[idtkid]

I have a rack and pinion driven gantry router that refuses to move accurately under Mach 3.

I have searched for days now as to a formula that will give me correct steps per unit, but every thread has a different way that simply hasnt even got me close.

I just chucked a dial indicator into the spindle, used the steps per unit calculator in Mach's settings to attempt to measure exact and get exact distance travels in return, nothing. After running the test, I attempted to mill a 1" square pocket but got a .9" square pocket in return.

I have a 24 tooth spur gear with 20 degree pitch and a 1.2 inch diametrical pitch paired with a corresponding gear rack, a 1.8 degree stepper and a 10 microstepping Gecko drive. There is no way that I am losing steps as the gantry will push me across the floor or drive me around if I sit on it. My steps per unit simply arent accurate at all.

This is my 4th CNC build (all previous have been screw driven) and its being a PITA. If anyone could shed light on this subject I would be quite greatful.

Thanks!

[LeeWay]

Ya gotta keep at it till you get it right.
I did have to play with mine on my mills Z axis after I installed a timing belt. It was close on the math, but not prefect. I would up adding over 100 steps to get it very near perfect. Mach thingy will get you close, but you will have to keep adding steps and measuring to get it perfect. You may not even be able to get perfect, but still be within your tolerances.

You must be sure that you don't have mechanical issues like backlash or loose parts of course.

I think you will have to add more steps if it is coming up shy. Just add 50 and measure. Rinse and repeat till you get better results. Then zero in and nail it down. Once done, backup your Mach 3 XML.

[idtkid]

Thanks for the help, oddly enough, it seems that when steps are added the axis moves less, when steps are removed the axis moves further, definenitely backwards from what you would think. Encouraging words, I will keep pluggin' away until I get it right, unless of course someone else has some advice?

[HuFlungDung]

What is the "unit" length in Mach? Inches I presume?

Your drive gear has a pitch circle circumference of 3.7699", and it appears as though in your 'lossless system' that your motor turns 2000 microsteps per revolution. So divide the distance by the steps and you get 530.5 steps per inch. Of course, there is no such thing as half a micro step, so you'd have to round up or down, I suppose.

Don't neglect to allow for cutter diameter when milling your 1" square pocket.

[jalessi]

How many teeth per inch is the rack?

[Torchhead]

How many teeth per inch is the rack?

Does not matter. The number can be derived from the Pitch Dia and it's teeth count. You don't really need that since the important question to answer is:



How far does the load travel in one gear rotation.
The answer is in the earlier post. You know how many steps it takes to rotate the gear one turn.

If there other transmission components between the motor and pinon then they have to be factored in as part of the equation.

You will see one effect that direct pinion coupling has: Loss of resolution and torque. Up side is it will be fast enough to win you bets! You can't get better accuracy than resolution (ususally a lot worse) so you start with .001886. Microstepping is "false" resolution since it can change with RPM and is different at different speeds. Your final accuracy is a factor of all the errors times the distance.

If you move a set distance multiple times and keep getting different values it may be a resolution problem. Also R & P has backlash if you don't keep the gear fully meshed with the rack.

TOM Caudle
www.CandCNC.com

[Torsten]

Does not matter. The number can be derived from the Pitch Dia and it's teeth count. www.CandCNC.com

That is what "Huflungdung" has calculated some circumference dia.
Should be very close but I dont think they make standard racks with such a uneven number of linear pitch.
May just take a physikal measurement and round to the nearest even number.
There should be a pitch or tooth per inch value to go along with this to make a correct calculation here.
Note you do want to know the theoretical steps per inch value.
Before you start deviations to make up for other variables in the system.
Good Luck

[HuFlungDung]

Standard racks are based on 'so and so' many teeth in Pi inches, so it might be unusual to find a particular gear of a particular diametral pitch that would move in even inch units in one revolution ( in direct drive). That is not to say that there couldn't be some magic combo in a compound reduction that might do what you want.

BTW, the pitch diameter of the gear multiplied by Pi gives the circumference of the pitch circle. This is the theoretical contact surface of the gear if it was toothless (a friction drive). The teeth are scabbed on just to provide good traction, they do not change the ratio at which the pitch circle rolls along the rack.

[idtkid]

The 531 steps per unit number gives me a milled square, in particle board with a 7/32 endmill, 1/8" deep, going very slow in hopes to avoid further errors, measuring .95" x .95"

I guess I would be okkkk with -1/16" tolerance on my machine however when I move the machine to what Mach thinks is 70" I measure 61".

