Unsolved Mystery. Weird Z0 behavior around build perimeter.

[Moeparker]

So far no joy for me. I don't know, it's driving me crazy.

I have two Rostock max printers. The normal sized one is having this Z0 gap issue between the towers. The 6 foot tall printer prints great between XZ tower, but has that gap issue between XY and YZ. So I'm just pulling my hair out at this point. I don't know.

My Snowflake was slightly uneven, so I thought maybe I'd remove that. In fact I took all the parts off and just took the nozzle down to the wooden frame to check that out to eliminate the glass, heated bed, snowflake, shims, etc. It wasn't level to the wooden frame itself. I took off everything down to my frame and there are still gaps between the towers at Z0. Is my Delta-Radius wrong with the length of the trick laser arms? Has my machine wore down some part that I just need to replace to fix this? Beats me.

I'm stuck. I'm going to play Skyrim for a while, give my brain a chance to reset it's logic centers.

[bot]

It is all in the math. I'm no wizard, but it's just finding the right balance. As I said earlier, you will need to bring some variables seemingly OUT of spec, to later equalize it by bringing another value conversely into spec... I can't tell you which values they are... but don't give up! It can be done! (somehow! when I get my printer I may be of more help).

[RegB]

I admit to having lost interest in this issue.

For ME it is almost irrelevant, although I was distracted by it for a while.
I am not trying to print HUGE parts, 10 cm parts in ABS seem to be about the limit of what I can get to stick, stay stuck and not curl too badly as it cools.
I made a couple of tries at printing a lot of parts on the bed at one shot, the math gears.
That would have taken a lot more hours to complete than I wanted to leave the machine unattended - IF it had started a decent perimeter, but it didn't, most likely due to this issue.

In any case, I'm not leaving it running for long periods unattended.
I'll just print one, two or three little parts near the middle, or one ~10 cm part.
If/when I have a project needing "production volume" I'll probably sub it out (-:

[626Pilot]

In any case, I'm not leaving it running for long periods unattended.
I'll just print one, two or three little parts near the middle, or one ~10 cm part.
If/when I have a project needing "production volume" I'll probably sub it out (-:

It's terrible that your printer corners you into having to make such a decision, forces you to give up on it and consider paying someone else to have their printer do what yours should be able to and can't. When they publish specs and you use those specs to decide whether to buy their printer, and the specs turn out to be literally wrong, is that your problem to fix, or is it theirs? I say it's theirs.

During the year since this problem was first reported on the forums, I've spent hundreds of hours reading headache-inducing mathematical papers, puzzling over countless depth readings and angles, developing my own Z-probe and figuring out how to wire it up, working with other users to try and figure this out, writing code for Smoothieware, etc.

During the same time period, SeeMeCNC has:

• Started work on a Z probe, and then given up
• Looked at one affected printer
• Issued a minor change to the carriage offset which doesn't actually fix the problem
• Sold a printer that cannot be relied on to achieve the published specs out in the field
• Failed to stop selling the printer when they realized the problem
• Failed to warn prospective customers that the published specs are more a matter of chance and hopefulness than something to be relied upon, concerning a thousand-dollar decision
• Let their users put in huge amounts of legwork to solve a problem that those users didn't cause
• Failed to issue a serious public apology for any of this, or made a concerted effort to show whether or what they're doing

From my perspective, they're standing still. Have I really done more work than they have, and I'm just one guy? That isn't right. They need to wake up.

To put it bluntly, which is regrettable but apparently necessary, SeeMeCNC's strategies for handling this have been woefully ineffective for far too long. This is their ship. When it comes to this issue, they have it on autopilot. This is not an autopilot issue. Someone needs to take the helm. They need to sit down and do the thinking they've been avoiding so far. I'm sure they can do it. They have at least one engineer, right? That's what engineers do, isn't it? Mistakes are okay, but failing to learn from them is a waste, and making your customers deal with that failure is unkind.

[bot]

Technically, this issue transcends SeeMeCNC. The rostock design is not theirs. You should have done a little bit more research before you purchased your kit because this "issue" is the name of the game with Delta style home built 3d printers... and the "issue" can be solved. It doesn't involve any crazy inverse kinematics or textbook theory. It involves tweaking.

[bot]

A thought I just had: perhaps the solution lies in calibration of the individual towers' steps per mm? I remember verifying that for myself when I had a homebuilt and it was around the time I solved my wavy plane issues.

[626Pilot]

Technically, this issue transcends SeeMeCNC. The rostock design is not theirs.

