Calibrating: Measuring the machine vs. measuring the part

I’m still on my journey to fix-up the Shapeoko 3 I purchased used. Just yesterday I got the machine square by adding washers behind the two rear v-wheels on the right-hand of the gantry. The right-side gantry plate was bent slightly, adding 1-MM behind each wheel (in addition to the stock washers) brought the machine into square. Awesome.

I was pretty certain the belts had some stretch, milling acrylic and polycarbonate parts was resulting in undersized components. On the X-axis, for instance, an expected 3.5" was coming in at 3.486". And on the Y, an expected 4.5" was coming in at 4.486".

I know deflection can contribute to undersized (or oversized) parts, so I didn’t want to rely entirely on the parts to determine my adjustments to $100 and $101.

So I mounted a cheapy digital dial indicator (from HF, with a button tip sourced from McMaster I think some time ago) to the table, and compared the DI to the values from CM.

An expected 23mm move was coming in at 22.91mm on the X-axis, and 23mm was 22.92 on the Y-axis. If you take this as a ratio, and multiply the actual part width, you get: (23/22.91) * 3.486 = 3.49969 (3.5 in the real-world). Awesome.

Doing the same for the Y-axis, I get (23/22.92) * 4.486 = 4.502, also pretty awesome. I am still getting a rougher finish on these edges so .002" oversized is not surprising to me at all. That is my next “issue.”

So I thought it was interesting how the two methods to calibrate/adjust for belt stretch correlate with one another quite closely, and figured I’d share my results.

I did use the values I measured via DI to adjust my $100 and $101, and then re-tested with the DI. Over the 1"+ of travel of the DI, I was never off by more than .03mm and was often within .01mm or spot-on. I don’t think these indicators can truly resolve to .01mm (that would be .0004"). So every few moves of the router by 1mm, the Carbide Motion #'s would be spot-on with the DI #'s again, over the full 25mm travel. I don’t think I can do any better than that.

1 Like

Did I miss where you talk about the table? Is it level to the Y axis rails and level left to right (X axis)? is the spindle square to the rails (-X to +X) and (-Y to +Y)?

If you didn’t do this yet, you’re not done.

It should also to be rechecked a couple of times a year.

1 Like

Yeah, previous thread had a DI mounted to the carriage and I found my particle board table to be impressively flat to better than +/0 .003" over most of it, dropping-off in one corner b y about .005 or .006", which I haven’t shimmed yet (but will, when I take my sacrificial board off).

I can see evidence of the drop when machining through parts and inspecting my sacrificial pieces, so I get a fairly constant reminder to take care of it.

I am kinda winging it, figuring things out as I go. My checklist:

(1) Check X/Y extrusions are square to table.
(2) Check X/Y extrusions are square to each other.
(3) Check table for flatness.
(4) Check bit for perpendicular.
(5) Check plunging for square.
(6) Check X/Y/Z for belt stretch.

Mind you, that isn’t the order I went. It is the order I’d go now, having walked in my shoes.

How does that look?

Am I right in thinking that making any adjustments during the process requires going back to step #1?


Most people use the “what you cut” method because it doesn’t require any special setups (ie. a DI or TI), tools (a good sized, accurate machinist square or two can help a lot), or a lot of care in setting up.

Nothing wrong with the “what you cut” method. I wanted to achieve accuracy to within .001 or .0015, if possible, and figured I could easily introduce that much error in my measurements of the part.

The DI eliminated a lot of uncertainty and I just attached it to a block of wood and did the masking tape/super glue method to attach to the table. I already had the indicator I use for setting-up my other (woodworking) machines, it is one of the digital ones from Harbor Fright, looks like it sells for $40 now but I paid I think $20 on sale at some point.

I did cut one of my test parts and got to within .001" over 3.5 and 4.5" distances, so I’m pretty thrilled with that.