Help dialing in precision

[2_many_hobbies]

I have a full 4x8 router mill running and everything is dialed in and reliable, it's closed loop so I can guarantee no missed steps or mechanical issues. I just started cutting test pieces to make sure I have full accuracy in all 4 corners as well as the center but have an issue.

I have a test shape that is complex, has nooks and notches on X and Y sides so that any errors are easy to find (it's a little 200mm stool to test slot fitting as well). My problem is that EVERY dimension is +0.6mm (+0.3mm/face machined) than expected in every direction and is fully repeatable in all 4 corners.

I first suspected my end mill was wierd but It is 1/4" (6.35mm) and is this one SOWA #103-854.
When I open the gcode and do some point math on the file my 200mm piece is 200mm + 1/2 the tool dia each side so 206.35 as expected. This results in cuts +0.3mm/face. I can eliminate this error by having a -0.3mm correction factor in CAM (I am using Cut2D Desktop) but I would like to figure out WHY this is happeneing first.

Looking at the design of the test piece I used a 9mm nominal thickness and cut it out of 8.47mm (3/8 ply). If I scale the design down 94.11% and use no offset factor in CAM I get a perfect fit in reality cut not the dimention I designed. This will work for fully new designs but if I want to make something to fit an existing project the scaling concept will not work at all.

I then did the most simple thing of cutting a 100mm square with a v-care bit in the router. When cutting on the line its 100% accurate.

Any idea where this 0.3mm/face error is coming from?

Similar Threads:

• 0.4MIL High Precision CNC Router Equal to CNC Lathe Machine Precision
• REG Precision Metal Arts - located in Massachusetts offering precision machining.
• Troubles dialing in A axis
• Need Help!- dialing in a 4th axis?
• dialing in tools....

[ger21]

Possible machine deflection. Are you climb cutting? If so, try conventional cutting, and your part will likely end up smaller.

[wizard]

Something just doesn't sound right here. When you say off in every dimension are you saying a box is cut big in both the X & Y by 0.6mm. At the same time a cut with a V bit is right on? This makes me wonder if something in your CAM software is screwed up.

It can be a huge benefit to have some hand coded software written for debug and maintenance. In this case I would write a bit of GCode to mill out a box. Write it in such a way that you understand what every line is doing and understand precisely how far the cutter is travelling for each side cut. Then measure up your freshly cut square. Do not enable cutter compensation or other wise try to compensate for cutter diameter by the way.

I just find it peculiar that you are off the same amount in both the X & Y directions. The idea here is to take the CAM software out of the loop.

[firepong]

Did you check the steps for the system? They might have to be adjusted. Using a calculator is a good starting point, but you have to take into account the design differences in your moving parts. For example, my leadscrew on my small machine was calculated out to 200 steps per revolution, but in actuality, it was 193. That extra 7 steps made the machine 5mm off over 600mm of travel.

I would look it over and check your using servos or just at the GCode for the steps of the machine. If using mach3/4, go in and do a calibration on the motors.

[fretman_2]

That happened to me too...the steps that I finally wound up using were slightly less than what was calculated. I used my dial indicator and kept adjusting the steps until the machine gave me the repeat movement accuracy I wanted in all directions.


QUOTE=firepong;2204970]Did you check the steps for the system? They might have to be adjusted. Using a calculator is a good starting point, but you have to take into account the design differences in your moving parts. For example, my leadscrew on my small machine was calculated out to 200 steps per revolution, but in actuality, it was 193. That extra 7 steps made the machine 5mm off over 600mm of travel.

I would look it over and check your using servos or just at the GCode for the steps of the machine. If using mach3/4, go in and do a calibration on the motors.[/QUOTE]

[2_many_hobbies]

Hello all, it's been a week of 105 deg weather and air quality warnings from the smoke and I only managed to do a few test cuts this week.

Hand written gcode for a 100mm x 100mm square in both climb and conventional cut. Both were off the exact same amount but it was a little hard to measure the climb cut piece as edge quality was poor, I had to do a quick hand sanding on it and take multiple readings to average. I will be sticking with conventional milling from now on.

I took some measurements of the cut kerf but the pointy bit of the calipers and the fact its hard to get a reading square always brings me an average reading of the correct bit width. I am going to try to measure the 100mm hole left over as it should be 106.35mm.

As for the steps issue, this machine is a servo/chain drive setup with encoders on the motors so there is no steps to modify. Software is p

[firepong]

Hello all, it's been a week of 105 deg weather and air quality warnings from the smoke and I only managed to do a few test cuts this week.

Hand written gcode for a 100mm x 100mm square in both climb and conventional cut. Both were off the exact same amount but it was a little hard to measure the climb cut piece as edge quality was poor, I had to do a quick hand sanding on it and take multiple readings to average. I will be sticking with conventional milling from now on.

