I don’t understand this. I’ve never had a problem like that. How are you exporting the dxf?
It’s been years since I used SketchUp but I do remember that before exporting to DXF you had to switch to “top” view and “parallel projection” (?) to get accurate scaling. I don’t know if that’s the problem here but it did at one time have to be done in a very specific way to get correct scale from SketchUp.
Isn’t Sketchup a 3D program? For modeling ? I know that all his models were don already and a fix was found but a easier way must exist or a 2D CAD for 2 D drawing I use corel or inkscape for 2 D and sorry Tinkercad for most 3D
It is nice to know the dfx from Sketchup work in Estlcam
Hi guys,
It’s nice that you’re interested in my little disasters. Thank you.
As I write this, the LR2 is cutting the first test-fuselage template. In MDF. 5mm feed, 2mm pass depth. Taking forever, of course, since this is a full 2500 x 1220 piece. But that’s OK…
Intersting observations regarding SU and DXF exports. I’ve posted pictures and commentary on the Sketchup forum, so let’s see what comes back.
I have a terrible confession to make. My original adjusting of the steps/mm (both X and Y axes) was on the Sketchup DXF. Which was out by 10-odd percent. I subsequently QCAD’d it, and the dimensions were ALMOST spot on with 100,100,400). I did have to adjust the X-axis steps/mm slightly to get it perfect, but it was only out by a percent or two…
Bloody hell - 5mm feed rate is awfully slow…
Regards,
Duncan
2 Likes
You should be able to run 500mm/m or better easy just work into it now that you got it sorted. Probable a lot better
I dont have a lowrider2 but it is designed for full sheet I use the MPCNC
Jeff is the expert with this but it isn’t much difference in speeds
I appreciate this perspective. I really do. I think we are just talking different things.
If you had steel belts and there was a break in the steel, it would stretch, but only in that section. So driving over that section might give you 4mm extra every time. If you measure over that section and get 104mm when you were supposed to have 100mm, and then change your steps/mm to match, you will get exactly 100mm over that section. If you then tried to move 200mm, you would get 196mm instead. Because you’ve compensated for that problem once, and then again, even though there was only one break.
If your far motor wasn’t working on an mpcnc, and it was just being dragged by the front motor (broken wire, loose grub screw, etc), you would end up with variable backlash where it would depend where the Y was when you measured X.
If your wheels were translating sideways you get weird issues that are not proportional to movement and even if you can get a 250mm move bang on, you will have a different error at 100mm or a different error if you make a clockwise vs. a counterclockwise square.
If bought 20T pulleys, then the calibration would be dead on, but you would have 25% less torque.
The pulleys, motors, stepper drivers have tolerances, but they move in whole integers. I think that is something most people don’t know when they think about this.
The belts have measurable stretch, but it is less than 1% and they hardly ever shrink. Usually much less. There are a set of errors, in the 5-19% magnitude that should be understood before you try to tune it out. I learned this on my 3D printer. I had a ton of backlash and I printed a 20mm test cube, like you do, and I calibrated it to print that cube perfectly. Then I printed a 50mm part and it was even further off than before I messed with it.
At the end of the day, I want you to own your machine and I want you to solve the problems on it yourself. That should be our “prime directive”. But I will argue with you enough to at least plant a seed that something is wrong.
To your point, I was being general. After you have a consistent machine, things like adjusting for belt stretch and surfacing your spoil board and finishing passes can greatly increase the accuracy without breaking the bank. I know you also use the steps/mm to get in the ballpark on prototype machines, and that seems fine too.
There are cases where it makes sense, and there are cases when you are solving the problem in your test, but making untestable situations worse. Is that nuance a bit more agreeable?
In this case we needed a standard to test using faulty cad was the problem we need some type of standard that has been tested on working machines and can give feedback
Even just using the command line and giving a x250y250 would have worked
Till the hardware is confirmed the gcode cant be trusted either
We have the crown how about a square and circle that are known good
I need to start linking Jamie’s calibration tool.
1 Like
Yes perfect just what we you need. 
