What measurements and observations should we record for the various linkages? Is the wiki post appropriate for this?
"A wiki post/admin wiki post is a special type of post that can be edited by any user on your forum.’
This post should now be open to all for editing and updating.
Values to Record
record the rotationalRadius value used as an estimate before calibration
record the rotationalRadius value determined by autocalibration
record the number of iterations required in autocalibration
I don’t think any of those calibration related things need to be recorded. We
need to be doing something like cutting squares all over the sheet and record
the errors that we see.
The calibration numbers, machine dimensions, and frame type would be interesting
to record in case we have a pattern or errors in the test cuts, but it’s the
results that matter, not the calibration (I still hope to replace the triangular
calibration)
Recording those values doesn’t hurt, and might help understand other results. The numbers might also be useful to future users setting up their machines.
If the linkage were the only variable, but across different systems with different sled bases, mount angles, frame geometry, work surface flatness, sled weight… we really need more than just squares cut. Cut speed, cut depth, bit age/sharpness, there are a lot of things that can affect the results. We need to see results from various machines to really reach an understanding.
Those measurements will be nice to know but the real thing we are after is how accurate are the different solutions (across the entire work area) and how easy are they to use, right?
The ultimate goal would be:
inexpensive
accurate
easy to use
easy to calibrate
easy to ship
So all kinematic solutions should be tested against those conditions (or similar, I’m not saying these are set in stone) in the same way. Things like “easy to use” are a little more subjective but can still be judged and documented. Things like “accurate” are totally objective and should be easy to compare.
Things like bit sharpness, cut speed, cut depth etc. should all be kept the same for testing purposes. (obviously we can test different speeds etc. but all kinematic solutions should be tested using the same conditions).
True, and a standard test file assures some of these, but one records the values to assure that they are used and to capture situations when for whatever reason they aren’t. It only takes a moment to record a number . As we will have multiple testers, we need to standardize the tests, but sled weight, frame angle, etc. will vary between testers, and could well affect results for different linkages. We might reveal factors that favor one linkage which had behaved poorly with other settings. Results from one tester should be shown to be replicable by others.
With more data this could be more than just a horse race.
@Blurfl: Thank you for making this thread. I saw it this morning but I spent most of my day fighting with a stubborn point-to-point machine, sorry I’m late to the party.
I had mentioned in the referenced thread that the sled was a bit too close to the top edge when I ran it last time. I’ve moved it down to 150mm from the top edge. That should be a little more stable at the top. I plan on using a strip of plywood above the test piece to give me more of a fighting chance. If we want any other changes to be made to the test pattern let me know and I’ll take care of it.
One thing I did not do in my last test was check that the locations of each of the squares were in tolerance with each other. In my notes, the center squares 5 and 14 were dead on, so I will be using them as a reference for the other squares. I think that using the edge of those squares as a “datum” should be fairly reliable:
May also want to check if any of them are not colinear in the Y-Axis. I doubt that any of the triangular linkages would have enough error in the Y axis we’d be able to measure it effectively.
On the note of tool sharpness and feedrates, I plan on using the 1/4" single flute bits bits I’m always recommending. I have the feedrates dialed in pretty well on those, and I think they should be a good baseline. The 500 mm/min (~20 in/min) I’ve been using for them is slow enough that any “pendulum effect” should be minimal and we won’t need to worry about acceleration planning. Honestly, it’s probably best that I do this before feedrate tests I’m planning on doing. That way we are all working from the same baseline hardware.
I’m writing up a spreadsheet to track the information that we gather from these tests. When I have it somewhere near a final draft I will post that.
I am temporarily out - triangular (back soon!), but I can’t wait for a world wide test of the same .nc
Even with different frames, or should I say because of them, I would like all the information gathered (YES! Also the temporary irrelevant ones) as a step forward.
@Gero Thanks for the kind words. I also would like to see as much relevant details gathered as possible from these tests. I had a chance over the weekend to put together a quick spreadsheet which should help gather information. Unfortunately life got in the way and I have not begun the testing process myself.
I added in some of the information from my previous test to make sure my math was working correctly.
I was unsure if the other testers would have access to excel, so I uploaded it to Google docs. The link below will serve as the source file, if you are running the test you should either copy it to your own drive or download it (see image below). I have set the sharing properties so that only I can edit it so someone’s test doesn’t accidentally get added to the source file.
While adding information to the tables, edit only the values in black. The orange ones are formula cells and will compute any error recorded in the table. Changing any of the values in orange will override the formula and reduce the functionality of the sheet.
Wow, nice and clean! Thanks for the time you put in. I think the hight of the motors is relevant. Now it is measured from the top of the ply, my preference would be to measure it from the bottom. You could hook the chain in the down beam and do automatic length as well. I would like to see motor hight and width in 2 columns, automatic and measured by hand for comparison.
Thanks @Gero! That is an astute observation that I has missed entirely. I’ve added the motor height to the spreadsheet as it’s currently measured in GC. I also made an auto and manual column for the measurements taken by the auto calibration routine. This probably will help @Bar figure out where there are discrepancies in the routine. I figured that there should also be values for the bed size in there, but I’m not sure how much that will effect the tests. I’m not sure I think the header is complete yet, but this is what it currently looks like. I tried to make sure there was more room to add more data if we need it.
