Automatic bed level - How does it work?

I'm struggling to get my Ormerod perfectly set up. Right now I have a number of problems but instead of trying to explain it all I'll take it in steps, one forum thread at a time.

Firstly, I believe my bed level compensation is messed up. I use firmware 0.78h-dc42 (reprap.htm/js 0.98) with dc42's diff IR.
config.g: [pastebin.com]
_automatic_bed.g:
M557 P0 X50 Y0
M557 P1 X50 Y180
M557 P2 X220 Y180
M557 P3 X220 Y0
M557 P4 X100 Y100

I start up Ormerod, home x and y. I'm still not comfortable homing z automatically so I go through the steps described in dc42's commissioning guide to calibrate zprobe and home z.

Then I run my _automatic_bed.g file and issue G32.

What exactly is it that Ormerod does here?

The head goes to P0, lowers the z to a level where it stops and proceeds with lower increments. The thing is, even though it lowers with less speed it still goes on and on and on. Even though it already hit the bed. Eventually it decides to stop and goes on to P1 where it doesn't drill through the bed as much as P0. At P2 and P3 it seems to do fine. All Ps have been working OK before. I'm just not sure what triggers it's using and why it goes on even though hitting the bed.

You have to have Z set up and homed before starting the auto bed compensation.

Hi bitfield, what I think is happening is that board returns a reading of 465 when the nozzle is low enough, which causes z-probing to slow down, but even with the head touches the bed it doesn't get the 535 reading. Here are some things to check:

1. The IR sensor head should be at least 10mm in from the corners of the bed and clear of the bed clips at each probe point. Check the P0 and P1 points in particular.

2. Check that the print head is level, by looking along the X axis.

3. In your G30 command, what Z value are you using? If it is less than 1mm, this may indicate that the board is mounted too high up. The bottom edge of the board should be between 1mm and 2mm higher than the nozzle. I amended the design of the modified heatsink duct a few weeks ago, because a couple of users found that the original design mounted the board too high up.

Hope this helps!

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Hi everyones,

I'm a new user of this 3d printer.

I've done all the calibrations but, I have just a remark on the technical choice for z axis.

Because of the choice of using "z gears" between the z-motor and the "M5 threaded z-rod" the automatic bed compensation is not very accurate.

You can see the very little rotations of the z-motor trying to automatically compensate the bed during printing process of each layers. But because of the spaces between the teeth and space between the "z-driven-gear" and the M5 nut inside it, the rod didn't turn (or not as expected).

So an advice is to be very careful when manually leveling the bed, it has to be "near perfect".

On my build, there is no detectable slop in the Z gears or between the driven gear and the Z rod. I have had slop between the Z nut and the nut trap, and I have had to replace the nut trap because of this.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
philrich123
Hi everyones,
I'm a new user of this 3d printer.
I've done all the calibrations but, I have just a remark on the technical choice for z axis.
Because of the choice of using "z gears" between the z-motor and the "M5 threaded z-rod" the automatic bed compensation is not very accurate.
You can see the very little rotations of the z-motor trying to automatically compensate the bed during printing process of each layers. But because of the spaces between the teeth and space between the "z-driven-gear" and the M5 nut inside it, the rod didn't turn (or not as expected).
So an advice is to be very careful when manually leveling the bed, it has to be "near perfect".

There should be close to zero backlash in the Z gears. If the motor turns without the rod turning it is likely that a gear is slipping on the rod or motor shaft. In any case as much as 10 degrees of backlash (which is almost a full tooth width) will change the Z height by less than 0.015mm.

Dave

Quote
dc42
Hi bitfield, what I think is happening is that board returns a reading of 465 when the nozzle is low enough, which causes z-probing to slow down, but even with the head touches the bed it doesn't get the 535 reading. Here are some things to check:

1. The IR sensor head should be at least 10mm in from the corners of the bed and clear of the bed clips at each probe point. Check the P0 and P1 points in particular.

2. Check that the print head is level, by looking along the X axis.

3. In your G30 command, what Z value are you using? If it is less than 1mm, this may indicate that the board is mounted too high up. The bottom edge of the board should be between 1mm and 2mm higher than the nozzle. I amended the design of the modified heatsink duct a few weeks ago, because a couple of users found that the original design mounted the board too high up.

Hope this helps!