Okay, I see a lot of “ifs” in there. First, I don’t use steel belts with a break in them… I don’t use obviously broken parts. And, I don’t test in just one section of an axis… I recommend, as soon as possible, to expand that commanded test over as large a movement as your axis allows. I don’t allow slaved motors to get – and remain – out of sync long enough (it’s pretty obvious when only one motor is working) to try to make a commanded test on that axis… that needs to be fixed. Again, the key is to see REPEATABLE motion on a mechanically sound axis for a given command before making any steps/mm adjustments.
Not sure about “wheels translating sideways”. I assume you’re talking of LR2 or rolling gantry-type machines… again, I recommend commanding a test as large as a given axis allows.
If I bought and used 20T pulleys, and calibrated dead-on, and the torque was sufficient for the machine and tooling I’m using… who cares if it’s “25% less torque”, as long as it’s enough?
I agree with you about the machined and electrical parts… motors, pulleys, stepper drivers… those are usually pretty accurate. MOST of the variability will occur in the softer, more flexible parts… where the belt tension and/or the wheel/bearing friction you have set in will have an impact on your particular machine.
As you know, I’ve generally moved away from building a stock MPCNC or LR2… not because they are bad designs or deficient in some way but because I’ve built enough of them that I want to try something new… that’s the old engineer in me. So, I’ve printed a R&P set off of Thingiverse and incorporated it into a basic MPCNC Burly machine configuration. I’ve download the firmware for that machine and… guess what? I don’t know the exact pitch of the gear teeth… but I can measure a commanded move and adjust the steps/mm to get the correct amount of movement out of the axis. And I printed rulers to prove it. And I still don;t know the pitch of the R&P…
I know you are trying to be helpful to as many people with stock machines and typical use cases as possible. I appreciate that… you are exactly what this forum needs. And you can run circles around me technically – I’m just a dinosaur among thoroughbreds – but I do remember enough to have the confidence I can pick up and reuse pulleys, motors, belts, wheels, controllers, etc from my stash – don’'t really care too much about their “specs” – and can put them together into a machine and get that machine to move the correct distances when commanded. Move, measure, adjust… and repeat… BOOM! Works for guided missiles, works for me.
1 Like
Hahahahah a love this, gonna learn more about guided missiles just so I can use it!
Really, as an outsider - ish on this convo, would either of you (dk, Jeff) agree that the only real contention here is whether all things have been verified before adjusting steps per mm? Seems like dk is presuming that all parts are good and properly assembled already, and Jeff is trying to ensure that (often new and inexperienced) operators have properly assembled good parts before moving on. Maybe I’m wrong, but I think you’re both on the same page. If all parts are good and correct, the adjustment may be necessary (dk) but should be very small (Jeff).
I just wanted to throw that out there because it may not be obvious to folks who couldn’t identify most of these parts a month ago (like me when I built my first mpcnc).
2 Likes
I didn’t say you had to run the whole massive job at that speed 
, just that in the test to resolve issues that backing off the feedrates and doc may eliminate some unknowns.
Fwiw I usually can run around 8-10mm/sec at a 4-5mm doc with a sharp 3.18mm bit without any issues. With a few test cuts I would imagine you’d be able to find a sweet spot in there for time management, and if it’s really thin material reducing the number of passes by cutting the full way through in one go would be a big time saving, same with a full depth finishing pass.
1 Like
I have nothing but respect for you David. I hope I haven’t offended you. I have never meant to say, or imply, that when you adjust your steps/mm, you’re doing something wrong. DKJ is master.
The point I was making with those “ifs” is, in this machine, if you’re off by 10%, the problem isn’t steps/mm.
2 Likes
You’re just worried because he knows guided missiles… 
1 Like
I use a free plugin where you just select a face and go to File → Export DXF or STL …
I think this is it: https://www.guitar-list.com/download-software/convert-sketchup-skp-files-dxf-or-stl
Unfortunately it looks like it’s only available up to version 2016 which is what I happen to use and refuse to “upgrade” from so maybe it’s moot.