The implementation is theirs. The slop in the implementation is theirs. The spec sheet where it says the print diameter is 11", when it may turn out to be half that, is theirs. It's hard for me to understand how you could think them blameless when all this stuff is out in the open.

You should have done a little bit more research

How many hours of research did I do, and how many should I have done?

...and the "issue" can be solved. It doesn't involve any crazy inverse kinematics or textbook theory. It involves tweaking.

How many Rostock MAXes have you tested this theory on?

[bot]

Sorry to ruffle your feathers. Have you considered that this is what an improperly calibrated delta printer behaves like? Even though, in theory, each tower is "identical" they, in practice, are not. There has to be a balance achieved in the firmware variables and the physical adjustments that are possible.

Can you point me to the "slop" that you believe contributes to this problem? I have heard that some people experience better results with carbon fiber arms. Are your delta arms flexing? Sliding around on their joints?

Unless you can point me to a deficiency, specifically, in the design or the implementation... I'm going to have to believe this is calibration related. I've had these EXACT symptoms on a home built rostock design, which incorporated NO similar parts to the MAX, and was able to flatten the bed and print huge pieces with good layer adhesion... all after pulling my hair out trying to figure out why the plane was wavy.

This is not unique to the SeeMeCNC kit...

[626Pilot]

Have you considered that this is what an improperly calibrated delta printer behaves like?

That is one possibility. Another possibility is that this is a design flaw that cannot be addressed in the assembly/calibration instructions.

Can you point me to the "slop" that you believe contributes to this problem?

A lot of the design elements rely on sanding, eyeballing, and loose tolerances when it would have been much simpler to use tight tolerances and restrict any unnecessary range of motion. You have to manage multiple degrees of freedom on multiple parts, where only zero or one degrees of freedom are needed. Let's start with the cutouts for the towers. On other Rostock designs, the towers are usually bolted into plastic or laser cut wood pieces that they fit tightly into, and thus, don't permit any wiggle. The tower cutouts on the MAX are wide enough that a tower may be 119 degrees from one neighbor and 122 from the other. I know this because I wrote code that tries adjusting the tower angles based on Z probe feedback, and indeed, the Z=0 error improves (but doesn't entirely vanish) when I let that code run about an hour and a half worth of test iterations. Of course, that side to side slop is not the only problem. When you assemble the machine, there is nothing in the base for the towers to sit on. You have to eyeball a little laser-cut line while you tighten them down. Also, the towers are supposed to press directly up against the inside edges of the cutouts, but even if you press firmly, they sometimes don't. So, you get the possibility of each tower being a slightly different distance from center. Likewise, the top plate has nothing that sits on top of the towers, so you have to eyeball them too, or cut a yardstick to length and hopefully get that lined up perfectly. When you tighten the towers down, the torque from the T-slot nuts may cause the towers to move up, down, right or left, closer to or further away from center, and it may also cause the tower to rotate to a less vertical alignment. Using an angle gauge, I discovered that it was common to get a tower that was 2+ degrees off from being vertical. With some care it was possible to get the deflection under 1 degree. I also developed tower clamps that would pull the towers in toward the center of the printer at the top and bottom. They helped a little, but they didn't fix the problem.

It would be better if the top and bottom assemblies fully captured the towers and restricted their range of motion. The towers need to be able to move up and down so you can seat them, but what's the point of all those other degrees of freedom? I would argue that there is none, and that they are counterproductive.

I have heard that some people experience better results with carbon fiber arms. Are your delta arms flexing? Sliding around on their joints?

I have had both the old and new versions of SeeMeCNC's delta arms, and the Trick Laser carbon fiber arms. The TL ones are better hands down. I honestly think SeeMeCNC should be shipping those instead of what they do now. Even though the new SeeMe arms are supposed to be easier to deal with, you STILL have to carefully deburr and sand stuff, which means you have an opportunity to mess it up by mistake. On the new design I was getting layer shifting because I had sanded somewhere a little too much, allowing the effector to wiggle slightly. With the TL arms there is no sanding of any kind. You do have to use an un-sanded effector with them. A sanded one will not fit tightly. Same with the U-joints on the carriages.

[bot]

I really sympathize with your frustrations. I've been there. I am somewhat of a perfectionist, so things not being done to their most precise bothers me. I will have to wait and see how I feel after I assemble my kit. Keep in mind that it sounds like you have the V1 kit (did you upgrade the laser cut components?) I will be evaluating a new V2 kit... so we may have different experiences.

Also... if you are truly unhappy with your frame, it is not hard at all to build your own. Maybe you should fix the design how you see fit?