1. This looks normal!

2. It looks leveled to me.

3. Z in G30 is usually between 0.5 and 0.9. So maybe it is to high up.I would say it's at least 2mm above nozzle. Not to much more than that though. You're saying I should try to print a modified version?

The sensor is behaving real strange on P1. If I center head over X60 Y180 I can lower the nozzle until it reaches the bed and still have zprobe report 0. If I put a sheet of paper under the sensor it reports 530. Also, if I push the head down it eventually reports 460 but the head is no longer level.

Bitfield, please check again whether the sensor head is 10mm in from the corners at P0. The ideal bed probe coordinates when using my board are not the same as when using the original RRP board. If Y=0 places the nozzle close to the edge of the bed, then the sensor head cannot be 10mm in from the edge at X50 Y0. It would need to be more like X50 Y10. The original IR sensor may have been OK at Y0 because it is offset from the nozzle in the Y direction, however my sensor board places the sensitive area in line with the nozzle in the X direction, so there is little or no Y offset. This is by design, so that the nozzle height measured is less sensitive to head sag. The sensitive area is approximately under the round black capacitor at the bottom of the board.

On my build, the Y axis travel is 210mm, and I place the corner bed compensation points at Y=20 and Y=200 (although Y=20 and Y=190 might be better).

Edited 1 time(s). Last edit at 12/14/2014 08:43AM by dc42.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
dc42
Bitfield, please check again whether the sensor head is 10mm in from the corners at P0. The ideal bed probe coordinates when using my board are not the same as when using the original RRP board. If Y=0 places the nozzle close to the edge of the bed, then the sensor head cannot be 10mm in from the edge at X50 Y0. It would need to be more like X50 Y10. The original IR sensor may have been OK at Y0 because it is offset from the nozzle in the Y direction, however my sensor board places the sensitive area in line with the nozzle in the X direction, so there is little or no Y offset. This is by design, so that the nozzle height measured is less sensitive to head sag. The sensitive area is approximately under the round black capacitor at the bottom of the board.

On my build, the Y axis travel is 210mm, and I place the corner bed compensation points at Y=20 and Y=200 (although Y=20 and Y=190 might be better).

I'm being really busy that's why it's taking so long for me to reply. I will verify ASAP but I'm pretty sure it's more than 10mm. More like 20-40mm actually. But I'll verify before ruling this theory out for sure.

The printer feels pretty unreliable. Now when I did the automatic bed compensation it performed much better. After that, I moved in all the corners even futher just to make sure. It still performs OK in the sense it doesn't hit the bed. But the distance from bed when it stops differs between the different points. How much is hard to say but it's visible to the eye at least. I tried to record it with a camera but it was hard to get focus and also, the white illuminating LED causes a lot of reflects so it's hard to see.

1. Is it possible to turn off the white LED?
2. Is it normal for the automatic bed compensation to stop at different hights above the bed on different P's? I thought the point was to have them aligned.
3. It prints really bad on one half of the bed (where X < 100). See this video! Why is that?

Hi Bitfield,

First, please confirm that you are using M558 P1 to activate the sensor, and never any other P number.

Quote
bitfield
But the distance from bed when it stops differs between the different points. How much is hard to say but it's visible to the eye at least.

Can you try to measure it? First, ensure that the nozzle has no plastic melted on the end. Then, for each of the bed probing XY coordinates, move the head to that position, do a G30, then put a sheet of paper under the nozzle and lower the nozzle in 0.05mm steps until it just grips the paper. Then read off the Z reading (which will ideally be zero) from the web interface. It may also be helpful if you post a photo of the print head at each probe point.

Variation in the bed surface will cause some change in the readings, but much less than with an ordinary modulated sensor. So if you are using strips of Kapton tape on the bed, I suggest you avoid probing at locations that put the sensor over one of the joins between strips of tape. Looking at your video, you appear to be using a single sheet of Kapton, however at the Y=0 end it looks like you may have an area where there are 2 layers, because the colour is darker.

Quote

1. Is it possible to turn off the white LED?

Only by modifying the board or removing the 12V supply from the board. You can remove the 12V supply by disconnecting the hot end fan connector at the Duet. Remember to reconnect it before heating the hot end!

Quote

2. Is it normal for the automatic bed compensation to stop at different hights above the bed on different P's?

No, the heights above the bed should be close to equal, provided you have chosen suitable coordinates that place the sensor head away from the bed clips and not too close to the corners, and the bed surface is reasonably uniform.

