What is the best way to verify accuracy of machine

[TTalma]

When I set up my machine I calculated the steps per inch at 2000. I cut a 5" square 1/4" deep in MDF and measured it with my calipers. it was a little small, I ran the numbers and came up with 2022 steps per inch for my X, and 2030 steps for my Y. Repeated the test and my calipers showed it to be right on at 5".

Now I am cutting a part to hold a small servo (the type used in RC planes). The servo measured 39.6mm with my calipers, so I put a 40mm opening in my model. When I went to put the servo in it was tight, when I measured the opening it came out to be 39.5mm, so .5mm smaller than desired. a few thousands I can live with but .5mm seems like a lot.

So I'm not sure what's going on, I am assuming it's probably due to backlash, or not being set up correctly, or some combination of both.

What I would like to know is there a good procedure for measuring and setting up to get the best settings for the machine?

A few more details that may or may not matter.
- X and Y are rack and pinion, Z is ball screw
- the max travel is 42 x 42 inches
- testing was done with a 1/4" 2 flute straight bit
- the part model was made in fusion 360
- the part model was cut with 1/8" bit in plywood
- There was a finishing pass of .004 mm on the part
- cut depth on part was max 1/8"
- the part was cut with 3D contouring
- there was a .2" silhouette around the part.

Similar Threads:

• Newbie- New User – ”No data to verify” error when trying to verify
• Accuracy of an inch machine
• Need Help!- Bobcad run fine when I verify but can't run in machine

[BalloonEngineer]

You probably should have different steps/inch on Z than for X and Y. This is a number that is calculated based on the hardware, not something you “adjust”. If you need to adjust this, you have problems with losing steps, backlash, bit not being it’s nominal size, or some other mechanical issue (loose couplers,etc.).

You are always looking for steps/unit so determine how many steps in one rotation of the stepper divided by how far it moves in one rotation. For example, on my machine I am going to replace a 1605 ballscrew with a 1610 (16mm dia x 10mm per rotation). So for me (200 steps per rotation x 10 microsteps) / 10mm = 200 steps/mm. Actually for my machine I work in inches, so it’s 5080 steps/in, but the math is the same.

[ger21]

Did you climb cut the hole?

Cut two 5" squares.
One climb cut, and one conventional cut.
See how much difference you see.

[Biggs427]

Are you sure that your tool is really 0.125? I bought a Freud 0.250 cutter but the cutter was in fact 0.239.

If you have access to a dial indicator that might help. There's lot of variables involved when you measure the parts. If you measure travel with a dial indicator you remove tool diameter and tool deflection and machine flex from the equation.

[wizard]

There are many things that can impact a parts final dimensions. If you are machining plastics for example they will at various times under cut or over cut depending on the specific material. Delrin for example drill or reamed in a lathe will often come out undersized, literally smaller than the tool that drilled the hole. Even steels can move over time as interanal stress relieve. So if you are attempting to get performance data by inspecting machined parts make sure the material is stable.

Second it doesnt pay to look at calibration with respect to machined parts unless you know the cutters actual diameter and your spindle run out. In otherwords dont assume the machine is off grossly if you dont know every thing about the setup.

Backlash is a real issue and likewise you need to know how your machine performs for each axis. Fortunately this is easy to mearure on most router type machines. You will never truely hit zero backlash but it needs to be very low to achieve tight tolerances and good finish. In any event you will want to document backlash for every axis and set the correct parameters in your CNC controler.

The gantry itself can be a huge issie with respect to holding dimensions as they can be subject to twists and tocking motions. In my opinion the gantry is often the weak link in a DIY router build. You will need to verify that you have acceptable Y axis stiffness to hold the tolerances you want. Some gantry issues can be likened to leadscrew backlash in the clearances in the bearings allow twisting and rocking. At other times you can get this unwanted deflection from poor mechanical engineering. In any event you really need to understand how much your gantry is contributing towards inaccuracy in your final part.