[626Pilot]

I have an acrylic V1 kit. I was first in line to buy the V2 upgrade when it was made available, but after another user told me the kit didn't fix this problem I realized it would be a waste of money. I was pretty sad about that. I have been thinking about designing a replacement for the top and bottom pieces that hold the towers in place, something that would exactly fit everything else and fully capture the towers with a very narrow tolerance. I think it also might be wise to completely replace the top plate with something much stiffer, like a T-slot triangle.

[bot]

Can you quickly summarize what you have tried and what has helped/hasn't? This thread got confusing for me. I'm about to build a v2 kit and would like any advice you may impart.

[Jimustanguitar]

Swapped Rambo boards, and that didn't do the trick. Also tried Marlin instead of Repetier (loaded from someone else's computer too)... I'm running out of conspiracy theories.

Things that have been replaced:

All wooden parts (V1 to V2 conversion)
Build plate (window glass, boro glass, and machined aluminum)
Arms (V1, V2, and Trick Laser)
Carriages (V1, V2)
Effector Platform (V1, V2)
Motors (swapped from Wantai to Kysan)
Pulleys (stock plastic to all aluminum)
Belts (different brands)
Rambo board (both 1.1a)
Wiring (not troubleshooting, just accessorizing)
Endstops (different brands)
Firmware (multiple versions of Repetier and 1 of Marlin)


Things that have not been replaced:

skate bearings, covers, and cams
belt idlers and spacers
aluminum extrusions


Unless my extrusions are bent, I can't imagine any of the parts that I haven't replaced yet having anything to do with this... I have extra aluminum to try this with, so that's next on the list. I've also got a smoothieboard to try in the next couple weeks, and a sanguinolulu if needed.

Am I crazy or is there something obvious that I'm overlooking? My only remaining theory is that there might be some gremlin hiding inside my Rambo board version since I tried 2 of the same version. Is anyone else who's having this issue using something other than a 1.1a? Maybe there's a math error and switching to a different microstepping setting or swapping the pulleys for a different tooth count is next?

Any suggestions besides seeking professional mental help?

That's how far I got on mine before the shimming trick solved my troubles.

[RegB]

I guess I view SeeMeCNC as a vendor to an open source community, they make and sell "parts".
OK, they go a LOT farther than that.
The kit is a GREAT way to save time and effort by not having to hunt down every little bit and it DOES pretty much all fit together.

There are opportunities for build errors and for tolerance build up.
I accepted that the kit isn't designed for a worst case scenario of their manufacturing in combination with every potential purchasers' assembly skills/tools/methods.

I no more intend to use the whole surface of the plate any more than I intend to ride my motorcycle at its (claimed) maximum speed (~300 Km/hr).
In the former case I won't leave the printer unattended for several hours, in the latter case I get helmet lift (due to aerodynamics) at about 220 Km/Hr - - IOW there are other PRACTICAL limiting factors.
Not least of which, anything much over 10 cm doesn't stay stuck and curls as it cools.
To ME this is NOT a production tool, it is a prototyping platform for hobby/pastime use that I might make a few service/repair things with.

Yes, it would be nice if SeeMeCNC could help us to get our machines working well right out to the corners of the build envelope, but as a practical matter I don't NEED it and I doubt that MANY others do.

[Moeparker]

Reg, makes sense. I feel ya. However I do need the entire area. I'm mass producing 3D printed items for sale. I leave my printer on for long stretches, 19 hours was the max I think. Nervous night's sleep but it paid off.

Also, I think given what 626Pilot and Bot have said I will be able to fix this issue.

I used to print on the entire print circle. Full thing, life was good. I printed myself a lovely full sized crown. Lately however, no full bed area printing. I've tinkered with it a lot, mostly those leveling screws, and I think I've done what they've both hinted at. I've lowered the nozzle so much at the tower that it's had that tetter totter effect and has raised it on the other end. It makes sense cause I added a new shorter hot end and I lowered the nozzle at each tower, when I should've just reduced my max Z height and kept it level.

I agree on the large parts curl for ABS, but I'm just printing with PLA so I can print those huge items. Last year I printed a 10 inch slide for a Halo pistol and it wouldn't work with ABS. I tried 8 times, curled so bad it tore the tape off the glass each time. ABS can CURL!

But yeah, it's not a point of wanting that whole print bed for the sake of advertising truth, I need it. And from the last few pages of this thread I think I'm on the track to full bed printing. I'll report back Friday. I have to go to Winterhold now, drop off some dragon bones.

[626Pilot]

Yes, it would be nice if SeeMeCNC could help us to get our machines working well right out to the corners of the build envelope, but as a practical matter I don't NEED it and I doubt that MANY others do.

