jleichne

I'm in need of such fine precision since I am working towards tissue & organ printing. I was hoping instead to use this section of the forum to discuss the advantages of each of your individual setups, in repeatable mathematical ways so as to truly test if, for example, direct drive will improve my variability enough to warrant setup. J

Oops - can a mod please delete the other topic, it seems to have duplicated after I went in for an edit and added a few numbers that I had left off the first post.

Hey there guys, I was here a while ago asking around to find some hard numbers on the printers accuracy, precision and backlash. Nobody really had much, so I picked up a gauge dial and did a series of measurements on the three axes. I was hoping to begin a trend here where people can measure and post the qualities of their machine, in order to begin devising a database of average assembled machine quality. In addition, this can really help give some true figures on the advantage of the direct drive design, the GT belts and pulleys, or any other combination of parts. So I guess I should start with my data It is very important to mention here that I was using at this time a gauge dial with 12 um resolution with about 15-25 um of hysteresis (aka the normal $40 dial you find on amazon) Since then, I now own a $350 dial that has 1 um resolution and only 3 um hysteresis, but I haven't retaken the measurements yet. For motion in the X or Y axis (my data was about the same for both): Protocol - tell machine to move using GCode by setting steps per mm to 16 and then moving the appropriate number of steps. Measurements repeated ten times. Full Step - (i.e. 16 steps): Expected Distance: 0.203 mm Average Distance - 0.322 mm Error in Distance Travelled as % of Avg Distance - 37.62% Std Deviation as % of Avg Distance - 3.71% Half Step - (i.e. 8 steps): Expected Distance: 0.1016 mm Average Distance - 0.1629 mm Error in Distance Travelled as % of Avg Distance - 36.93% Std Deviation as % of Avg Distance - 9.64% Microstepping (6 steps used here) Expected Distance: 0.076 mm Average Distance - 0.1245 mm Error in Distance Travelled as % of Avg Distance - 38.78% Std Deviation as % of Avg Distance - 15.91% This data points out an important trend - the increase in step variability as you move from full to half to microstepping. The 36-38% consistent offset in the desired motion was surprising, but it is clear that the consistency of that across all three trials demonstrates the difference between 'accuracy' and 'precision' here. In addition, I looked into my backlash, and it averaged at 773 um, with a standard deviation of 133 um. For this measurement, my protocol was to move on one particular axis and measure at that position. Then the machine would be told to move right 10 cm than left 10 cm. The difference would correspond to the backlash. After I finished these measurements I began the search for a ballscrew and recirculating guide system, since I need this box to move with about 4-5 um resolution with about 3-5 um of accuracy. Let me know how your box holds up! If other people are willing to take the challenge I'll even retest mine with the more accurate dial in order to get more precise results.

Hey! So I want to remove the Bowden tube from the hot head so I can attach it to a new mount for a syringe-extrusion setup. How do I go about pulling it out? Do I need heat? Do I need to hold down a little white thing like when you remove the bowden from the back extruder? Thanks! I'm cautious about pulling too hard before asking the community. Jared

Hey yall' Two things: (1) I'm wondering if there is a good online resource for understanding the numerous commands in GCode. I know there is a list of all the commands but I haven't found any simple tutorial (although if there isn't I'm ready and willing to dissect the GCode of basic circles and squares for starters) (2) I also want to update the firmware for Marlin, I've used Arduino environments before but before I start wading through the source code blind I would like to find a good resource. I plan to add additional printheads and I want to be able to add in the functionality of moving to a particular XY position and sending a trigger signal to the additional printhead (piezoelectric dropper) extrusion electronics. Thanks for the help! I'm hoping there is a good place for me to start but I'm ready to start making my way through it anyways if there is not. J

Based on my calculations, 78.74 steps per mm is equivalent to 12.7 micron steps. Is this with or without microstepping? In addition, the largest source of error is probably the backlash (which I wouldn't be surprised if it was anywhere between 15 and 50 um. I want to switch out the pulley system too and use ballscrews with minimal-backlash nuts. I'm okay with near-perfect positioning at the cost of speed.