As for backlash that is fairly simple to measure with some sort of dial indicator on the X and Y. Basicall put the controller in a fine jog mode, move the axis under test a good 20 mm zero ot a dial indicator and the reverse you jog direction and very finely jog until you see the gage move. Add up the number of jog increments taken or read the delta seen in the gui and you will have your backlash number for that axis. If it is excessive you will need to correct the mechanical issue creating the backlash and test again.

There is one gotcha when testing for backlash, you can have conditions where friction can mask backlash so that you dont see it under light loads. This can be caused by a shaft fitting tight on a bearing for example and only moving under load or after heating up. So dont test a cold machine, actually warm it up for a good half hour. It doesnt hurt to check the leadscrew bearings for proper installation too.

As mentioned above, once you have an accurate value for steps per mm or whatever your controller requires this value should never require changing. Once calibrated the only thing that will impact this is thermal growth during actual machining. This can be a nasty surprise if you are tring to hold very tight tolerances. How you mitigate the thermal groth issue is really beyound this discussion and in any event does not appear to be the problem here.

[BalloonEngineer]

If you calculated mechanically it should take 2000 steps/inch then changed it to 2022/2030, parts should have come oversized. (Would have commanded more steps than necessary to move any set distance) . Also note that your .004mm finishing path is well under a half a “step” so no material was removed. Set your steps/in to the numbers based on the mechanical design and run the tests suggested by Wizard. That is going to reveal the system accuracy. You can’t adjust the steps/in to remove the error. Make sure that in working between inch and metric sizes you don’t get confused. Just because your displays read to 4 decimal places, does not mean your system can actually resolve to that accuracy.

[TTalma]

I managed to run some tests on my machine last night. My results are:

All dimensions in inches.

Spindle run out: .0002" (a perske vs 50.09-2, tested with 3 collet sizes, 3 different router bits per collet, 9 tests in total) results very consistant
Bit used for square cutting tests, new .25" Onsrud 2 flute spiral bit, flute length 1". actual dimension .246"
Cutting 5" squares, bit protruding 1.25",

I cut nine 5" squares 1/8" deep in a piece of 40" square MDF, cutting like a tic-tac-toe board. One square 1" from each corner, the others centered in between the climb cut on one side of board, standard on the other side.
results left to right top to bottom

Climb cutting:
X: 4.983, 4.982, 4.983, 4.983, 4.984, 4.983, 4.983, 4.984, 4.984 - Average 4.983
Y: 4.971, 4.972, 4.973, 4.970, 4.970, 4.971, 4.971, 4.971, 4.971 - Average 4.971

Traditional Cut:
X: 5.011, 5.010, 5.010, 5.011, 5.011, 5.010, 5.011, 5.011, 5.011 - Average 5.011
Y: 5.000, 5.001, 5.000, 4.999, 5.001, 5.000, 5.000, 5.001, 5.000 - Average 5.000

So now that I have this info what if anything should I adjust? It seems to me that I am right on for a traditional cut, but averaging the traditional and climb makes them a little smaller.

I understand what BallonEngineer is saying, and in an ideal world I could just punch in the math, and set my pulses per inch and be good to go. But if the rack and pinion each have a manufacturing tolerance Of +/-.002". Making a worst case of a .004" error per inch, over 40" this is .16" more than 1/8". By adjusting the pulses per inch I can overcome the manufacturing tolerance.

Backlash bit deflection etc. is harder to overcome I'm not looking for perfection, my goal is .005' or better tolerance. And as I work with the machine I will learn it's nuances, so I will know what to do to get better results. But I want to get the set-up as best as I can from the start.

[ger21]

It would appear that you are measuring the hole, rather that the square part you'd be cutting. When cutting actual parts, clumb cutting would be larger.

Personally, I would set the steps per unit so that the largest parts you cut are accurately sized.
If you adjust your steps/unit for a 5" part, a 40" part could be off a large amount.