If my steps per unit were wrong then I would assume I would multiple my .05" shortage by 70 to find how short I would come up. This would give me 3.5", 70" - 3.5" should give me a physical measurement of 66.5" but as stated before I get 61".

I must be loosing steps someplace no?

Any suggestions would be greatly appreciated.

[HuFlungDung]

You'd best show us a copy of your g code, just to be able to double check on the possible nature of the error.

[Torchhead]

Did you account for the tool offset? The number would be to the center of the bit. If it's off then you need to get a RATIO of the error (divide the actual by the intended) and use that ratio to increase or decrease the steps per unit in the same RATIO. Work everything to 3 decimal places. You CAN have fractional steps (it averages them out over distance) and you need to use them. Take the ratio and get a number of steps to add or subtract. In this case you add if you are under the size.

You have to measure to the center of the tool or you have to offset the cut by the radius of the cutter OR you can let the CAM program do that for you.

TOM Caudle
www.CandCNC.com

[idtkid]

Here is the code:
I promise the right bit (7/32" Endmill) is in the chuck

G17 G20 G40 G54 G64 G90
(2 1/2 Axis Pocketing)
T1 M6
(0.21875)
G00 Z0.1094
X0.2750 Y0.2750
G01 Z-0.1250 F40.0
X0.5062
Y0.5062
X0.2750
Y0.2750
Y0.1500
X0.6312
Y0.6312
X0.1500
Y0.1500
X0.2750
Y0.0250
X0.7562
Y0.7562
X0.0250
Y0.0250
X0.2750
G00 Z0.1094
M05
M30

Here is the exact steps per inch that I have entered into Mach 3: 530.51647697298445256294587790858
(Mach chopped it after a handful of digits obviously)
This number yielded a square measuring .91" x .89" this problem just keeps getting weirder. This is a much more accurate calculation but it outputs less precision.

Again, any further guesses would be appreciated!

[jalessi]

I could be wrong however the step over in your G code looks very aggressive, it may be causing missed steps.

[Torchhead]

If the percentage of error is not constant it would indicate you are losing steps. Slow your acceleration and velocity down to about 1/2 what you have and rerun the tests. The percentage of error should be constant.

[Torchhead]

At this point you need to not cut pockets and concentrate on getting one axis at a time calibrated. Start by marking a zero point on the table and setting the DRO to zero and running one axis out as long as you can on your table. Measure the actual. Send the axis back to zero (before you change any settings) using the MDI windows. Now find the ratio of error and use that to take your current steps per (with 3 or 4 digits of decimal) and determine the number of steps to add. Then run the test again. You should be a lot closer. Keep doing the decreasing error calulations. I use a small stylus with a point to get closer measurements.

Once you have one axis set use the same number on the other (if the gearing is the same and you should be closer.

After you have the machine so that it is dead on at a long move then you can start running your small tests.

If you can't get repeatability then either you have serious backlash (poor gear mesh), something is loose or you are losing steps. Slowing the tuning parameters down will confirm the last possiblity

[idtkid]

Thanks again for the suggestions. It would seem to me that a formula would prove much more accurate than measuring with a measuring tape of 80+ inches. Why the formula yields goofy measurements though is beyond me.

I can assure you that the step over is by no means aggressive. It is running at 16,000 RPM, an 1/8" deep and 6IPM. Ssuuppppeerr conservative to avoid any lost steps.

I will try your advice Torchhead and get back to you shortly, thanks again, appreciate it!

[Torchhead]

As in most formulas the fact there are tolerances and variations in each number makes them more ballpark than exact. This is one of the shockers to new engineers that instead of doing pages of math to do circuit design (especially with analog circuits) it's much quicker to have a grasp of the approximations and work around a range of common values. There are inherent errors in everything from the rotor position in the motor to the backlash in the gears.

Tom Caudle
www.CandCNC.com

[idtkid]

Well over 60" in X I am off 1/16 and over 40" I am off 1/32", "tape measure tolerances" are fine for me. Now when I run my pocket I get a .95" x .95". Still doesnt quite make sense but this is the closest ive been.

Shown below is a screen shot of Mach, units/min are showing ~80IPM, im seeing no where near that, more like 30IPM. Why would it be showing me something different? Thanks

[ger21]

Did you close and restart mach after changing settings?

[HuFlungDung]

I did a backplot from your gcode, and then offset it by 7/64 (ie., the cutter radius, and the square comes out to .950 x .950

Always check your 'rithmatic