Unsolved Mystery. Weird Z0 behavior around build perimeter.

[lordbinky]

My best bed probe so far is the switch from a Sanwa 30mm joystick button. I just glued it into a hotend mount but it turns out to be much more precise than the other two microswitches I was using. I should make a probe out of it instead of use the raw switch sometime.

[lordbinky]

If you think about what the printer is doing, every movement is relative to the steps/mm. Essentially the belt is what makes 80 steps == 1mm. So rather than change that, you change everything else.

It's still possible the belt is wrong. If you have a precision ruler/calipers you can count the teeth of the belt over some long distance.

I'm starting to think the belts are indeed my problem (dipping between towers). Assuming the belts are accurate when made there still is the stretch to account for, as we use the belts under tension.

I've set my steps/mm to 81.0 and proportionally reduced the arm length and delta radius. This gives me the same flatness but better dimensional accuracy. And the programmed arm length matches what I measured off the machine.

If your tower is angled toward the center, it will cause dipping on the opposite side of it as well. Also the tower position, their individual distance from center and their orientation (their angular position) are more significant on such differences.

If you want to bring some peace of mind towards your belts, do repeat-ability checks. Probe a few different points over and over and see how accurate and precise the output is. The easiest is just to G28 and G30 to check the center position a few times. You'll find the consistency of the endstops, the motors, and the tightness of the carriages affect the precision of the result here the most, the belts are downplayed (unless you are straight up skipping teeth) in this from the home position since it does not require any the the carriages to move back up when they travel. One thing I did experiment on is that on a 12V power supply I do not have issue with over-skipping steps supplying my motors the max amps I could. I'm using °0.9 steppers with heatsinks and additional fans cooling them though (and regardless of what they say I have copper tower heatsinks all over my X5), so this may or may not be helpful, but what you may be accounting for is missed steps by increasing the steps/mm. If it's reliably missing steps, then I see no reason why it wouldn't work. There's still likely a few imps to banish in your build, but when you just want it working, do what it takes right?

[rpress]

Thanks for your thoughts. I've measured the tower angle with a straightedge and they are all pretty much straight up. I measured 0.01" over 16", pretty much nothing (atan(0.01/16) = 0.035deg). In terms of the 120 angle between them, I don't know what that is but the dipping is consistent between all 3 towers implying that they are equal angles.

My repeatability with FSR sensors is 0.02-0.04mm. Not amazing but not bad, I imagine most of that error is due to the FSR sensors. I'm not sure what my microstep accuracy is, but I'm sure it's not missing whole steps.

I'd like to find the source of the real error. It's difficult when the measurement standard for the machine (the belt) is an unknown entity.

Does anyone know who makes the belt SeeMeCNC supplies? I would like to find the datasheet to determine accuracy, and to see if there are any considerations for belt tension.

[KAS]

I ended purchasing a digital angle gauge as i wasn't having much luck with my squares. Come to find out that I was actually damn close to being square to the bed, although the the left and right tilt of each tower was off by a lot.

I'm not sure how you would correct that without a digital angle finder or come up with some ingenious way to mount a bubble finder on the end of a square.

[626Pilot]

626Pilot, I published a smoothieboard mount on thingiverse last year that you could use to mount on the rostock. It was originally designed to bolt to the tower on a mini kossel. This might actually be cool in the base. You could use one of the mount points as a rotating hinge point to be able to access the board with the door open and then rotate it back inside.

I did it up. The board faces out, so you can watch the blinkenlights and see whether it has crashed.

[img]http://thingiverse-production-new.s3.am ... atured.jpg[/img]

[lordbinky]

That is awesome. I have my Azteeg floating in mid-air with the cables holding it up so the heatsinks (top & bottom) are being cooled by a fan blowing across it It's a little much but it removes any worry from my brain. Looks horrible though.

[mhackney]

Hey cool! By the way, if you use Rhino and want source for anything like this, let me know.

cheers,
Michael

[Eaglezsoar]

That is awesome. I have my Azteeg floating in mid-air with the cables holding it up so the heatsinks (top & bottom) are being cooled by a fan blowing across it It's a little much but it removes any worry from my brain. Looks horrible though.

Levitation from the Lord?

[lordbinky]

That is awesome. I have my Azteeg floating in mid-air with the cables holding it up so the heatsinks (top & bottom) are being cooled by a fan blowing across it It's a little much but it removes any worry from my brain. Looks horrible though.

Levitation from the Lord?

I shoulda been a magician.....

[DavidF]

Can you print from a sd card with the smoothie board?

[lordbinky]

Yes, the SD card holds the configuration and firmware files as well. There's not an EEPROM to access though, or at least it doesn't work with the hosts I've tried so far.

[DavidF]

Yes, the SD card holds the configuration and firmware files as well. There's not an EEPROM to access though, or at least it doesn't work with the hosts I've tried so far.

Bit confused....so you can have your G-code on the same card? And view files thru a lcd as well? Guess i should do some surfing a figure out how to. Havent seen it done yet..