The build envelope was one of my primary reasons for buying this instead of something else. If someone came to me and said, "I'll pay you to print this beautiful huge sculpture," I wanted to be able to say, "Okay." There are much smaller things, like hand tools, that may only require 30% of the print area - but that 30% needs to be wide. I did some tools for a guy that were pretty simple, and roughly wrench-shaped. It took a lot of work just to be able to put down the first layer properly, and this irritating glitch meant I had to print fewer of them at a time than the printer ought to be capable of. Of course this means more time spent setting up the jobs, monitoring the first layer to ensure it prints, etc. It's wasteful. This is not the kind of unpleasant surprise I want to deal with when someone is paying me for my time, and I have a deadline to think about.

[626Pilot]

That's how far I got on mine before the shimming trick solved my troubles.

Did you try watching the belts to make sure they aren't sliding left and right on the stepper pulleys? I notice mine do that sometimes even though I have them fastened straight up and down to the carriages. They tend to ride up against the clamshell halves of the idlers by the steppers, now riding up against the right and later riding up against the left. That seems like it might cause a positioning error. But since the clamshell halves are further apart than the width of the belts, the belts are not constrained and are free to wander back and forth like this. I have been thinking this might be resolved with a simple printed belt guide, with the belt-touching parts lined with Kapton tape to make them ultra-smooth.

[Jimustanguitar]

That's how far I got on mine before the shimming trick solved my troubles.

Did you try watching the belts to make sure they aren't sliding left and right on the stepper pulleys? I notice mine do that sometimes even though I have them fastened straight up and down to the carriages. They tend to ride up against the clamshell halves of the idlers by the steppers, now riding up against the right and later riding up against the left. That seems like it might cause a positioning error. But since the clamshell halves are further apart than the width of the belts, the belts are not constrained and are free to wander back and forth like this. I have been thinking this might be resolved with a simple printed belt guide, with the belt-touching parts lined with Kapton tape to make them ultra-smooth.

When I swapped out my pulleys for all aluminum ones as a test, the new ones had a much narrower toothed section and high walls around it, so there was very little play side to side. Much less than the dimensions of the stock pulleys would have allowed. If you're worried about your belts walking on the pulleys and idlers, get some pulleys from robotdigg and print 3 guides for the top idlers.

[RegB]

From a "clean sheet of paper" it would seem to make some sense to do all this by actually MEASURING the HEIGHT ABOVE the plate at or close to where the hot end carrier currently IS -- not by calculating it as a DEPTH BELOW three switches.
Follow the actual plate.

WW II - dam busters, bouncing bombs, height range finders, etc....
(-:

[Steve123]

Yes, it would be nice if SeeMeCNC could help us to get our machines working well right out to the corners of the build envelope, but as a practical matter I don't NEED it and I doubt that MANY others do.

The build envelope was one of my primary reasons for buying this instead of something else. If someone came to me and said, "I'll pay you to print this beautiful huge sculpture," I wanted to be able to say, "Okay." There are much smaller things, like hand tools, that may only require 30% of the print area - but that 30% needs to be wide. I did some tools for a guy that were pretty simple, and roughly wrench-shaped. It took a lot of work just to be able to put down the first layer properly, and this irritating glitch meant I had to print fewer of them at a time than the printer ought to be capable of. Of course this means more time spent setting up the jobs, monitoring the first layer to ensure it prints, etc. It's wasteful. This is not the kind of unpleasant surprise I want to deal with when someone is paying me for my time, and I have a deadline to think about.

This^^^
Same reason i chose the Rostock over other printers! If i buy a bike that states it goes 300km an hour, it better go 300km an hour! So is this being worked on? Or am i going to have to shim my bed?

[Polygonhell]

This^^^
Same reason i chose the Rostock over other printers! If i buy a bike that states it goes 300km an hour, it better go 300km an hour! So is this being worked on? Or am i going to have to shim my bed?

It depends what you believe the problem to be.
If you believe it's accumulated error because of combined error from manufacturing and assembly, I'm not sure what they can do about it.
If you believe it's software configuration, then it should be fixable, but every machine should have the same issue.

My V1 doesn't do this
I did do the math before I built it to understand the geometry and what were critical dimensions and I was particularly careful during assembly in those places.
I also run different firmware than everyone else.

As I posted a few pages back, the problem is that while you can get exactly this result from a combination of errors in delta radius and arm length, pretty much any significant mechanical error manifests int a similar way.

[RegB]

This^^^
Same reason i chose the Rostock over other printers! If i buy a bike that states it goes 300km an hour, it better go 300km an hour! So is this being worked on? Or am i going to have to shim my bed?

