHill]

Terry,

How many decimal places are you using on your g-code? Four places would be best. If you are using three places, the error could be in how the control handles the arcs.

[tikka308]

I agree with cam1 - put a dial indicator on the table and see if that's the problem first.

[Terry G]

Thanks for all the suggestions. I did double check my positioning and indeed my table always returns to the same spot. I am using 4 decimal places for all my numbers.

I did some more testing today. First I drew a 1" diameter circle, then I drew a second circle but with straight lines every 5 degrees. So the gcode for the first circle was a G03 command, and the second circle were many G01 commands. I double checked the gcode and the numbers were all correct (I have never had a problem with this).

I cut both the circles in .125" 6061 aluminum using a new 3/8" 2-flute EM for aluminum (double checked it measured exactly .375"). I did a rough cut (leaving .010" for finish) at a feed of .002 (2500rpm, 10ipm). Then I did the finish cut at a feed of .001. I used arcs for entry and exit. The finish was perfectly smooth.

Both circles came out identical right to 0.99". I measured this with my calipers, then I took out my b&s hole gauges and checked them with my micrometer. I consistantly measured 0.99". Also, these holes are very circular. I get the same .99" anywhere I measure around the circles.

Then I spent the time and upgraded my controller software to the latest version. I took another piece of aluminum, ran the exact same program, and once again I measure the 2 circles at .99". That just proved it was not in the sofware, and had nothing to do with how it deals with arcs.

I also wanted to mention once again that I cut some .125" holes with a .0625" end mill in .06" thick plastic. And it was exactly the same thing. The holes came out to .115". I re-drew the holes to a diameter of .135", and they came to .125". I was cutting these holes at a feed of .00075 (10,000rpm, 15ipm).

If there is a problem with my servos missing steps, is it possible that the problem can be so consistant. It doesn't matter what size end mill I use, what material I cut in, or what size hole I cut, it always comes out .010" to small, and a really true circle.

Even if it was my backlash, how could it always be so consistant and true?

I'm open to any other suggestions anyone might have. At this point, I am pretty much at a loss. I'm starting to think it is something simple I am overlooking. The problem is easily repeatable and very consistant.

Thanks!

[Halfnutz]

Usually the only consistant thing about missing steps is that it is inconsistant. If you are duplicating the results then it probably isn't missing steps, but may be an offset or a setting somewhere that you haven't found. I deal with these types of problems every day and they are frustrating to say the least. Put it down for a while if you can afford to and take a fresh look at it in a couple of days. It sounds possibly like some thing to do with resolution, something is looking at 4 places and some other part of the software is dropping or picking up a decimal place somewhere.

[CNC_BOB]

Terry,
I do not know anything about "Taig" mills so bear with me, I know you said it is closed loop but are there glass scales that talk to the encoder to makeup for lost motion? We had a company bring out a "Renishaw ball bar" which is a simple tool that when hooked up to your machine and you do some simulated circle mill routines can give a computer hardcopy report of accuracy. Second, Programming in metric mode is always more accurate. Last but not least, looking at your part it looks like quit a bit of material is removed, are you roughing everything to let the material relax, re-clamping and then finishing? or do you just do it all in one shot? hope this helps