Diagonal rod length on delta printer issues.
-
Second update.
7 factor error was 0.0508
8 factor error 0.3048mm after multiple iterations- Calibrated 8 factors using 19 points, deviation before 8.525 after 0.146
Deviation before is obviously due to having something wrong with my probe setup and it changing my H number. Deviation after is overly optimistic like the least-squares calibration calculator online calculator I was using before.
I am borrowing another square to see if my error is there. Still can not comprehend using an incorrect number (and math being done correctly) and getting such good results.
Cory
- Calibrated 8 factors using 19 points, deviation before 8.525 after 0.146
-
You can adjust the baby step rate in DWC
(this is on DWC 2.0.0-RC6) -
@dougal1957 said in Diagonal rod length on delta printer issues.:
You can adjust the baby step rate in DWC
(this is on DWC 2.0.0-RC6)That doesn't seem to effect the z steps in machine control or in the probing cycles.
-
So my square is within .1mm as confirmed with and indicator on a CNC milling machine.
I'm back to what error can be covered up so perfectly by using a wrong number when doing the math.
Below is a quote by dc42 in the other thread I linked too. Can anyone elaborate?
- Having the arms go vertical or even beyond when printing close to a tower isn't a problem. What you need to avoid is arms going below about 20 degrees to the horizontal
Thanks
Cory -
i got similarly whacked results when my frame wasn't as straight as humanly possible (i.e. calibration aids suggested absurd rod lengths). After rebuilding the whole thing I mounted a dial indicator instead of nozzle and manually adjusted the delta radius until it was flat.
if i recall correctly if the center of the bed is low you increase the delta radius (or smooth rod offset or whatever its called in the various firmware). I think it will vary with bedsize but a factor of 4 gets you close quite fast. I.e. your center is 1mm low, add 4 to the radius; split the difference when you overshoot. (or if i'm mistaken, its the other way around, but you'll see)
after spending days with auto calibration, re-doing the whole thing by hand took an evening, rebuild and all. The wiki highlights the important aspects of the physical build, get that right and the calibration is pretty straight forward by hand.
https://duet3d.dozuki.com/Wiki/Calibrating_a_delta_printer -
Another source of error for me has been not having the belts parallel to the frame, caused by the toothed gears having slipped on the motor spindles. Measure belt-to-frame at each carriage and at the bottom just above the motor spindle, get them the same.
-
@bearer said in Diagonal rod length on delta printer issues.:
i got similarly whacked results when my frame wasn't as straight as humanly possible (i.e. calibration aids suggested absurd rod lengths). After rebuilding the whole thing I mounted a dial indicator instead of nozzle and manually adjusted the delta radius until it was flat.
if i recall correctly if the center of the bed is low you increase the delta radius (or smooth rod offset or whatever its called in the various firmware). I think it will vary with bedsize but a factor of 4 gets you close quite fast. I.e. your center is 1mm low, add 4 to the radius; split the difference when you overshoot. (or if i'm mistaken, its the other way around, but you'll see)
after spending days with auto calibration, re-doing the whole thing by hand took an evening, rebuild and all. The wiki highlights the important aspects of the physical build, get that right and the calibration is pretty straight forward by hand.
https://duet3d.dozuki.com/Wiki/Calibrating_a_delta_printerI loosened all the bolts and re squared the printer before the last tests. I also checked between the towers with the outside of calipers, really annoyed that I didn't think of that myself. They are within .25mm of being consistent. Didn't have any luck with the way the frame is made getting them any closer, I think one might have a bit of a warp in it.
Also the subject of changing the radius to get to the head to track center to outwards... This works great, until I get to the area between the towers. Or opposite a tower, however you wish to describe it.
@robm said in Diagonal rod length on delta printer issues.:
Another source of error for me has been not having the belts parallel to the frame, caused by the toothed gears having slipped on the motor spindles. Measure belt-to-frame at each carriage and at the bottom just above the motor spindle, get them the same.
Thanks, I just checked that, no luck there either.
I'm pretty stumped by this and have to get away from it for a while. Unless anyone else has a better suggestion, I'm planning to decrease the print area size and see if I get better results with the correct arm length.
Thanks
Cory -
Hi,
My first three printers were mini-deltas kits. I spent hours and hours tweaking them.
Then I got a Folger Tech FT-5 kit. It worked so much better with so little effort I got rid of the deltas.
Frederick
-
@fcwilt said in Diagonal rod length on delta printer issues.:
Hi,
My first three printers were mini-deltas kits. I spent hours and hours tweaking them.
Then I got a Folger Tech FT-5 kit. It worked so much better with so little effort I got rid of the deltas.
Frederick
Thanks, I guess.
