I spend a lot of time in the forums, and most people don’t care if they asked for 500mm and they got 500.5mm. More often than not, someone is trying to use steps/mm to adjust it when they asked for 500mm and got 505mm. And then they have more trouble when they adjust it and it is sometimes 495, and sometimes 500mm. The root cause is something like a loose pulley, or a stretched belt (one where the fiberglass or steel has stretched). It ends up being very hard to diagnose when they have adjusted the steps/mm.
I’m very glad you are happy with your adjustment, and I encourage anyone who is down to that level of error to measure and adjust it themselves. But am of the opinion that (without context) it will do more harm than good in the instructions.
I’m not a dictator though, feel free to disagree with me. I can accept that.
True words. Same if one stepper stalls because of a crash or something.
Imho squaring the core is a benefit. If you square a non squared core with dual endstops (which i did not installed yet) it will put tension on this. This releases if you cut power.
But hey, its plastic and maybe it doesnt hurt, but definitevly will not help for getting the last bit of precision.
If you power down and back on, you could avoid homing, if you dont have tension compensated with the steppers.
But there come the other issue:
I also thought about this.
Maybe this for the last bit of fine tuning: energize the steppers (but do not send any movement command) so they snap in at a full step. Than adjust the belt position (not the tension) on one side to the desired position. Not sure if this works in every case after powering up.
But I think dual endstop squaring is a must in case you want precision.
Compensating dimension errors caused by deflection, wobble, vibration, loose parts, wear or whatever with the steps per mm is wrong and will cause unexpected things.
Compensating the belt stretch with the steps/mm is neccesary, if you aim for precise parts.
My very first project on my machine was an plywood box with interlocking parts, that have to fit exactly.
After adjusting the steps/mm I got all dimensions below 0,1mm tolerance and it fit all perfect.
Before the adjustment I could get it precise either on inside or outside distances by adjusting the tool diameter, but never both.
That drove me crazy and I started to doubt that the mpCNC concept is suitable at all for my projects.
This should be avoided at other builds with a small note in the instructions somewhere. Maybe an extra finetuning section? But first, we should collect finetuning ideas somewhere.
I think what this really boils down to is the user. I have been a user for 5+ years and multiple machines. I have never been unhappy with my accuracy or precision (other than human error). I have also never checked the accuracy of the belts to the degree you or Jamie has.
I think anyone chasing ultra precision on these machines would start with verifying there axis travel and would understand the need to fine tune and adjust everything down to run out compensation as well as backlash compensation. Meaning if you want ultra accuracy beyond what this was designed for you will be checking everything and adjusting everything any way as you do with any machine you use to that degree of accuracy (to the point some also record temperate at which measurements are accurate).
I will be sure to add a note somewhere for those seeking this sort of accuracy but +/-0.1% is well within our user bases expectations.
I think I’m pretty much in agreement with Stefan on this. Basically, I believe that when I build a machine, I want the motion system to accurately and consistently move the empty carriage – i.e. unloaded – the exact distance I ask for, every time. IMO everything else is CAM/extrusion/rigidity/power/etc related. Hopefully I’ve built a machine with sufficient rigidity and power to do useful things with the loads I subject it to… if not, I just hang a laser on it
I also realize I’m the odd duck, proof-testing my machines by creating rulers with it… and I can easily see the difference between 100 and 100.5 steps/mm. I also love Jamie’s test pattern rulers which also allow me to see evidences of backlash… or loose grub screws I also understand that many/most folks don’t need that kind of accuracy, and 100 steps/mm is just fine. No argument from me.
I am not generally opposed to correcting stretch with a steps per mm correction, but I don’t recommend it as general advice because novices will often misidentify a different error like runout or backlash and try to correct it with a steps per mm correction which in those cases makes the problem worse.
That is some, maybe even most of the reason for the test pattern generator that makes rulers: to truly identify a length scale error and separate other effects like backlash or sticky, uneven movement that could be misidentified as a dimension scale error.
I, again, agree with Stefan… as belts are part, stretched or not, of the unloadedmotion system they can, and should, be compensated with the steps/mm setting. All those other things he mentioned are CAM/tooling/load/wear/etc effects and are not an inherent part of the motion mechanics and should NOT be compensated with the steps/mm setting.
IMHO Jeff and I established some time ago that our “target” audience was different, though sometimes over-lapping, and our primary purpose for our postings/advice are often a bit different. He’s the exceptionally generous, ace troubleshooter, trying to help the relatively inexperienced “many”, new, maybe first-time, users get their machine built and running as Ryan designed it. Keep it stock and let’s get it running. He, Ryan, and many selfless others are what makes this forum special and great.
I OTOH am just an old, experienced, builder who’s picked up a few “tip and tricks” along the way that might be helpful to someone trying to get a little more out of their machine. I’m not nearly so technical as Jeff,/Jamie/Stefan/Ryan/etc but have been doing this stuff long enough that I really do find this stuff “fun” and a bit second-nature… and tune in to hear what they have to say. A lot (most) of what they talk about is over my head but the old engineer in me is still trying to pick up a trick or two from them.
I do also believe that a section of the docs should target the more experienced user and provide information for those trying to get a little more out of their machine. But how you keep new and inexperienced users from inadvertently finding their way into a possibly confusing (and maybe even contradictory) maze of information is beyond me. I’ll leave it to you guys to sort it out
Okay. I will not let this go yet. So we are in agreement we have a precise system, this could bring up the accuracy for free. A small percentage but if we truly have a systematic error, might as well take care of it.
I have a couple machines here, I will run a test ruler or two. If I measure something similar we can figure out how to proceed.
You all know how much I hate to change things like this, and the implications I am most worried about are people updating their firmware and running previously validated gcode and getting new results. This would have to be done with care. I flipped one of the axis directions (to decrease marlin manual edits) early on, made a huge note in several places, and still got a bunch of crap for it.
My Primo is slightly more than 50 cm X 50 cm and it was accurate at first but developed a curious tendency to add 5mm to the first 50mm of movement. I was using steel reinforced belts because… why not? But when I posted my dimensional error problem here, it was suggested that the steel reinforced belts were the cause, so I got fiber belts and my accuracy is now nearly dead on. I tried measuring the old belt but decided counting all those little ridges was an exercise that wouldn’t improve my health.
Yes, its a bit more accuracy for free.
I think the thesis you made earlier count also here.
User with higher demands on precision will easily understand why and where to adjust the step/mm parameter.
No matter what you decide…updated firmware or not…a note about the issue is necessary in my opinion.
I also understand that many/most folks don’t need that kind of accuracy, and 100 steps/mm is just fine. No argument from me.