Ordinarily, waking up to a lengthy something I’ve written late last evening, while “under the influence” (JD is my preference…)… I’d feel rather bad/guilty about my involvement in a lengthy exchange/debate with someone. But in this case, Duncan (the OP) has literally invited us to speak up with his thread title… “How accurate can I expect the LR2 to be?” So, IMO we’re still “on topic”.
Here’s what I see… Duncan has asked about “accuracy” and “expectations” for his LR2, a larger-format machine.
If you build an LR2 as your first CNC machine and you just want to see what all the fuss is about… you can use all stock parts, assemble it with care, use the stock firmware with its round-number steps/mm… and get “pretty close”, maybe within 1%? That probably meets or exceeds your [little/no] “expectations” and if that’s good enough for you and any use case you can imagine… you’re done. NO CALIBRATION NEEDED. You can have a lot of fun making all kinds of stuff that doesn’t require extreme accuracy.
But if you’re building a custom sailboat and can’t tolerate (your “expectations” are pretty high) every 8’ piece of deck/hull planking you cut being off by the better part of 1"… then you need to calibrate your machine. AND – GOOD NEWS! – IT IS CAPABLE OF BEING TUNED FOR BETTER PERFORMANCE THAN STOCK… if you are willing to test and adjust; i.e. get it “dialed in”. This is CALIBRATION.
So, IMO it depends on the person. What are your expectations for the machine… why did you build it? And how accurate do you need this machine to be… and at what point is it “good/close enough” for you?
Yes, I suspect we are talking about the same thing… just different stages of machine construction and operation. I’m addressing the particular machine in front of me, that is basically up and running, and asking what I can do to improve its performance. Jeff, I think, is concerned with the machines of the “masses” who come to our forum, with a machine under construction and/or having genuine problems that need fixing. He’s troubleshooting and being helpful to others… helping folks get their machines up and running. He’s addressing the machine’s basic construction and operation… I’m addressing the next stage, the machine’s calibration.
Again, no disrespect to Jeff, or anyone else here… and I’m not offended when somebody pushes back a bit. We all love this stuff or we wouldn’t be here. I’ve worked in industry and as a hobbyist… and been involved in making stuff most of my adult life; i.e. I’ve been around the block a time or two. So has Jeff… and so have many of you. We’re all friends here, with varying backgrounds, and I think it good to share our experiences… that IMO is what a forum is all about. So, keep it civil, share your experience, and always see yourself as a “student”… that way, we all learn and grow and have tons of fun interacting with one another.
Again, my $0.02.
– David
3 Likes
Please, Jeff… I’m not offended. Our respect and admiration for each other is mutual. We’ve been civil and this has been healthy, instructive, and constructive IMO.
That depends, IMO. If I knowingly (or perhaps unknowingly) use/substitute/interchange 16-tooth and/or 20-tooth pulleys – knowing that my MPCNC can be made to operate properly with either one — is my MPCNC broken, if I’m not using the one recommended? IMO it’s not “broken” and not a “problem”… my initial motion tests might be “off” a lot more than I expected but all I have to do is change the steps/mm to match the pulleys I’m using. Or, maybe I’m using different drivers… or different micro-stepping … than recommended.
My point is that nothing has to be necessarily “wrong” or “broken” to see pretty significant amounts of change in the motion tests… and maybe all it needs is a steps/mm setting that matches the components and jumper settings actually used. As long as I use components and settings that allow my machine to do what I need it to do… then it’s operating properly and nothing is “wrong” or “broken”.
I was surprised that Duncan’s tests were off by so much and he had to adjust the steps/mm from 100 to 90-ish… maybe it’s something akin to what I’ve suggested here?
– David
All of this exactly the reason I even understand what y’all are talking about here. Even if it WERE off topic, it’s valuable. I knew zero about any of this when I got started with my mpcnc - maybe even less because I didn’t know what I didn’t know.
Thanks to both of you in particular, and everyone else in general. Oh, and K especially for reminding us all not to mess with the guy that knows guided missiles.
2 Likes
He said the problem was the 64mm circle in sketch up was being scaled funky from cad to cam.
1 Like