Quote

3. It prints really bad on one half of the bed (where X < 100). See this video! Why is that?

I think the nozzle is starting too close to the bed on that side.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

First of all. I really appreciate your help and patience, dc42. Right now I just want to get started printing again.

Quote
dc42
First, please confirm that you are using M558 P1 to activate the sensor, and never any other P number.

Confirmed.

Quote
dc42
Can you try to measure it? First, ensure that the nozzle has no plastic melted on the end. Then, for each of the bed probing XY coordinates, move the head to that position, do a G30, then put a sheet of paper under the nozzle and lower the nozzle in 0.05mm steps until it just grips the paper. Then read off the Z reading (which will ideally be zero) from the web interface. It may also be helpful if you post a photo of the print head at each probe point.

Is G30 necessary? Anyway, with this technique I got the following values:
P0 +0.3
P1 +0.3
P2 -0.2
P3 -0.1
P4 -0.05

Ideally, it should read 0 on all of them but it's not unfortunately. P4 is probably not 0 because I used 0.1 steps when I calibrated. Would photos help or is there anything else I can do? These values are somewhat confirmed when I print a large circle on the bed. On high X's (>100) I get a "normal" extrusion but on small X (<100) the filament is squeezed to the bed a lot more.

Quote
dc42
Variation in the bed surface will cause some change in the readings, but much less than with an ordinary modulated sensor. So if you are using strips of Kapton tape on the bed, I suggest you avoid probing at locations that put the sensor over one of the joins between strips of tape. Looking at your video, you appear to be using a single sheet of Kapton, however at the Y=0 end it looks like you may have an area where there are 2 layers, because the colour is darker.

Correct, the kapton tape is 200mm wide. At Y=0 (on both ends actually) I've folded the tape over on the other side. That's why it looks darker.

Not sure if now's the time to introduce even more strange behavior but maybe it's related. Trying to print this big pixel star something's happening when it's at the left side of the star (~ X60 Y100) doing the contour on first layer (not sure about the second layer since I stopped). At two points it seems to run the Z motor heavily. Have a look at this video. It's really hard to see but you can hear the bad sound coming from the Z axis/motor. The model I'm trying to print is actually a scaled version of the original STL. Here's the gcode file.

Edited 1 time(s). Last edit at 12/17/2014 01:53PM by bitfield.

I'm sorry, I should have mentioned that you need to raise the head to 5mm or more before you do each G30. What G30 does is a single bed probe at the current XY position, and then reset the Z value to whatever you said in your G31 command in your config.g. So by raising the head to around 5mm, doing a G30, and then measuring how high the head is actually off the bed (or equivalently, lowering the head to grip a piece if paper and then reading off the Z height from the web interface), you can check whether the trigger height really is different at the 4 or 5 probe points.

One other thing to do: look along the X axis. Is the X plate straight, or does it have a slight twist or wave in it? That would cause the head angle to vary, which will also affect the sensor reading slightly (but not as much as with the standard Ormerod probes).

PS - the folded over Kapton could affect the readings slightly as well, because the reflection from the bottom of the glass contributes slightly to the reading. However, I wouldn't expect that to give you an error in the X direction. Nevertheless, it's probably a good idea to ensure either that all the probe points place the sensor over the double-up Kapton, or that none of them do.

PPS - if you are using the web interface, set "Half Z steps" on the Settings tab to get 0.05 mm steps.

Edited 3 time(s). Last edit at 12/17/2014 02:23PM by dc42.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

None of the points are over the double kapton tape. Well inside actually. My current points I'm experimenting with are:
M557 P0 X70 Y20
M557 P1 X70 Y160
M557 P2 X200 Y160
M557 P3 X200 Y20
M557 P4 X100 Y100

By "along x-axis", do you mean looking from X200 towards X0? It might have been slightly bent to the right. I could just straighten it up. Did a re-calibration but with the same result.

Here are my findings (I measured from bed to bottom of the X-carriage, because it was easiest):
Point - mm above bed after G30 - Z value after lowering until gripping a sheet of paper
P0 - 10.5 - +0.3
P1 - 10.5 - +0.35
P2 - 11.0 - -0.2
P3 - 11.0 - -0.15
P4 - 11.0 - +-0

But frankly it's really hard to get a decent read from the ruler. When running G32 and by looking with my bare eyes, it looks like P0 and P1 stops closer to bed than P2 and P3 where P4 is somewhere in between.