So after many hours I can not repeat the extremely flat results I had before with the incorrect rod length I posted a video of. I don't know if my attempt to re-square the printer made things worse, or if I was able to compensate better by feel to set the tilt. I get slightly better results with the incorrect rod length than I do with the correct number.
I rigged up a probe and have ran many tests. Much easier to do, but harder to feel for what is happening than using an indicator.
I am still of the mindset that if you do the math right with the wrong numbers it is impossible to get the results in the video of my first post. Would be nice to have access to the actual step data, or actual axis position data. Something doesn't add up.
Cory
-
Correct rod length 291.06
294.9 rod length
Messing with the tilt some more.
Unlike in the video I have a bad high spot by the Y tower no matter what I do. But the regardless more of the bed would be usable.
Going to try and forget about this for a couple weeks while I'm on vacation
Cory -
Do the rods bend at all as the effector moves? If so then that will change the effective length of the rods, affecting the geometry.
I am suspicious of rods with springs between them, because the stretch of the springs, the force they apply to the rods and the direction of that force will depend on the XY position of the effector.
Duet WiFi hardware designer and firmware engineer
Please do not ask me for Duet support via PM or email, use the forum
http://www.escher3d.com, https://miscsolutions.wordpress.com -
@dc42 said in Diagonal rod length on delta printer issues.:
Do the rods bend at all as the effector moves? If so then that will change the effective length of the rods, affecting the geometry.
I am suspicious of rods with springs between them, because the stretch of the springs, the force they apply to the rods and the direction of that force will depend on the XY position of the effector.
I tried removing the springs early on and didn't see any real improvement. In the thread I linked to in the beginning I think they were having the same issue with magnetic rod links.
It seems that longer rods help. But then you trade build height, torque, and resolution
Thanks
Cory -
Going to reinstall the old control board to test when I get back from vacation. Seems that 3 motor and 3 home switch connections are about the quickest next test. Thanks
Cory -
So after more exhaustive testing. I did find that the wood frame under the onyx bed was warped up(convex) pretty bad. I flipped it over and shimmed it to the frame. It was warping the glass a little bit. I kept checking the glass for flatness, but not on the machine. Now unfortunatly the onyx bed isn't really supported and the glass is the structure there. All in all that setup sucks and I'm thinking though a better option. Problem is every time I start thinking I end up with a whole new machine with a metal frame and linear ways. And, and, and. lol
I removed the vibration dampers I had on the motors, I had pretty much forgotten they didn't belong. With the duet board it is way quieter without them than the rambo board was with them. That made the largest improvement. They add deflection, plus moved the pulleys way out on the shaft to give the angular deflections more deflection. If that makes sense.
Of course the rambo board didn't do any better, but I was stuck on the thinking the Duet must be making an calculation error since others had the same results.
Latest results, with the correct arm length. Big improvement over previous results.
I think the belts are tight enough but the screws have pulled into the wood frame, so I'm going to make some plates so I can tighten them a bit more. Might try the longer carbon fiber arms for the rostock too. Seems to help others.
I'm thinking this is far from ideal, is this close enough for bed mapping to handle? Most of the issue is way out on the corners. Should I map with a closer spacing than 20mm?
Thanks
Cory -
Went ahead a put .9 deg steppers in.
Some tests suggested by dc42[code]
Position Trigger heights
X0 Y0 6.224, 6.217, 6.224
X100 Y0 6.235, 6.235, 6.229
X140 Y0
X100 Y0 6.210, 6.210, 6.217
[/code]
Looks pretty good to me.So I'll try with some numbers over by the towers.
[code]
Position Trigger Heights
X0 Y0 6.249, 6.242, 6.242
X86.6 Y-50 6.263, 6.276, 6.263
X121.24 Y-70
X86.6 Y-50 6.226, 6.226, 6.220
[/code]If I go anywhere on the bed and feed Z down to zero the indicator, jog Z negative direction, then back up, I'm within .001" of being back at zero. The numbers in my first test at X100 and X140 matches pretty much exactly the ,001"(.0254mm) backlash I've been seeing. That is how you check machine tools for backlash, seems like it should work here. Actually seems to work, at least in normal realm... Not over near the towers when quantum mechanics seem to take over. But it's still not horrible.
So I wrote a little program to pick back and forth over in the trouble area. My bedmap was done at 13mm spacing. So that is the number I'm using. Points and trigger heights listed, moving one direction, then reversing the path.
