I was trying to fine tune the calibration on our machine today.

I thought I had it nearly right-on for the X and y axis, but have some confusing results.

I am using two techniques for the calibration - as I don't have a set of gauge blocks I am "roughing in" with big moves and measuring with a ruler (1/20th" scale - 50 mil/division) and then finally "fine tuning" with short moves (<= 1") using a dial indicator and magnetic base.

The problem I have is that the measurements between the two techniques do not agree.  I don't think it is a machine problem - I think it is my measurement technique.

I can get the calibration right on the money either way - I can set the counts per inch using the dial indicator for a 1" move +/- 1 mil - and this is for both directions - I will measure 1" and then do the opposite move and return exactly to 0 - +/- 1 mil for the opposite direction move.  I have the same results with both G0 and G1 moves.

If I go with the numbers I get from the 1" move with the dial indicator - I will have, as a example for an 8" move on the X-axis, a bit more than a 1/8" error.  Yet I can return to zero position where I started and I've got a zero on the dial - and a 1" move will still yield +/- 1 mil - either direction.

Here the travel is too far - and the cal value for the counts per inch is ~8100.  I am using velocity between 10 and 20, based on how the stepper motors respond.

So - using the "ruler technique" - with a "calibrated eyeball" I will perform the cal again - and come up with something like 7975 counts/inch.  Now the 8" move is nearly perfect (with my eyeball) - and also incremental 250 mil, 500 mil, and 1000 mil move are right on the money!

I certainly understand that my eyeball is not as accurate a a dial indicator - yet am perplexed by my results - as certainly I can discern two divisions of the ruler (100 mil) with my eye.

As I have the same results in both directions on both X and Y axis - I don't think this is a mechanical thing like backlash - I think I have a flaw in my measurement technique.

Tomorrow I will first get the machine cleaned and lubricated - as I have found that the proper lubrication makes a big difference.  I have been just spraying WD-40 on the crosslide and bed rails for the short term, but the manual recommends 10 weight machine oil.

I will get this machine as clean as a rifle in boot camp, oil it well according to the manual, and try again.

Is white lithium grease OK for the drive screws?  This adheres better and longer in my experience with engines.

You really only have to do it once in an accurate measure (dial indicator) make a move say 1" and then apply what teh dial indicator says it moved let mach do the math and reset the values. Retest to confirm it is correct.

IF the 1"move is correct to .001 then rest of it is correct as well as far as mach is concerned.

Each axis is done the same way.

NOW if your screw mapping is way out then that is another story

(;-) TP

Yet that IS the problem - I do the cal with the dial indicator for a 1" move and I can get it correct withing 1 mil.

Yet using the value I obtain from that cal (for the pulses per unit) a big move will be off.

Yet if I use the ruler and cal for a big move - small moves - across the full travel of the axis - seem to be correct - much better than the gross error when using the value obtained with the small move and dial indicator.

Are you doing the autocalibrate routine   What are your axis specs  motor,gearing feedscrew , microstepping ?

What settings are in the config setup ?

(;-) TP

Are you doing the autocalibrate routine Huh  What are your axis specs  motor,gearing feedscrew , microstepping Huh?

What settings are in the config setup ?

(;-) TP

Well - I did not measure all the teeth and gear ratios - I first used the autocal routine in Mach3, and using that zeroed in on the steps per unit, and fine tuned it with measurements.

Yet - as mentioned - the perplexing part of this is that I have different results using a dial indicator vs. a steel rule.

I went through this process again Friday - I will setup the dial indicator by loading it with about 10 mil - then zero the dial - and then make a 500 mil move.  Here I am using the motor cal value (counts per unit) I obtained with the "ruler method".

When I make the 500 mil move - it is right on the money - yet when I return to zero - the thing will be 10 mil off.

If I make a 1000 mil (1") move with the dial indicator - and using the autocal routine setup the counts/unit in the config - both positive and negative moves will be within 1 mil - i.e. the dial indicator moves 1" and properly returns to zero.  For example - for such a move on the X-axis the counts/unit will be ~8750.  

Yet - using this value in my motor tuning config results in a GROSS error across a 4" move on the same axis - over 1/8" (125 mil). I can see the error with my eye - I don't even need to measure it.

Using the "ruler method" with my calibrated "carpenters eyeball" I will come up with a count/unit of ~1790.  Using this value in the config, I can move in 1000 mil increments across the full 8" of x-axis travel and cannot discern an error on my ruler which has graduations of 1/20" (50 mil) (using my eye).

I have just bought a 24" machinist rule from McMaster-Carr which is graduated in 50ths and 100ths of an inch.  I will scribe a small reference line on the bed rails for the y-axis - and the crosslide rails for the x-axis - and use that to calibrate it.  As we are an electronics manufacturer - I also have access to binocular inspection scopes which I can setup plumb and level - and use that to see the results on the rule.

It is really not "fair" that I am asking all these questions here on the Mach3 forum - as the software works very well - is straightforward to use - and really delivers allot of "bang for the buck" IMHO.

I need to read some books - but for the meantime this job needs to be done by next Wednesday.....

I will go to CNCZONE and see if they have a thread for "People who want to be a machinist but don't know their ass from a hole in the ground".  

...I did not measure all the teeth and gear ratios.... ...I will go to CNCZONE....

Don't turn to the Dark Side!

Do you have a link to the machine supplier? The spec. page might give the required info. If you can, you should really try to calculate the figures rather than rely on an auto-routine. Start with the leadscrew on the axis. Is it metric or imperial? Make sure that you set Mach to the same units as your leadscrew. Make a mark on the dial and on the machine body. Measure how far it moves for ten turns of the screw. It _will_ be a sensible number. How then is the stepper motor connected to the screw? Is it through any kind of belt or geared reduction device? What size are the gears or pulleys? How many steps does your motor take for a full revolution (don't worry if you don't know - we can guess a sensible value, probably 200)? Does the stepper driver do microstepping? If so, how many microsteps (again, we can guess if you don't know)?

Set up a pen in the spindle and draw a line that is supposed to be 10" long on a piece of paper. Measure with you ruler.

I will assume that whatever units your axis is that is the units you are using for the axis calibration and also that the velocity and acceleration are reasonable for the moves you are doing.
It doesn't suprise me that when the axis is calibrated for  a long move and then checked on a short move  using an indicator that there are differences.
Let me clarify.......
With a good scale you can read quite accurately and the shorter move will be +- short move / by the long move. Realise that you are limited in small move accuracy of movement by your steps per unit.....1/steps per unit would be the best you can do if all was perfect, and, nothing is.  A number of things can cause the differences ie;
the screw is not linear, backlash ( it could be in the nut but in addition it could be in one or both  thrust bearings), other components of the driven train.
Cause of the differences become harder to identify as one tries to refine their steps per unit to very small values and at some point it becomes a rationalization
of what is correct.

For the most part the calculated value will be correct unless one knows the reasons they are changing it.

RICH

There is part of your answer,

" When I make the 500 mil move - it is right on the money - yet when I return to zero - the thing will be 10 mil off. "

Even IF your cals are wrong the machine should return BACK to zero every time dead on zero. 

You have either a drive problem,noise problem, binding axis stalling an axis or Vel/Accel too high for the conditions. The factthat yo have the same problem on X/Y indicates the Vel/accell setting may be TO HIGH for the motors you are running, OR the drive is not setup correctly and is loosing steps.

(;-) TP