Adventures in Measurement

[@MrNlce30]

Your commitment to pen storage is commendable. Keep up the good work.

[@smert_okupantam]

2:40 I would at measuring Start use the yellow button ZERO on caliper, so i would not need to subtract

[@JohnJones-oy3md]

Thank you for showing the methodical steps to fine tune your results. I for one would have never even considered half of the parameters you tweaked. I've learned a lot today.

[@mogelzmo]

i woud calibrate in a 45° angel cause when you move just right or just left on core xy both motors turn // but maby im dooing sth wrong when I calibrate my printer that way

[@zzing]

The way you measured x and y seems to me to be prone for inconsistency between measurements.

[@countk1]

I have to agree. You need to be absolutely sure you're measuring exactly parralel to the axis and that is not the case here. You also need to make an extra travel to take out clearance on your travel direction. If there is a difference on these measurements, you need to get on that issue as well.

[@zzing]

@David I would measure your reply against the fact that this is the same person who made a rather impressive and popular printer over more than a dozen videos.

[@havenview]

@David His results prove otherwise. Its a home made 3D printer being used as a CNC, his technique is more than adequate for the expected tolerances

[@ovDarkness]

@@zzing "popular" doesn't mean "good". There are quite a few errors in this design. Some are obvious, some derivate from low quality of build materials and lack of equipment (as this chinesium calipers that can haave about 0.1mm error on the farther end). Some can be connected to extremely low budget build. People somhow got the idea that FDM/FFF 3D printers are cheap and easy to build. This is only when you're not building them right, or you aim just for simplest materials like PLA or PETG, or you've decided to ignore physics.

[@zzing]

Wojciech Lewenstam you do realize that everything has a budget right? For the vast majority of people in this hobby this machine is an upgrade. If there are problems on this build help us fix them. Not everything has to be perfect.

[@Redtv12Gaming]

Why did you not round the steps

[@linecraftman3907]

For precision

[@Nappen87]

If u get diffrent movement values in x and y u need to tighten the belt. because then u will have uneven tightness.

[@pen25]

one other thing i noticed. to prevent tearout use a down cut end mill

[@das250250]

I think in fusion you can use parameters measurement variable to dynamically adjust the tolerance and keep your drawings accurate dimensioned to transfer to another person or machine

[@ExtantFrodo2]

If your runout is indeed exclusively from bring off centerm then you could shim the short jaw of the chuck to bring it true using an appropriate thickness clip from a feeler gauge (or a piece of tin foil (folded if necessary)). You can't do this if the runout is conical from an angle change to the tool in the chuck. If you have or can borrow a test indicator you should be able to determine which jaws to alter to fix this issue if it concerns you.

[@Tech2C]

That's a good idea. I could place a small strip of kapton tape along the shank of the end mill and rotate it on each jaw to determine if I can eliminate the run-out.

[@ExtantFrodo2]

@@Tech2C Also, I don't know how deep your chuck hole is, but inserting the tool as deep as you can never hurts. (Some even find it pleasurable.)

[@bbogdanmircea]

Nice, when I reassemble my second Hypercube again I shall make it a CNC too! That base is still too narrow lol!

[@metallsimon]

shouldn't the x-axis and y- axis steps/mm be the same on a coreXY System?

[@muski282]

Yeah, but due to manufacturing variances e.g. in the belt pulley radii on the stepper motors, they may deviate.

[@alperenarslan6036]

what you think about marlin bed skew compensation

[@roidroid]

This is a more accurate way to measure sizes: You *make 10 of them,* butt them all up against eachother and *measure the entire block,* divide the result by 10 to get your answer. To measure your kerf-width holes more accurately: Cut out a long rectangle, measure it, then use the router to cut that rectangle into 11 separate side-by-side pieces (ie: make 10 cuts). Put those 11 pieces back together, all butted upto eachother, and measure that whole thing again. Subtract those 2 measurements from each other, then divide it by 10 to get your answer. Because you're measuring something much larger, you get a much bigger surface to measure, it's thus easier to keep the calipers straight & square. Also any small cutting errors from the router/printer are being averaged out over 10 results. Learned this from Ponoko.

[@BikersGarage101]

As usual, very nice video. If I may. There are a couple things going on here: 1- Measuring technics need revision. Points of contact for your ref measurements should be flat, parallel/perpendicular. Using the bars and ref points that may vary slightly from measurement to measurement will introduce errors. Try fixing the calibers or better yet, use a dial indicator with a dial indicator holder designed to clamp on any repeatable reference surface. 3- The calipers may be off too! You would be surprised!!! Knock-off digital calipers may be off by as much as .04 from measurement to measurement and this is a huuuuuuge 4- Repeatability. This is the point that concerns me the most. Identical parts should have identical measurements in such short runs and using this soft materials. Something is going on there 5- Measuring the slots like you have is not going to give you accurate results at least not to the level that you are chasing.