[626Pilot]

The SD card has a file called config, which is the stock config and has lots of named variables like zprobe.enable, zprobe.probe-pin, etc. It also has a file called config-override, which is written when you type M500 or dumped to console when you type M503. That contains only commented G- and M-codes, like this:

; Do something
G123 A1 B2 C3
M456 D4 E5 F6

You can also put G-code files on the SD card wherever you want. It's better to take out the card and mount it on a computer rather than doing it over USB. You can do it over USB, but it's slow and if a print is in progress... say goodbye to it, because messing with the SD card during a live print gives the firmware conniptions.

I am working on a Smoothie branch with heuristic calibration and I found that after adding some G and M codes to config-override, that file is corrupted whenever I type M500. (I have a separate file of my own which stores depths - it NEVER gets corrupted.) I think there is some file size limitation going on, which they claim to have fixed, but really isn't. So, when I change something, I have to type M503 to dump it to the console, copy, remove the timestamps, and then paste into config-override. It is a huge hassle and one of the reasons I don't say my fork is ready for prime time yet.

[626Pilot]

I made my own thread for my Smoothie fork with heuristic calibration. I just committed a new release that updates a lot of things behind the scenes, as well as fixing the M500 file corruption bug.

Thread here: http://forum.seemecnc.com/viewtopic.php?f=82&t=7640" onclick="window.open(this.href);return false;

[longshot]

Still a little new here . . . but I've been lurking on this post for a bit hoping someone finds out something.

I've got an original Rostock Max and I'm also noticing the "potato chip" feature where my hot ends take a dive in some regions and hits hard on the build plate in some areas despite following the instructions regarding calibration.

I am going to check the towers tonight with a digital angle gauge. it is an area I haven't looked into. I've used gauge vials to eyeball center, but with how tall the towers are I can see how if one or more of the towers are tipped in or out of center can cause problems.

However the last two pages are talking about replacing the existing Rostock board with a smoothie board, installing a z probe, etc. Is this really the only way to fix this? I don't mind going the manual approach (i've been messing with mine for months now, so i don't mind the extra time) but I'm way too afraid to by more hardware for this machine only to end up in the same spot.

Are there any thoughts or advice where we don't have to buy another control board for this to work?

[Nylocke]

You can do similar work with a dial indicator and adjusting the rotation values from what I can tell. I doubt he's done much work in this front considering his software ability, but if you ask him I'm sure he'd have some advice on how to do it manually. Alternatively you could check out Mhackeny's Z compensation program. Its essentially a g-code post processor to correct for the dips by adjusting the z values.

[JFettig]

so I got some TL arms so I decided to design and print new mounts for the carriages and a platform, I measured and looked and decided to beef it up by spreading the arms way out, they are 77mm center to center vs 50mm. I printed up an indicator mount and I'm actually reading +.001" opposite of the towers now!

I do have the rotational offsets configured as described earlier, they are just a little off of stock, I haven't messed with those yet as I was eager to print some stuff.

[img]http://i.imgur.com/9Jh0lcb.jpg[/img]

[Nylocke]

Thats sweet. Would you mind sharing those files? If I can get my printer to do that without a smoothie or Michael's correction code that would be sweet!

[JFettig]

Let me know if they do the same for you.

All 3 carriage mounts are the same, just positioned around the bed together - I arranged them in SolidWorks then saved the STL. I printed them out arranged such that the Z axis mount corresponds to the z-axis side of the platform and keep them together for the best accuracy.

http://repables.com/r/483" onclick="window.open(this.href);return false;

[Nylocke]

Awesome, thanks dude. I'll post my results here and in my build log. If these work you'll be the man! When you say the rotation offsets are different, do you mean you used the quadrant system that Pilot came up with a few pages back?

[JFettig]

http://forum.seemecnc.com/viewtopic.php ... 375#p58696" onclick="window.open(this.href);return false;

[longshot]

Ok so I printed the calibration thing from thingiverse. After some adjustments to the tower angles (within 0.5 degrees), checking belts, skates, etc.

This is what I have.

The areas between Z and X is high (not much adhesion) and the area between Z and Y is Low (squished plastic) . Areas at the towers seem to be ok. Any ideas on what my next step should be?

Still stumped. . .

[redoverred]

Not having a z probe, I determined using a sheet of paper that the points directly across from the towers are significantly higher as far as the nozzle distance goes, i.e.: the nozzle doesn't touch a sheet of paper at these points. Using the z height menu, I can see they are all over .2mm away from the bed. My z height at each tower and center are perfect. I am using points 120mm from the center for each of these points. I figure rotating tower angles is pointless with this error being on each tower-tower arc, so what the heck else can I do to fix it? I can get decent prints near to the center of the bed, but a calibration circle will not stick properly on the outer diameter (240mm diameter circle).

[626Pilot]

It's not pointless. Tower angle and radius offset can both have an impact on that. The printer thinks its towers are somewhere they aren't, and that's what causes that problem. If it only knew where they really were, it would lay down a perfect 1st layer every time.

[redoverred]

It's not pointless. Tower angle and radius offset can both have an impact on that. The printer thinks its towers are somewhere they aren't, and that's what causes that problem. If it only knew where they really were, it would lay down a perfect 1st layer every time.

Even if they are all high opposite the towers? I guess I'll have to fool with it a bit, but even after reading this whole thread and others it just seems like people are randomly moving things around until it fits. Is there even any math involved or steps to take to fix this in some sort of logical manner? Either way I guess I'l mess with the rotations a bit.