Z depth test.

OK, so this is what I’m doing. I have about a 32"x50" cut area. I made a grid of 3 rows and 5 columns. Each column is 12" apart (for 48" wide) and each row is 10" apart (for 20" total). I cut a 1/2" helical drill in each location. I cut those into the spoil board, and I used the depth guage of some digital calipers to measure the depth of each cut. Since it’s all from the same gcode, any difference in depth should be because I fail at calipers, or errors in Z.

I expected to see three rows of nearly the same numbers, but with deeper cuts in the middle than the outside, because I expected the tubes to bow in the middle. I only cut three rows because I wanted some more data points.

What I actually found is that the difference across the 48" axis is actually as much as the difference across the 20" axis… Here’s some raw data:

average depth of each column (in mm, this is the platform moving across the pipes):

Left 0.96 1.55 1.37 1.30 0.23 Right

average depth of each row (in mm, this is the gantry moving on the wheels):


I wish there was a better way to visualize the data, but any plots I’ve made just don’t tell the story.

All in all, I’ve very happy with these results. The complete range of z values over that entire area is 2mm (48in is 1200mm…). That’s about 1/16th of an inch over 48inches. But the “standard deviation” in that area is 0.6mm, since except for the corners, it’s very consistent, indeed.

The measurements for the columns makes a lot of sense to me. Shallower on the ends, deeper in the middle. The rightmost column is shallower than the left, and I can only figure that the pipes are bent, or the vacumn tube is causing some error.

The measurements for the rows don’t make sense to me. I don’t know why the rear of the machine would be cutting shallower than the front. I’m going to look at the machine a little closer with this in mind, and see if I can figure it out. Some ideas I have are that the wheels aren’t completely flat, and the machine landed on a different side for the rear and front rows, or maybe the table thickness is slightly different, and at the front, the bottom wheels are pulling the gantry lower, whereas in the back, there’s less downward pulling of the wheels, so the top wheels flex a little less. I might cut a groove in the spoil board from front to back and measure the depth along it to see if I can tell if it’s a smooth sloped pattern, a discrete one, or a wavy one.

At any rate, I definitely think these numbers are better than I could have hoped. I am sure they are better than the MPCNC I made, because I was adding more to the depth of cut just to consistently make it through material. Ryan said at some point that in the size range of a smaller low rider, or larger MPCNC, that the low rider might be the better choice. I think this supports that. I would be curious to see what other folks find with a test like this.

BTW, here are all 15 measurements, if you’re curious:

[pre] Rear
Left 0.66 0.84 0.79 0.77 0 Right
1.03 1.72 1.5 1.39 0.24
1.19 2.09 1.81 1.74 0.44

1 Like

Some images, to give a better sense of direction:

[attachment file=34289]

[attachment file=34290]

I ran it again, but a little deeper, to avoid that 0.0 in the rear right. This time, I did it with the lower wheels loose, so they can’t pull down the gantry. I can tell that the table is a little bit thicker in the back (0.4mm). Basically, I ended up with the same results. I am going to leave it like this (Like I said, I’m pretty happy with this result). If I wanted more precision, I think I’ll have to surface the spoil board, or somehow build that mesh into the controller… Here are some graphs:

The Y axis is difference from the middle depth
The X axis is left-right (columns)
The Colors are the front-back, 0 is front.

[attachment file=34293]
Test 1 (wheels tight)

[attachment file=34294]
Test 2 (wheels loose)



1 Like

It might be possible that your board is not entirely flat in the first place.

How did you make sure it was flat before running the test? Since it is your reference for all measurement, you better make sure that it is correct in the first place.

Running a surfacing bit on all your table making a very shallow pass will indicate you where are the high spots.

Otherwise, that’s quite an interesting test process, thanks for sharing your method, I think I’ll use it on my MPCNC.

1 Like

Cool idea.

Lots of variables. I think the biggest would be the rails the wheels are on and there contact with the table. MAybe do a test run on just the table.

The numbers are good so far but I would love to see them a bit better.

I have a project on my table right now but I would love to try this myself. I have been trying to come up with some simple projects that might require pockets, I should probably test this first.

1 Like

I think MDF will be the most consistent thickness if the ply is the problem.

There is way more variance on the roller axis vs the rail axis. It almost seems to slope pretty evenly.

1 Like

The wheels travel on the surface of the spoil board, which is why I thought it would be quite consistent in the roller direction.

By that, I mean I have two pieces of 3/4" ply. Each one covers half the table. The wheels travel on top of the plywood, and it’s what I used as a reference surface. If the rear part of the plywood was thicker, then I would expect the rollers to just ride up higher. They aren’t exactly the same Y as the bit, so maybe there’s some discontinuity.

I agree that surfacing the board would be good, and would tell me a lot, but it seems like it would take days.

I also agree that I wish it was better, but I’m more than happy ATM. I can’t think of any large projects I have that need more precise Z. Otherwise, I’ve had a great couple of cuts, and I’m going to try some big, more expensive things next.

If I was doing some carving or pockets that needed precise Z, I might try to make separate gcode files for those parts, and set the z locally. It seems very consistent over any 12" area, except maybe the right side.

I’m just looking at those graphs again. They look more like footballs, with more variance from front to back in the middle three points. I wonder if the rear middle is sagging more than the gantry, because it’s either not well supported or the table is slightly warped. The variance at the left and right side is much smaller. I’m​ pretty sure I will never run out of ideas. I’m just going to fille these spoil boards with little holes, aren’t I?

This student went into his physics professor’s office and said, “Last night, I noticed that all the smoke coming from chimneys on my street was flowing up in a clockwise direction.” The physics professor’s thought about it and said, “I can explain that…” The next day the student returned and said, “Sorry, professor. It times out they were all moving in the counter clockwise direction.”

The physics professor thought for a moment and said, “I can explain that too…”

That’s how I sound sometimes.

Yeah it really could be anything. ±0.6mm is not bad for a large area, on plywood. If it matters later on i’m sure a new table will help. Mine might be worse I only planned on through cuts but now I am reconsidering.

I really think it is necessary before doing any further measurements. Otherwise you just cannot know where the problem is actually coming from.
For instance, on my table, I use two 20mm thick MDF panels, supposedly one of the flattest wood possible. Well I still have a variation of around 0.5 mm between some places. Plywood is usually a lot worse than that.

Surfacing could be pretty quick actually, you can do very thin passes at high speeds. First cover your entire area with pencil or anything that can leave color on the surface without being absorbed too much by the wood, then run your surfacing bit on very shallow passes, like 0.2 or 0.3 mm at a time, at whatever speed you feel is fast enough (not too fast for the first time, but can be maximum speed for other passes). It will instantly reveal which areas are lower since you’ll still be able to see colors at those places. You’ll instantly see if and how those patterns match the holes you previously drilled. Works kind of like a 2D height map

Before and after flattening your work surface, I suggest you use a level or any very flat metallic bar to check how flat is your table looking at naked eye. After completely flattening it so that there is no color left anymore on the work surface, this method should help you to locate the spots where the issue is coming from your rails, wheels, vacuum hose, etc.

I know it sounds like a hassle, but I think that’s really the only only solution for having precise measurements. Bonus: You’ll also be able to see how perpendicular to the work surface your router actually is, by observing the pattern the flattening bit is leaving on the surface, which is a quite important data too. I keep on postponing this operation for weeks now and I regret it every time I try to attempt a big 3D print, don’t be like me!