[@prxrb]

Is there a good reason to empirically calibrate X/Y step-count instead of using theoretically calculated values given by motor_step_angle * pulley_tooth_count?

[@the-mush]

yes, it is called "manufacturing tolerances".

[@prxrb]

​Actual position of the hot-end should be basically step_angle_radians*steps*Pi*pulley_diameter + error, where that error is due to tolerances, backlash, belt stretch etc, and is likely normally distributed with an approximately zero mean (or at least, a mean that does not scale as a function of the desired travel distance). It seems to me that by empirically calibrating step-count based on single measurements done at 100mm, you do two things. 1) You take whatever error happened to manifest at a short distance, bake it into your firmware, and literally multiply it by 3x for full-distance, 300mm travels. 2) You also take that random error that was observed from a single trial, and bake it in so that all subsequent travels are systematically off.

[@the-mush]

@@prxrb seems about right, but how would you determine the error component of your equation if not by empirical measurements?

[@chloemcholoe3280]

use the flexible shaft to drive your extruder!

[@samc5898]

Very interesting. Goes to show how much the details matter

[@miketorri4776]

Hi, I 3D printed and build you Hyper Cube. I have build other printer before. But, this is the first one that has Auto Bed Leveling. The inductive sensor is the same as yours. Marlin 1.1.9 is configured from your two sample files. I am quite savvy with Marlin. But, I need to tune in the Z axis. With bed leveling feature "G29". The Z height is always two high to print. Do you have another Video to do this calibration on your machine? I did try using C Riley bed leveling You tube video. But, I think I need more detail on how to do Z Cal on your design.

[@mezje]

Great video, subscribed! The rounding did get me though :P

[@avejst]

Nice, thanks for sharing😀👍

[@bigmac1598]

Morning. I think you need to check how vertical to the table your driver head is. If it is 0.1° off the cutter at depth will produce a tapered slot

[@pen25]

is that a taper endmill?

[@chriswisneski6071]

Good info - glad to see you were able to solve the accuracy issues of the milling operation, thanks for sharing. I like your videos because I always learn something. One minor point though - run out is generally measured as a difference in radius, so technically you had 0.115mm run out, not .23mm. Didn't affect anything in this case, but if you run into a problem where the runout is to be plugged into a standard equation (or software), it will cause you problems.

[@Tech2C]

Ah yes that makes sense. So it's out by half the error.

[@HevTec]

I think you can't calibrate the steps this way on a Core XY system. Let me explain. If you choose different steps on the X and Y axis, they will be more accurate, but in the printing or milling process you will get angle errors on the part. This is due to the fact that one motor takes more steps and the other less. But since both motors are responsible for moving the X and Y axis, both motors have to take the same steps. Ich denke du kannst die Steps auf diesem Wege nicht auf einem Core XY System kalibrieren. Lass es mich erklären. Wenn du unterschiedliche Steps auf der X und Y Achse wählst fahren diese zwar einzeln genauer erzeugen dir aber im Druck oder Fräsvorgang Winkelfehler am Bauteil. Dies ensteht da durch das der eine Motor mehr Schritte macht und der andere Weniger. Da aber beide Motoren dafür veranwortlich sind die X und Y Achse zu bewegen müssen beide Motoren auch die gleichen Schritte machen.

[@ronversteeg4627]

What plug-type is used for the changeble print/cut/laser head ??

[@Tech2C]

mpx

[@jmtx.]

Thanks for an excellent calibrations video. I had wondered about how CoreXY steps got calibrated, now I know - math is dealt with by the FW.

[@3dprinting4life10]

What program do you use when using your 3D printer as a CNC, thanks for your usual good videos, regards RobUK

[@speesy]

thank you! very informative.

[@ar-mory]

Outstanding!

[@mikeflight9]

Excellent information. It just goes to show that calibrating your 3D printer is very important for engineering pieces. :D

[@Tech2C]

Yep ;)

[@keyz182]

Note: I am not a machinist. From my understanding, the Collet type chucks have less run-out than the type of chuck you're using, so maybe switching to one of those would improve things?