To make sure I understand you correctly, you are suggesting removing all the belts and making the two axes parallel to the two ball screws driven by the sheer force of the ball screw on the opposite side (meaning I will need to replace the opposite wooden blocks with something on a super frictionless bearing). I feel that while this will better level the printhead, it will probably cause strange torque beyond a particular load I am applying at the printhead. (Still need to run simulations) This is more or less my current plan already - although since I'm going to be reworking the frame in acrylic and since I'll need to buy new stepper motors that can handle high humidity environments I'm planning to keep the Ultimaker on the side and use PLA to build little bioreactors. I like the idea of just using motors with twice as many steps for my Ultimaker (although I still want ball-screw repeatability for my cell printer). J

True - I've needed to adjust those, but the grinding sound was happening only during the initial motion towards the center of the bed after homing. (but its now gone away as mysteriously as it came)

How low can it go? I've seen the difference between the 0.2 and 0.1 setting, but how low can it be put? At some thinness I presume the error in filament regularity overpowers the resolution - but where is this limit? J

Alaris: I didn't plan to add any extra axes. I will add a photo later when I have more time. My plan is to simply replace the particular X and Y axis that is timing-belt driven with a long ball screw. A similar belt system will pull the block on the opposing two axes. This will combine direct drive with a ball screw and I believe the two axes driven by belts will have much less of an impact on error than the ball screw. Snowygrouch: I'm happy to see that you've also considered this. I agree that its definitely not a speed-friendly approach, but for my purposes (small-diameter arteries), I don't need to go too fast. I feel like a ball screw would have incredibly more positional accuracy (the pricey ones at least), so a rotary encoder on the coupling should be unnecessary, right? MSURunner: Yeah, I will need to rework the slice software to automatically parse instructions to separate printheads and yes this is essentially a cell dripper. This is why my main concern now is my underlying XY positional accuracy (which I should hopefully be reliably testing soon). Are you saying that I can avoid belts by using the screw to push both blocks through the central bar? I was concerned about that approach because I felt that the belts would be needed to prevent any strange long-term axis deformation. I suppose I'll need to model it first. I think that if users are going to make significant mechanical updates to the device, there should be a section of the forum for people to design replicable experiments to properly characterize improvements in the positional accuracy or mechanical improvements. This way we can over time build up metrics of proper device performance.

Hey yall' My printer used to be perfect and quiet. Real quiet. I've only had it assembled for a week and everything has been great. Two things have happened recently: (1) I set my printer to the bottom for the first time and the bottom Z switch didn't activate perfectly, creating an ugly grind sound for the half second it took me to turn off the machine. (2) The timing belt pulley on the Y axis became so loose earlier that it messed up a print. I retightened it with amazing results at first. I just found it slidable again only a few hours later! So here's the problem: Running Cura, after the initial 0-0 homing, the printer GRINDS REALLY LOUDLY moving to the first print location. It's fine afterwards noise-wise (although the prints aren't coming out right again yet). Anyone have any ideas? I'm pretty sure item #1 up there is definitely not causing it, since any Z-translation doesn't cause the grind until the simultaneous Z and XY found in the post-homing movement in Cura. J

(3) Keeping the timing belts and replacing the pulleys on the main XY axes with ones that are twice as large in order to double the theoretical precision of the device.

I agree that direct drive is a crucial step in removing an unnecessary source of error. How do you feel about the following two adjustments as well (in tandem with direct drive): (1) Using Ground Ball Screws to drive two primary axes and the belts and pulleys to respectively drive the parallel two axes. This may drastically raise print times but do you believe it will achieve a more reliable, consistent accuracy over solely direct drive. (2) Adding two more vertical beams like those to the sides of the main vertical leadscrew at the front of the device to better stabilize the platform J

Thanks for the quick and thorough reply, however a few things: (1) "The bed moves around about half a mm at the front under severe jerk forces during printing" Is the jerk force in the direction of the Z-axis twist? If it was possible to add two more rods and linear bearings to the front could this be accounted for and removed? There's definitely room for it, that's for sure. (2) "the frame is made of wood, which will be expanding and contracting with humidity all year round" I will be recasting all of it into acrylic so I can do cellular printing. I will also be changing the printhead to a cell droplet mechanism so the filament thickness doesn't matter here. (3) I would honestly just make yourself some nice geometric primitive shapes, print them and measure with a nice set of Mitutoyo vernier calipers, which are far more accurate than the error of the machine itself. I don't think that printing filament would be a good calibration approach, since not only does it shrink (even if only a little), but the thickness of the strand I print is far larger (I believe) than the positional accuracy I am trying to measure. I don't see how I can design a gel-droplet slicer software that will provide me prints that are accurate enough to facilitate cell-cell recognition unless I am positive that the positional accuracy of my printhead is within reason compared to my droplet size. My only point of concern is that I see no evidence of any record of positional accuracy measurements (except for a random article I stumbled upon on the topic of Ultimaker calibration but was completely in German :\ ) Thanks for your time, J

In another post, I tried to describe a mathematical model to describe my error: Final Position (along one axis) = (Xo) + Etiming + Elong Xo = what software considers the initial position Etiming = Positional Error due to (1) Slip of timing belts and (2) Improper Orthogonality of Timing Belt to Long Belt Elong = Positional Error due to (1) Slip of long belt, (2) Improper orthogonality of both long belts (Also errors due to parts not being screwed in tight but is removed due to simplicity) Would you agree with this simplistic description? I also came up with what I consider to be a reliable calibration method: UV LED coupled into a fiberoptic cable that is mounted on the moving head. The beam diameter of the resultant light determines the precise resolution of this method, but with proper optics can probably be put to at least 10um without significant costs. On the build platform, place a sheet of the UV-reactive copper used for etching circuit boards in the 70s. After drawing shapes using Gcode, put the copper under a microscope (with calibration distance lines built in) in order to properly test positional accuracy. How do you feel about this? I would be very willing to put one of these systems together if it means I can get some real numbers behind my positional accuracy.