[code]
G28
G1 X120. Y-70. Z11. F1500. ;point #1 Stopped at height 6.383 mm
G1 X120. Y-57. Z11. F1500. ;point #2 Stopped at height 6.295 mm
G1 X120. Y-44. Z11. F1500. ;point #3 Stopped at height 6.271 mm
G1 X107. Y-44. Z11. F1500. ;point #4 Stopped at height 6.232 mm
G1 X107. Y-57. Z11. F1500. ;point #5 Stopped at height 6.310 mm
G1 X107. Y-70. Z11. F1500. ;point #6 Stopped at height 6.315 mm
G1 X107. Y-57. Z11. F1500. ;point #5 Stopped at height 6.260 mm
G1 X107. Y-44. Z11. F1500. ;point #4 Stopped at height 6.244 mm
G1 X120. Y-44. Z11. F1500. ;point #3 Stopped at height 6.340 mm
G1 X120. Y-57. Z11. F1500. ;point #2 Stopped at height 6.364 mm
G1 X120. Y-70. Z11. F1500. ;point #1 Stopped at height 6.364 mm
[/code]
These seem to sometimes look like backlash, but other times be pretty good.Am I asking too much from the printer? Is this normal? I've burnt up so much time on this it's getting ridiculous.
Thanks
CoryMachine info:
.config
[code]
M665 L291.060:291.060:291.060 R144.257 H368.890 B140.0 X-0.009 Y0.368 Z0.000
M666 X0.088 Y0.191 Z-0.279 A-0.09 B-0.24; Drives
M569 P0 S0 ; Drive 0 goes forwards
M569 P1 S1 ; Drive 1 goes forwards
M569 P2 S0 ; Drive 2 goes forwards
M569 P3 S0 ; Drive 3 goes forwards
M350 X16 Y16 Z16 E16 I1 ; Configure microstepping with interpolation
M92 X160.0 Y160.0 Z160. E92.5 ;92.65 Set steps per mm
M566 X800.00 Y800.00 Z800.00 E1200.00 ; Set maximum instantaneous speed changes (mm/min)
M203 X18000.00 Y18000.00 Z18000.00 E1200.00 ; Set maximum speeds (mm/min)
M201 X2000.00 Y2000.00 Z2000.00 E2000.00 ; Set accelerations (mm/s^2)
M906 X1500.00 Y1500.00 Z1500.00 E800.00 I30 ; Set motor currents (mA) and motor idle factor in per cent
M84 S30 ; Set idle timeout; Axis Limits
M208 Z0 S1 ; Set minimum Z; Endstops
M574 X2 Y2 Z2 S1 ; Set active high endstops; Z-Probe
M558 P4 H3.3 F30 T1800 ; Disable Z probe but set dive height, probe speed and travel speed
M557 R140 S13 ; Define mesh grid
G31 Z6.256
; Heaters
M307 H0 B0 S1.00 ; Disable bang-bang mode for the bed heater and set PWM limit
;M305 P0 T100000 B4138 R4700 ; Set thermistor + ADC parameters for heater 0
M305 P0 T100000 B4425 R4700 L0
M143 H0 S120 ; Set temperature limit for heater 0 to 120C
;M305 P1 T100000 B4138 R4700 ; Set thermistor + ADC parameters for heater 1
M305 P1 T100000 B4725 C7.060000e-8
M143 H1 S280 ; Set temperature limit for heater 1 to 280C; Fans
M106 P0 I0 F500 H-1 ; Set fan 0 value, PWM signal inversion and frequency. Thermostatic control is turned off
M106 P1 S1 I0 F500 H1 T45 ; Set fan 1 value, PWM signal inversion and frequency. Thermostatic control is turned on; Tools
M563 P0 D0 H1 ; Define tool 0
G10 P0 X0 Y0 Z0 ; Set tool 0 axis offsets
G10 P0 R0 S0 ; Set initial tool 0 active and standby temperatures to 0C; Automatic saving after power loss is not enabled
; Custom settings are not configured
; Miscellaneous
T0 ; Select first tool[/code]
Bed map file
0_1563076106043_heightmap.csv -
I realized that I skipped a post here as this is on going on 2 forums. Here is what I was getting.
I found a macro posted by dc42 to test backlash, I lost the results, but they were good at the X100 Y0 points. Then I rotated the program to run to the X tower and back, then the backlash shows up there.
I tightened the belts tighter again, I have washers under the heads and nuts so they hold without pulling into the wood(whatever it's spelled). Not much improvement there. They are too tight in my experience with timing belts, but I'm aware my experience isn't really valid here. At home position, the resonate frequency of the belt below the skate is around 130hz. Anyone know a good number?
The only play I can find in the system is that end play in the R4ZZ bearings in the cheapskates. The old ones had it and the new ones have it too. I looked for a couple hours in vain to be sure no one made a suitable bearing this small. It does make sense that it only shows up when the arms are so horizontal and the load is sideways on the skate. Another plus for longer arms, they'll shift that play into a direction of less influence. But there goes more build height. I was sure it wasn't effector tilt, because I assumed it would tilt about the center and that would take an obvious amount to create the error. But no, it tilts about the balls on the arm that is vertical.