On the topic of where to measure the motor height to, I think this is the perfect use of our newly discovered poll feature. What do the other testers want to see?
Motor Height Measured to the Top of the Spoilboard (as GC currently records this data)
Motor Height Measured to the Top of the Fence (As Gero suggested we may want to collect this data)
If ‘Top of the Fence’ is the bottom of the cut sheet, Yes, if I may vote for my preference.
Good to put the Bed size in, I forgot that.
A poor guy who starts with a ply sheet from a garage sale might end up with a random different motor hight during calibration. Not so if calibrated from the bottom, I think.
The math is taking this into account with the Bed Width I guess.
On the standard frame with have a slightly rough estimate of the bed already. Where the motors should be.
A calibration of the frame instead of the spoil-sheet?
Edit: I must admit that I have a hinged frame and no permanent spoil sheet. I use easy exchangeable 4mm and if I don’t have a complete one, a little double sided tape and some scrap pieces secure that I’m not cutting into the frame.
The height of the motors does matter, that’s why in the triangular calibration
topic it’s one of the two unknowns that are being solved for.
That said, we have two different things that we are looking at figuring out
the overall accuracy of the machine
and
how the different linkage/ring kits compare to each other.
this discussion started with the second one, and hile it’s nice to try and
figure out the first one as well, we really need to get a comparison of the
different kits to figure out which is better to ship with the next batch
With respect, I’d say that the overall accuracy is more important. This will be a lot of effort and we should glean everything we can from it.
Further, there have been some fundamental improvements in the last week which would affect the accuracy measurements. I wonder if there are any more we should nail down before going through the exercise.
Since every maslow is a little different (some a lot), maybe add room for a picture and some text. I would think build quality, craftsmanship and materials used can be a significant factor in the accuracy as well. If you think something that doesn’t have a dedicated field makes your maslow better then everyone elses you need a field to put in.
Or maybe those things are better suited for the Miss Maslow contest we have to do in a few months.
Also we should all do the tests in the same material, which should be cheap and available everywhere, so that the people who don’t want to spend €50 on a sheet of 18mm plywood to throw away can also join. Maybe even make a maslowable pen holder so we can do the tests without a router.
In my rough draft I had a “machine notes” field, which I wasn’t sure if it was necessary. I can add that back in. As for pictures, it probably is a good idea, I will probably have to add a second tab that users can put pictures in.
I have a sheet of 3/4" CDX that I used for the last round that I plan on using for these tests too. A sheet I believe is somewhere around $30, so it isn’t that bad. Honestly, thickness shouldn’t matter too much so if you wanted to save some money on a 1/2" or 5/8" sheet that shouldn’t be a problem. I also like the idea of using a pen, since you could easily put the shapes on a sheet without “ruining” it. However, how do you accurately measure that? IMHO, there is a lot of room for user error there. I feel like it’s much easier to take the calipers to a hard edge.
I think i does, a little error in one pass and a little error the other way in another pass give bad test results, we should take material that can be cut in one pass.
Duct tape is your friend here though fuzzing it up a bit first so it will come off is important! Seriously, it’s quite possible to tape or clamp the dumb end of a tape measure in a place that will allow a loner to accurately measure. I pre-flight everything plotting it first.
The NC file I posted above only cuts 3mm into the material. I programmed it so that each of these patterns only gets cut as a shallow grove into the material. This should do a few things. First, it should reduce the effect of sideload on accuracy. That way, we should get the best possible results from the test. Second, the cutting should go faster. With less Z-travel, the machine should be able to make more cuts in less time, given that the Z feedrate is currently one of the slowest movements of the machine. And thirdly, we won’t have to worry about a part getting cut out and shifting, which would throw off the accuracy of the test. Generally, cutting out a part with tabs doesn’t shift, but I’ve seen some unexpected things happen while cutting and I’d hate for that to throw off someone’s tests.
It could be difficult to get an accurate measurement (especially when we’re looking at < 1mm of difference) with a tape measure. This is a bit of a moot point with the positional error columns since it would be difficult to measure such long distances with a pair of calipers.
Also, it’s a little subjective measuring to a printed line rather than a hard, cut edge. The operator would have to account for the thickness of the line and always measure to the same points. At least taking a pair of calipers to a routed shape you would be able to consistently measure the 100mm shape of the square.
I don’t know what kind of pen others use, I’ll be using a 0.5mm pen - no Sharpies here😉. As @dlang points out, the rules for measuring pen plots are the same as for cuts. Plotting on 20lb. paper with a fine tip makes easily measurable traces. An added plus is that you can change to a different color pen, replot, and have a visual check of repeatability.
I’ve been meaning to ask if the .SVG file of the test pattern is available? I’d create a ‘cut-on-the-line’ version so that the pen traces would match the insides of the cut squares. I’ve created one, but the order of cutting is different. Does the order matter?
. As we will have multiple testers, we need to standardize the tests, but sled weight, frame angle, etc. will vary between testers, and could well affect results for different linkages. We might reveal factors that favor one linkage which had behaved poorly with other settings. Results from one tester should be shown to be replicable by others.