Snowygrouch - Any way to demonstrate your improvement in XY accuracy with a repeatable test?

In addition, I just read SnowyGrouch's excellent Direct Drive modification here: viewtopic.php?f=6&t=1429 Lets say we describe my accuracy as a mathematical model, which goes like this: Final Position (along one axis) = (Xo) + Etiming + Elong Xo = what software considers the initial position Etiming = Positional Error due to (1) Slip of timing belts and (2) Improper Orthogonality of Timing Belt to Long Belt Elong = Positional Error due to (1) Slip of long belt, (2) Improper orthogonality of both long belts (Also errors due to parts not being screwed in tight but is removed due to simplicity) In a simplified model, can this be described as the primary source of error? I understand Snowygrouch's technique may introduce minor error relating to alignment of direct drive motor with axis, but we can neglect that too. Does this make that machine substantially more accurate? I feel like we need positional accuracy testing metrics. The gear-ratio approach may include the extra source of error, but maybe by lowering my step size enough it is a viable method to achieve my desired resolution(?) J

Why would they be working harder to complete prints if I am printing extremely slowly to get the accuracy I desire? Jared

A few questions: Have you ever calculated the precise accuracy and long-term repeatability? When you say 'Gearing on the Steppers', are you implying that if I move the XY steppers over laterally and make room for a certain size gear I can use gear ratios to improve my baseline 'microstep' size? Is the microstepping of the motors embedded in the firmware or the software? I'm assuming that the microstepping capabilities are probably already maxed out as much as possible by talented electrical/software engineers so maybe I shouldn't play with that. Another concern is that in modifying the structural loads at the printhead with a different extruder, I'd need to recalibrate the 'step to um' plot since I need to precisely speed up and slow down the motor per motion in order to make sure the printhead stops moving in the perfect spot. I'm very tempted to hook up a linear encoder system so I can know exactly where the head is at all times. I'd probably need to rewrite all the software and firmware though to properly account for new Force-Distance curves, etc. Other options include radio frequency modulated visible light sources or triangulation systems. So nobody has every tried to supercharge their Ultimaker accuracy with an awesome positioning system? Thanks, Jared

I've been trying to decide on the best way to properly characterize the positional accuracy and repeatability of my printhead. My ultimate goal is replacing it with an alternate printhead for cell hydrogels so I'm not concerned about parameters relating to the fiber extrusion (although I'd like to work on a multi-spool feed system one day). I came across a good post by Daid recommending attaching a pencil to the printhead - have any of you tried this and what have you found? I'm trying to decide whether its necessary to install a positioning system to ensure that I can get below 20 micron positioning accuracy. Any thoughts on this? J

Hey yall' I've only had this printing for like two days but already I have been impressed by the ability of the built-in retract mechanism in Cura (which I have barely played with but was able to print the vertical DNA model rather accurately). Here's my question: When I warm up my head, it always starts pushing out tiny strands of filament. When it begins printing, I always need to make sure that the filament is a certain length so that it does not mess up the first layer and hence the whole print. Do you all just do what I do, and pull it off real quick right as it starts printing? Jared

Hey there Daid! I gotta say - I'm impressed that you took it upon yourself to try to write custom slicing software. I've just received my ultimaker recently and my application requires a custom slicing module that allows me to generate particular sequences of G-Code based on predefined geometric parameters to guide a modified multi-headed system. The only problem is that it means I need to write a 'slicer' of sorts that works with predefined geometries. I'm excited to possibly contribute to this work once I understand exactly the protocol in which G-Code can be generated from a 2-D slice of a 3-D structure. Jared

--- No Longer Applies ---

Hey all, My front sliding block, when assembled like this: http://www.flickr.com/photos/ultimaker/6237389712/ has almost no clearance with the top of the frame. My sliding blocks don't exactly sit perfectly vertical (although I think that changes once I tighten the belts) but the slight lean causes some undue friction on the top of the frame when moving the tread by hand. Will all of this be fixed once I tighten the pulleys later? Thanks, Jared

I took apart the box and carefully reassembled it and it is sitting much more flat now. Thanks for the tips - I suspect it was the black cable guides. Time to put in the motors and axes! Jared