[@ronversteeg4627]

Very nice and informative. I like all your video's and have build an Hypercube which i am calibrating now. One remark: pencils should be stored tip-side down. As ink is going to the down-point en your storage makes the tips run dry... :-)

[@Tech2C]

Oh that is a good point! Let's rename it from a pen holder to a pencil holder instead ;)

[@dougingraham5807]

As was mentioned in a previous comment, you probably should not be correcting the X and Y motion by futzing with the steps/mm setting. If set at 80 this should be quite accurate. If not you need to look into why it is not. Assuming everything is working properly (not skipping steps) then the error could be in the motor pulleys or the belt tooth spacing. Since you were seeing it short in both X and Y cases I am guessing the belts are not under enough tension which was also mentioned in another comment. Changing the steps/mm is a temporary solution at best as the length of the belt is set by the reinforcing material when under correct tension and the rubber when not. Correcting belt tension will also help to reduce ringing which is the primary cause of ghosting in 3d prints. As for dealing with the run out issue, eliminating the 3 jaw chuck by switching to collets should help greatly. And I did watch the previous video where you switched away from a collet setup. Because your cutting speeds are low I would mount the router motor directly to the gantry and eliminate the issues with the flex head. The flex stuff is meant to be hand held so there is no need for precision when they design/manufacture them.

[@das250250]

The measurements are very consistent in the x y . A tad too big in one and a tad too small . You could adjust them to be bang on .

[@Tech2C]

Yes keep iterating the calibration until it's perfect.

[@bardenegri21]

Wouldn't this be more accurate if you measured cuts made in a plastic rather than wood?

[@das250250]

Did you test your micrometer against a known accurate measurement block ? Otherwise how do you know all your measurements are calibrated?

[@Tech2C]

Good point. I guess we can say the machine is now calibrated to the micrometer used.

[@faxxzc]

First of all, you do not calculate steps/mm by the pulley diameter, but by the Number of teeth and belt pitch. Second, the way you measured the travel distance does not account for backlash. the correct way would have been to 1) move the head while the caliper is in place 2) first move 50mm to take out the backlash, then zero the caliper and then move another 50mm. If the first 50mm are not equal to the second 50mm, then you have backlash. 3) use the second 50mm to measure your steps/mm.

[@Tech2C]

Prudent method to calibrate. Thank you. I wonder how much backlash, if any, this printer has.

[@havenview]

Pulley diameter is intrinsically linked to teeth quantity and belt pitch so you could use either in this application but in any account he wasn't directly quantifying steps per mm he was calculating a relative change. In these applications the backlash on a belt is near zero. On a leadscrew you would have a point, which is why in the domestic 3D printing world leadscrews are often only used with a gravity assist which normally negates the issue.

[@BaldurNorddahl]

The precision of the part seems to be within 0.1 mm. I am not sure anything better can be expected with wood as material. What is the resolution of the 3D printer? The smallest distance it can move - the part can never be more precise than twice that value. This probably is part of the explanation why the parts where not identical. I am not sure how much it matters with wood being a soft material. But with metal that slot would definitely not have been the cutting width because you went both ways. That will make the slot wider.

[@thebeststooge]

X/Y has more precision than Z. X and Y would be 10micron for 16t gears with a 1.8° stepper at 1/16th microstepping using a 2mm belt. For 20 tooth gears it would be 12.5 micron.

[@foxabilo]

@@thebeststooge No screw drive for Z on this printer? that's an odd ball

[@foxabilo]

Typicaly I find 0.0125mm mm x and y and 0.00125mm for z, plus whatever backlash I have, I don't have tools fine enough to measure my backlash. But as you said not in wood with a dirty great motor vibrating about 5 tho is way more than adequate

[@thebeststooge]

Z has an 8mm per revolution leadscrew so way larger than X/Y but for 3d printers you do not need ultra fine layers. For me my printers normally use 1 or 2mm per revolution leadscrews.

[@foxabilo]

@@thebeststooge I getcha, the ones that look like barber shop poles ... i have to put in 800 stepper pulses to get 1 mm of travel on the ones I use.

[@Celcius1]

Remember your rounding rules!