So the auto-calibration route is obviously messed with by the crazy numbers at the towers. I added g1 x0 y0 z10 f6000 before every probing point in bed.g to pull the lash the same direction. Looking at the new bed map I should feed inwards before probing instead.
But this map is better for using the machine, I ran a large test print last night and got good results.
Coming from it as a machinist, then machine tool repair tech, I nearly have a new printer designed in my head. I'm certain that I'd be assembling it by now if I'd not spent all this time on this printer. And at this point I feel it is the next step.
Thanks
Cory -
I'm actually excited, I got worse results. The difference is that the pattern has changed a little. Not completely dependent on the direction on travel anymore. This time I ran the belts as tight as I had adjustment for. Now the axis feel pretty bad.
I have checked the bearings and belts several times at this point. Maybe a bearing is going lumpy under load, so hard to tell with the steppers in the system. The servo systems I'm used to don't ever feel rough, unless they are bad. Of course lots of times manually moving them isn't an option at all when they weight tons.
I keep ruling the bearings, pulleys, and belts out since the axis are consistent. But it's the only thing left to try IMO, it's either backlash from belts, or backlash from tight axis. Or both at the same time I think.
Seemecnc sells a flanged bearing for idlers on some other printers. Does anyone have an easy hack for the max v2? Maybe a toothed idler in in the places where the teeth are against the idler? Seems like that would be a smoother solution. I don't know if any line up on the V2. I'm thinking that I might machine out some replacement metal holders for the idlers, maybe screw them together through the bearing so the bearings will stay squared to the frame? Jack screws for the ones to tension the belts? I've tried over and over to get the belts to track perfectly, very little luck there.
Any thoughts on this? I"m going to at least replace the belts, pulleys, and bearings. Since I don't feel like I'm ordering parts that will fix the machine, I'm trying to upgrade as I do it. maker713's metal frame looks interesting, but at that cost I want the V3's bed mount and go with the AC heated bed. Then a 24V power supply would be easier and cheaper. The V3 frame is available but it looks like I'd be missing parts(like motor mounts) if I order the V3 frame kit from there.
Thanks
Cory -
@clytle374 said in Diagonal rod length on delta printer issues.:
Below is a quote by dc42 in the other thread I linked too. Can anyone elaborate?
- Having the arms go vertical or even beyond when printing close to a tower isn't a problem. What you need to avoid is arms going below about 20 degrees to the horizontal
This is a "rule of thumb" to keep errors from accumulating, particularly when "between the towers on the opposite side" like you describe.
@clytle374 said in Diagonal rod length on delta printer issues.:
M665 L291.060:291.060:291.060 R144.257 H368.890 B140.0 X-0.009 Y0.368 Z0.000
Given arm length of 291, delta radius of 144, printable radius (bed) of 140, and a smart effector, I believe you'd be WAY beyond 20 degrees when past the line between towers. Something more like 10 or 15. This was calculated on a thing that helps set up 2020 and 2040 extrusion framed printers with CNC corners... so it may not be perfectly applicable to the Rostock Max V2...
If it is applicable, you need longer diagonal rods. Somewhere around 307.
-
Thanks for the input.
I'm using the stock V2 effector and e3d hotend. Well actually an indicator and no hot end for testing.
I guess I was trying ask when I quoted dc42 is what error is accumulating? I agree it appears to be backlash, but it's not find-able by standard methods of measuring backlash. I understand that once you reach vertical that the end of travel. But until that point error of the horizontal arm should have very small influence in the Z axis, that is nearly solely dependent on the vertical arm. I did find dc42 mention that the delta solving method was 'probably more susceptible to rounding error' is this the issue?
I said a couple posts up that I thought it's effector tilt... I don't believe that to be the case.
I've been back and forth with belt tension. Last attempt was break it or fix it. So at this point, I think I've abused the belts and idlers plenty, bearings aren't free anymore. I'm going to replace them next.
Thanks again
Cory -
@clytle374 said in Diagonal rod length on delta printer issues.:
... error of the horizontal arm should have very small influence in the Z axis, that is nearly solely dependent on the vertical arm.
Hmmm... makes sense.
Just grasping at straws: Looking at the map of the mesh, it certainly looks like a periodic error on one tower. I'm basing this on the "ridges" or "waves" of yellow.
Shaft bent? On a motor or an idler?
======================
And, just another "oh by the way", since you are on a path to replace things: 9mm belts have about half the stretch of 6mm belts. Very worth it on larger printers.
A tiny bit harder to source; they are available. I just replaced the belts and idlers on a larg-ish delta (1.5 meter tall) after two years of heavy printing. Probably not absolutely necessary, nonetheless made me feel better. The point being: I found sources for the 9mm parts much more easily last month than I did two or three years ago when I built that printer.