If you need very accurate parts, I'd probably adjust the tool diameter in the CAM software.

[fretman_2]

For the X and Y, would using a pen and measuring the drawn lines be just as accurate as cutting? I drew a circle within a square and measured that.

[TTalma]

I was measuring the part. The command was
G01 X5.25
G01 Y5.25 ...

I choose a 5" square because I only have 6" calipers

Should I bother adjusting anything, I'll be making parts of all different sizes?

99% of what I do will be wood, and I know wood can move a few thousands overnight, so even if I had perfect cuts every time in every direction, wood still moves.

The only time I would need a high degree of accuracy would be when I do an inlay.

[BalloonEngineer]

For the X and Y, would using a pen and measuring the drawn lines be just as accurate as cutting? I drew a circle within a square and measured that.

That will represent the no load best case, but not the loads due to cutting, bit deflection, etc. If the structure is not stiff enough may even get gyroscopic loads fron spindle running.

[wizard]

Most likeky you have more going on here than a calibration ptoblem. To that end id check backlash on X & Y and machine stiffness. You wont accomplish much calibrating the leadscrew pitch with bad data resulting from backlash and deflection.

The climb cut values are telling me that the cutter is able to pull or deflect something easily be 0.015" to 0.020" fairly easily.

This could be simple rigidity issues or it could be backlash. What you might want to do is get something like a fishermans scale to apple repeatable pressure to an axis to look for defection and other issues. Ideally the scale would go up to 100 pounds. 50 pounds of force would be a good starting point.

Also what was your depth of cut. Dont forget that cutters deflect too. Ideally you would use the biggest cutter that the collet system supports. Even then you are at risk of some deflection. Also by varying the depth of cut you may end up seeing different error values due differing loads.

Edit
Pictures of your axis construction might help. Also rack and pinion can have significant backlash. You will want to make sure your pinion engagement is perfect. Further if the pinion is spring loaded you want to make sure that the spring isnt moving under normal loads.

Im still leaning towards a rigidity / backlash issue here.

[fretman_2]

Ahh...gyroscopic loads...haven't thought of that, but it makes sense.

That will represent the no load best case, but not the loads due to cutting, bit deflection, etc. If the structure is not stiff enough may even get gyroscopic loads fron spindle running.

[skrubol]

So your test cuts were slot cuts, just cutting a square, not making pockets with islands, right? So the "climb" cuts were just clockwise and the conventional cuts were CCW?

[TTalma]

Wizard: I did some backlash and load measurements last night. My dial indicator only has about .25" of movement. My test:
Move X +10", Set up dial indicator against shaft of bit in spindle, Use mach 4 to command .2" movement in neg direction, I did this 4 times eyeballing the spacing across the 40" of travel. Then repeated test in opposite direction, movement measured was .1970", or .1979" each time.
Repeated this test with Y, measurement was .1983", .1984", or .1986" each time.

I was amazed it was this good!

to measure deflection I used a fish scale that went up to 100 lbs. For X I hooked it to a spot on the Z carriage, near the bottom, about where the spindle collet is, and slowly pulled it towards me, Had the indicator set to display peak value. Found that I was overcoming the holding force for X around 80 lbs. So I only tried going up to ~70, max deflection with Z elevated was .0026 With Z about 1.5" from bottom (as low as I could go and still get my dial indicator set up) was .0039".

Repeated for Y with scale hooked about center of Z carriage, and Z carriage as close to center as possible , Did not overcome holding force, at 100 lbs max deflection was .0064, with Z lowered.

It's a pretty robust machine I attached a picture of one from the Internet. Mine is different in electronics only, I stripped all the AXYZ controls, and replaced with smooth stepper and Mach 4.

Skrubol: Yes basically just 4 lines making a box, no pocking operations or anything like that, running CW and CCW. Cutting 1/8" deep