[@ArtemKuchin]

I really thought that Tech2C is smarter than this. Didn't expect such crap. Terrible. Ruined my day. Just joking. But really, one should never ever, unless in total dispair, do such thing! The step/mm are based on mechanical calculations only (not measurements!). Because if you think about it then you realize that if steppers are ok, drivers are ok and mechanical parts are ok than there is no place where error beyond some very small error (sub 0.01mm) can appear. Measurements you do with caliper are always less accurate than accuracy of normal mechanics. So, you cannot measure accurate enough to make you mechanics even more accurate than you measurement tool and method. Basically, you measure with +-0,02mm of calipers + +-hell knows how much of curved hands and bad measurement method. If you have problems then you should not change the steps - check you mechanics!!! Could be stretched belt, bad Chinese pulleys and idlers (even a little wrong diameter), somethig's not square, etc. Solve the mechanical problems. Now about CNCing :) You set cutting width based on the cutting bit measurement. That's good, but you did not measure how much it vibrates sidewise during cutting. Your setup is not really CNC friendly. Not much firmness. So, no wonder if the real cutting width is a lot more than the cutting bit width. Make a cut, SCAN IT (do not use calipers) and measure the real cutting width. I think it would solve all the problems. Return step/mm back how they were.

[@Tech2C]

Every printer I've owned has needed slight calibration from the theoretical steps/mm because as you've mentioned, manufacturer inaccuracies are almost guaranteed at this price point. And yes I agree the printer may be vibrating or similar causing the run-out, but you saw when using a .1mm smaller end mill the error observed stayed the same 0.23mm, so I think the caliper was adequate for this test.Great ideas you have keep them coming ;)

[@havenview]

Artem Kuchin - Consider giving yourself a reality check, its a home made 3D printer making plastic parts primarily for fun and personal interest. It is not NASA! With all the will in the world there is no way you can ever expect plastic to extrude within those tolerances anyway. Doing what he's done is a simple, viable and effective solution to the problem with zero money spent. And about CNCing, the clue is its NOT a CNC, its a 3D printer that's been cleverly adapted. The runout on the collet shaft was the major culprit. The setup was friendly enough to produce a working and application accurate pen stand and my guess is that his steps per mm will stay modified

[@ArtemKuchin]

@@havenview Exactly because there are already tons of problems with such devices one should not create even more mess with correcting already existing error by creating new. One error does correct the other, but this is plain wrong. Eventually the error will show itself but will be harder to diagnose. Reality check passed, i guess. As for CNC see the suggestions i provided. Actually i am sure the size error came for the described reason. Just fix that and everything will be fine.

[@PokeyDoggo]

Artem does have a point, as you can see from his channel he has a bunch of idler issues, and the Core XY HEVO group also had these problems with cheap Chinese pullies. (tbh where else would we get them from, real ones are atleast 8 euro's a piece, and we need 8 of them). It is still a real thing, and a Core XY is not rigid enough for this, eventhough it's much more rigid than any prusa i3 clone printer is, and it can be compensated to produce a pretty piece. Then again i highly doubt all stepper motors, and their drivers, and the general wackyness of 8bit controllers, voltage fluctations, are not at fault here. I think these all can have a fault way above 0.01mm. And if we tried to fix everything, we'd be sitting here for ever, with very expensive measuring equipment that is way above out budgets... to mill a pen holder ffs. I'd recommend you all to look at the Mostly 3D printed CNC, with the modifications from Brauns CNC (always funny to see youtube channel calling themselves CNC or ''teaching tech'' r whatever but they are actually noobs and know jackshit, and have to learn along the way to even get a mediocre part from it). It is can do the job pretty well, and can be build for under 300 euros, depending on your end mill choice of course, which start at 50 to 150 euros for a decent one. A Core XY Hypercube costs atleast 300 euros as well. then again not everyone wants to throw another 300 bucks around for a pen holder or acrylic cutter.

[@ArtemKuchin]

@@PokeyDoggo Cheap idler generally are bad, usually are off center. Look at the VORON 3d printer - he author dropped all idlers and replaced them with pulleys on a shaft with 5mm ID bearings.Pulley also have problems but a lot less. As for error. TECH2C has 80 steps/mm, that's 5 full steps per mm, 0.2mm per step. 1.8d stepper has 5% error on full step. That's the worst case. 5% of 0.2mm is 0.01mm error from stepper motor. Idlers and pulley which are off center can add about the same 0.01 if set properly and up to 0.1 if set badly. The belt itself can add some (up to 0.1mm of longer distances) too if it is too tight, but properly tightened good belt can be very precise, the error will be less than 0.01. Bad frame (bad alignment and not right angles) can introduce HUGE error. I never heard that driver introduce positioning error when set proper current. The firmware is ok, because it just send pulses and it is pretty gooo at calculating the number of pulses. :) So, as you see the main two factor are idlers/pulleys and frame and maybe belts (must not be too loose or too tight, just about 3-4kg force). Also, motor can have bend shafts from factory or because of overtightened belts.