I took some measurements of the cut kerf but the pointy bit of the calipers and the fact its hard to get a reading square always brings me an average reading of the correct bit width. I am going to try to measure the 100mm hole left over as it should be 106.35mm.

As for the steps issue, this machine is a servo/chain drive setup with encoders on the motors so there is no steps to modify. Software is p

The last statement is true, to an extent. If you are using Clearpath or Leadshine servos, they have a software setting for setting the steps. Clearpath's usually sit around 1200step/rev, but majorly depends on what resolution you are running on them. Can't say what the leadshines run at as I have never ran them or had a chance to tinker with them.

These steps will have to be fine tuned on the machine to get completely 100% accuracy on cuts.

[wizard]

Hello all, it's been a week of 105 deg weather and air quality warnings from the smoke and I only managed to do a few test cuts this week.

Sorry to hear that.

Hand written gcode for a 100mm x 100mm square in both climb and conventional cut. Both were off the exact same amount but it was a little hard to measure the climb cut piece as edge quality was poor, I had to do a quick hand sanding on it and take multiple readings to average. I will be sticking with conventional milling from now on.

I'm beginning to think that your lead might be off slightly thus needing an adjustment tot he steps /mm (or whatever) settings your CNC has.

This can be checked with a set of joe blocks or precision 123 blocks. Using inches as a measurement basis; you would need to set the blocks up in a string parallel to the axis so that you can remove blocks in integral inch values. This would require a decent dial indicator with a stiff indicator stand. With your string of precision blocks set up you would jog the axis such that the dial indicator would contact the first block in the row. The dial is zeroed out on the gage and one block removed from the string. You then command the axis to move precisely the distance of the block. We will use 1" for example here. The axis should stop at what it believes is a 1" move, the dial indicator should show any lead error. YOu will likely get some error, maybe a few thousands depending upon the quality of the screw. A larger error likely indicates that the lead of the lead screw isn't what the CNC controller thinks it is. If the error is large you need to take many readings with different distances and along different positions on the lead screw. You should be able to come up with a more precise value for steps per millimeter (or whatever your controller uses).

This is just one way to check your lead you can use digital callipers if you have the ability to take high quality measurement. The goal here is to get a base value without machining anything. This takes cutter diameters, deflection and other ills out of the equation. The trick here (no matter the method chosen) is to get a high quality number for the actual lead of the lead screw. The reason here is that rolled lead screws especially, can have a lead error. You can also have an error in the current setup of your CNC controller.

At this point I would want to eliminate the potential of a lead screw lead error or a mistake in your CNC setup before doing much more. The fact that you keep coming up with the same error on both axis has me leaning in this direction. I About the only other possibility is a gross error in cutter diameter or a compensation value set some place that shouldn't be set.

I took some measurements of the cut kerf but the pointy bit of the calipers and the fact its hard to get a reading square always brings me an average reading of the correct bit width. I am going to try to measure the 100mm hole left over as it should be 106.35mm.

As for the steps issue, this machine is a servo/chain drive setup with encoders on the motors so there is no steps to modify. Software is p

Sure there is. Some place in the CNC controller it is counting values or increments based upon how it is setup. I'm not sure what sort of servo drive you have but it it may very well be taking step and direction signals from the CNC controller. In any event some place in that system encoder counts are being counted. Those encoder counts however are meaningless unless transformed by the CNC controller into counts per millimeter (or what ever) that your controller is using as its basis for positioning.

I'm trying to make this clear so to look at this another way say you have a belt drive with stepper motors only. Say we went through the math and determined that each step for the stepper equalled 0.01mm of linear movement. Some place with in our CNC controller there has to be a vale that relates how many steps are required to get 1 millimeter of movement (again whatever your controller uses). In this case 100 steps. The controller bases all of its math on this number which maps to the linear world of your machine. Now take the same system but through an encoder on a servo motor. The CNC controller needs to know what count is equal to on that system. On a system with a 1000 count rotary encoder you might need 500 counts for the same 1 millimeter movement (this is a bit contrived for simplification). So the controller does its math based on 500 counts per millimeter. In either case it rotates the motor shaft and thus a sprocket or pulley one half revolution. The controller doesn't know how far to rotate the shaft until you tell it how far the linear system is expected to move for a move the controller specifies.

Another way to look at this is that the controller does its calculations in a virtual world and then has to translate that at some point in to values that related to the physical world. The specifics of how the controller does this is dependent upon implementation but it really needs to know what a step or a feedback bit means in the real world. This is always part of the controller configuration.

Out in the real world we often have machines where every axis has a different motor, gear box and even linear conversion method. One axis might have a rack and pinion, another might have a lead screw, another a belt drive all on the same machine (not a router in this case). Each axis has to be properly setup in the controller so that the controller knows what is required to move a millimeter. Think of it as a translation, the controller has to talk to the physical world in a way that makes sense for the indvidual axis.

I hope this helps. A quick review of your controller documentation should help locate the parameters that make a difference.