Mach 4 axis scaling, not with G50/G51

So, I’ve been making a few parts recently that required being milled out of 1/2" mic6 aluminum plate. because that material is expensive, I was prototyping the parts on my router out of 1/2" MDF and noticed first that my router was skin cutting even when it said it had moved 1/2" down, but then on inspection with dial indicators, gage blocks, and calipers that all 3 axes of my router were being called down to about 99.63% of what they should have done.

First, I checked that G50 was active so no code scaling was applied.

Then I looked around in Mach for some “scale factor” type multiplied, but I couldn’t find one.

. After some math, I corrected the steps / inch in Mach to fix this and now it works fine, BUT,

Is there any way to add a scale factor anywhere in Mach that could cause this? on Y, Y and Z I was seeing a uniform loss of .033"/7" (I have a probe and 7 inch gauge block, so that’s how I checked, and then verified with a dial indicator.).

I have no suggestion except the obvious one. Had you done this calibration before?

How did you check this? I’ve been wondering if there was a place in Mach4 that displays the current status of all the modal G-codes. I had something go wrong in the middle of a run and had a really hard time restarting it because G17 mode was active, but it needed to be in G18 instead.

The standard Mach screen set does show this, not sure about the Avid one.

The other answer is, I had not validated in as much detail before, but, I have 6400count encoders on my servos, and different mechanical systems on all axes, so the possibility of being off by the same scale factor everywhere says software, not hardware. My other machines are dead-on the calculation.

@abakker I’d ask on the Mach support forums

@Stephen The Avid screen set does, it’s at the bottom between the Enable/Disable button and the History panel.

Anecdotally, that number is almost exactly the difference between pi/20 inches and 4mm - if you have a pro setup and assume the rails are 4mm pitch, your movements would come up this short.

Despite the rest of the machine being metric, the drive gears are imperial, with a pi/20 inch pitch.

So, I do have pro R&P on the long axis, but gantry and z axis have 10mm and 5mm lead ballscrews respectively. I would think I’d have to have some freakish inconsistencies to replicate everywhere mechanically.

I said “anecdotally” but sometimes it’s good to verify the stupid things first :wink:

And you said that Y was a Pro R&P and it’s off; I’d double check the math on that one and measure it with a tape measure across the max move. I was off by only 1/4 inch across 100 inches, but that’s how close those factors are.

It sounds like a custom machine with some Avid parts, and maybe not running Avid’s Mach4 customizations. Interested to learn more about it and, ultimately, the solution.

Honestly the only Avid machine parts are the R&P drive, and the ESS/Avid 850 and Avid 860 boards and their sensor setup. The rest is custom. But, I’m an Avid machine fan, so I hang out here.


Are you measuring with MDI commands to rule out CAD/CAM?

Edited for clarity - when you were performing your measurements and found the uniform loss, were you using the MDI?

So, my exact method of measuring was using a drewtronic touch probe and a 7" gage block.

I would probe the X positive side of the block, note the Machine Coordinate for that probed point, probe X negative side of the block, note machine coords, and then subtract. then compare the “measured” distance with the gage block, find out it was .9963, multiply that by my steps per inch and then change my steps per inch until I got a near-perfect result.

this way, I was only ever using MDI and known standards.

Do you see an error in that approach?

Are you probing in opposite directions for each side? If so, it would be more accurate to place a gauge block against one side and probe that, so that you probe both sides with motion in the same direction, the same side of the probe, and independent of the diameter of the probe.

Plus, I’d double check that with the lowly tape measure across the entire width of your machine. The further apart your measurements are, the more accurate you can be.

As for “tweak steps until it works” - I really don’t trust or like that approach. If your measurements don’t match the math, FIX THE MATH. Don’t just fake it until you make it. If you have the wrong steps/distance then you don’t understand your machine enough.

Did the math, and the math didn’t match reality. That’s what the whole point of the thread. Something is making my math wrong by a uniform scale factor in 3 mechanically dissimilar axes.

Yes, why are you not using the calibration tool built into Mach4?