It depends what you believe the problem to be.
If you believe it's accumulated error because of combined error from manufacturing and assembly, I'm not sure what they can do about it.
If you believe it's software configuration, then it should be fixable, but every machine should have the same issue.

My V1 doesn't do this
I did do the math before I built it to understand the geometry and what were critical dimensions and I was particularly careful during assembly in those places.
I also run different firmware than everyone else.

As I posted a few pages back, the problem is that while you can get exactly this result from a combination of errors in delta radius and arm length, pretty much any significant mechanical error manifests int a similar way.

Care to elucidate ?
i.e. WHICH are the critical dimensions ?
Is it worth stripping a machine down and re-assembling it with particular care to those ?

[Polygonhell]

Care to elucidate ?
i.e. WHICH are the critical dimensions ?
Is it worth stripping a machine down and re-assembling it with particular care to those ?

Outside of ensuring the towers are all vertical in both dimensions relative to the build plate which is critical because of the lengths of the columns, I honestly don't remember, there was no complete manual when I built mine and I did the math as an exercise to understand the mechanics while I was doing the build. Pretty much everything is a function of the combination of Sin's or Cos's of the angles of the arms.

There is a practical limit to how accurately you can assemble something and small errors in arm position << 1mm are not critical.

The one assumption I made when I was building was that the laser cut parts were largely accurate (though it's a questionable assumption), so I would try and ensure all of the gaps were at the very least consistent. I know I did not follow the advice (which I vaguely remember from the time) of ensuring the towers being fully seated in the top recesses because it did mean on my build the towers would not have been vertical.

[626Pilot]

From a "clean sheet of paper" it would seem to make some sense to do all this by actually MEASURING the HEIGHT ABOVE the plate at or close to where the hot end carrier currently IS -- not by calculating it as a DEPTH BELOW three switches.
Follow the actual plate.

Endstops can be mounted at the bottom, but the carriages can't get near them if they're there, so you have to have the printer reach down and touch each one with the effector. If the glass plate was perfectly level to the support, and if the endstops were all mounted at the exact same height over the support, it could work.

The towers can still have issues, though. Every variance corrupts the coordinate system to some degree. They can lean in or out, or side to side. They can be one or two degrees off from being 120 degrees apart. They can be at different heights, which means the top support plate is not orthogonal to the bottom, which means the whole coordinate system is compromised to that extent. I think this could also introduce some torsion on the towers, though probably not much. Getting the geometry as close to perfect is very important. It solves as many of those issues as possible in the best way possible.

The rest has to be made up in software. The favored solution from what I've seen is to use a Z-probe with Marlin. I tried this last year and got bad results. I got similarly bad results using a Z-probe with Repetier. It looked so cool while it was mapping the bed, but the end result was a first layer that was slanted. One half of the print surface was too high, the other too low. I then spent a few weeks implementing my tower rotation methods into Smoothieware. That code runs dozens of tests to see whether it helps to rotate the towers a degree or two in either direction, and it tries combinatorics with rotating two towers at the same time as well. It takes about an hour and a half and produces a moderate improvement, but I still can't hit the edges even. I would like to put the code on GitHub but GitHub is such a pain in the ass dealing with anything other than the main branch. I will figure it out at some point I guess.

[bot]

Okay, I'm going to suggest what I remember my steps for resolving this were:

1) set the 0,0 point at your desired height (paper thickness).

2) Set all three tower points at the same height. This should now mean the plane is level? Not necessarily.

3) Move the hot end from A tower position and 0,0, watching very closely to the path of the nozzle. Does it travel STRAIGHT across the bed? Or does it lift up and then back down from the tower to the center point? Adjust your firmware settings until this is FLAT. My making this flat, your plane will likely be sloped in an odd direction. That's okay.

4) Re-set every tower height as in step 2.

5) Now you will move the hot end from the tower points to "opposite" tower points. As in the point of the bed directly across from the tower, in between the other two. You now have to set these opposite heights by adjusting the physical end stop screws. Continually adjust the screws, set all the height points, move the hot end across the bed.

6) Repeat steps 1-5 until you are happy with results. By repeatedly adjusting these variables, you will slowly bring your machine closer and closer into spec. When adjusting this, you'll maybe want to bring the hot end farther away from the bed and use the firmware movements of 1mm and .1mm to "measure."

If any of this is unclear let me know... it's possible you've tried repeating this procedure, but it was what lead to my solution. When I get my max v2 kit I'll take pictures and detail my calibration process (if it even works).