Maslow Home Maslow Community Garden

Optical Calibration Demo and Three Hours Working on a Bug


My vinyl banner arrived last night so I’m running my first full calibration :tada:

I also have my PWM set to 32kHz so the motor whine doesn’t drive me nuts. Seems to be running fine with 32kHz.

@madgrizzle do some of your squares have fuzzy edges? I noticed some of the squares in the banner are not the most well-printed

But some are definitely better than others

Dimensional accuracy was pretty close at 0.998263889 scale in the X dimension (shorter than expected) and 1.002604167 scale in Y dimension (longer than expected)

Project of the Week / Community Gardener for October 17, 2018

I’ve never really looked that close, but generally I think they look like the latter rather than the former.


Hmm, would you mind taking a closer look at yours? I’m wondering if i got a dud or if there’s something I did wrong.


I didn’t scan every square, but I couldn’t find one that looked like the first picture… they all pretty much look good like your second. How does the optical calibration seem to handle it?


Seems to do ok but i haven’t put things in graphs yet. I’ll do that now.

I wonder if I somehow bumped a scaling setting or something. There wasn’t anything special you did when uploading right? Just picked 4x8, indoor vinyl, and uploaded the file? Perhaps they have multiple printers and I got the janky one

Don’t know if it will make a material difference, but definitely has me a little concerned about accuracy of the edge detection


Same file, same settings. You can always try to complain about the quality with them. We use them for business cards and when our company’s blue didn’t come out like it should, they reprinted the whole order. Never know. I feel about it since I suggested them.


Where are the bad squares (i.e, top rows, bottom rows, etc.)


Don’t worry about it! I’m mostly just trying to avoid inconsistent results between our testing.

It seems like it was by column. Like some columns had bad sides and others were fine. The tops and bottoms of the squares seemed to be fine. Makes me wonder if the scale was somehow off in the x direction. I’ll head down to the shop and take a closer look and some more measurements (and get my results into graphs)


It seems like a printer issue to me.


Yeah I buy that. it seems fairly periodic, like there will be a good square, a slightly worse square, a bad square, a better square, a good square etc. I wonder if we’d have better results asking them the exact resolution they can print for that size and providing a bitmap. I also noticed my banner isn’t actually 96" wide, closer to 95 9/16". None of this is probably that big of deal as my calibration values looked totally fine. I’ve added my latest run to my google sheet.

How did you generate the graph to visualize your curve fit values?


I put the curve fit formula and coefficients into Excel and copied it across the 31x15 cell array.


Looks like a little tilt. Funny how bad our machines are along the top edge… I might try a skirt to see how that helps.


Cool, mind sharing your Excel file for easy spreadsheeting?

Yeah the second graph maybe shows some clockwise tilt. I struggle to understand the first graph. I think it tells me my test pattern is offset 1mm to the left (intersects the y axis ~1mm below 0), my scale is slightly wrong (slight downward slope), and the edges are inaccurate (probably over-compensating for chain sag, or just weirdness from not having a skirt)


Hmm I also got the connection timed out thing again

Sending: $O=31,15,42,42
Connection Timed Out


Hmm and now groundcontrol crashed mid move and it thinks this is the center point :confused:

Is my only recourse to start over with resetting the chains on the sprockets?


Sending $0=value sets the workarea width to ‘value’ in the stock firmware. Or is that a capital letter ‘O’?


I think that’s capital O (the letter)

@blurfl are you all set up to try out the optical calibration routine as well? Would be great to have another set of eyes on this!


Ran another calibration after re-setting my chains. My values (especially my Y values) are a lot better. I wonder if somehow my chain length had drifted over the past couple weeks. Full results in my spreadsheet

I also used this opportunity to shoot some PTFE SuperLube into my motor gearboxes and tighten down all the screws (turned out 1-2 had gotten a little loose)


It’s in the MaslowTestData repo now… under excel files… just dummy data in the spreadsheet. I just recreated it… think it works :wink:

I could be wrong, but anytime I see a constant slope from one side to the other, I think it’s indicative of tilt. Both charts seem to show a downward slope from right to left, though it’s easier to see in the first chart.

Your chart does indicate an offset of the pattern… but that might not be the pattern (though it could be), but could also be that the motors are not spaced equally apart from the center of your spoilboard (i.e., the center between the motors does not coincide with the center of your spoilboard). But, if you aren’t trying to do cuts that go all the way through to the edge, this isn’t going to be a problem.


So these new ones don’t suggest frame tilt… got me on understanding that one.