The half z step trick was nice. I've been fiddling with G91, G1 Z+-0.05 and back to G90 for that purpose previously.

Anywho, I'm out of ideas. Really not sure what's going on with my printer. But it's of course even harder for you trying to troubleshoot a printer you don't have access to.

Edited 1 time(s). Last edit at 12/17/2014 03:37PM by bitfield.

Quote
bitfield
But frankly it's really hard to get a decent read from the ruler. When running G32 and by looking with my bare eyes, it looks like P0 and P1 stops closer to bed than P2 and P3 where P4 is somewhere in between.

It's best to use the paper method rather than a ruler. After performing each G30, raise the head 2mm or so, put a piece of ordinary copier paper under the head, and lower the head 0.5mm at a time until you can just feel the paper being gripped as you slide it around under the head. Make a note of the Z height displayed at that point, raise the head 5mm and do the same with the next location.

Dave

OK, the Z values for P0/P1 are about 0.5mm higher than for P2/P3. This is not normal. I suspect that the head angle is varying as the head moves along the X axis, probably due to a twist in the acrylic X-plate. It must be quite a large variation because my board is designed to be less sensitive to the head angle than the standard board, by placing the sensitive area in line with the nozzle n the x-direction.

Please can you take a couple of photographs looking along the X axis, one with the head at the P0 point, and the other with the head at P3, so I can see whether this is happening.

Note that on the earlier Ormerod 1 kits, the tabs on the X rib often wouldn't fit the slots in the X plate without imparting a twist to the X plate. This can be avoided by filing out the ends of one or more of the slots a little (or filing down the ends of one or more tabs) until the X rib just drops into the X plate with very little force.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Not sure how it's possible to determine anything from these images. Please let me know if I should use a better camera (I didn't find the CF card reader so I took photos with the smartphone).

As you say, it's hard to tell anything from those images. Can you take some more photos, this time with the lighting off? You can either disconnect the hot end fan connector at the Duet, or just turn the power off when the head is in the right position. What I am particularly interest in is the relationship between the bottom edge of the board and the bed - and anything else in the vicinity such as the bed clips.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

I'll come back with better images tomorrow. Just got myself a new card reader for the camera.

I tried to take some pictures but I'm not sure it's worth it. It's really hard to get any decent photos that's of any value. Anyway, while doing this I realized that it might very well be the x plate that is bended in some weird way. It's hard to tell by just looking at it.

This is what I did.
Put the nozzle over P2. Raised head just before going from 5xx to 4xx. I.e red LED on sensor board is lit. Then I moved the head closer to P1 but in small steps. After just a few steps (like at X150) the LED went black. Then I twisted the x-plate gently with my hand. Pushing the top of the end of the plate towards Y0 and the bottom of the plate towards Y200. It didn't require much before the LED was again lit. So I continue towards P1. Every now and then I had to twist the x-plate harder in order for the LED to lit. Does this give any clue at all? Even if it is the x-plate that's twisted, is there anything to do about it? I tried to look closely at the nozzle and the sensors while going from P2 to P1 to see if it turns but it's really hard to tell. Maybe a little, but again I'm actively trying to look for that behavior so I might be biased.

Edited 2 time(s). Last edit at 12/22/2014 03:44PM by bitfield.

Hi Bitfield,
When I had the original acrylic X plate/rib (now have DaveK's aluminium one, excellent) I found that my X axis was similar to what you are describing. What I did was to loosen the X idler a touch and 'twist' it on the rod to get the X plate/rib as near as I could to vertical. However, be very careful to not twist it too much as the acrylic plate WILL break at the point where it goes into the idler where it has a slot cut into it.

Pete.

Hi bitfield,

The sensor is designed to have low sensitivity to head rotation about the X axis, so if that is the problem then I would expect you to be able to see it.

Here are a few things for you to try:

1. At the P0 or P1 position, check that the nozzle is truly vertical. Adjust the X idler if necessary until it is. Then check that the bottom edge of the sensor board is parallel to the bed. The mounting holes in the sensor board are a little larger than the screws, so you can turn the board a little if you need to. If the edge of the board is not parallel to the bed, then it will be more sensitive to head rotation about the X axis.

2. To check whether the head is rotating significantly about the X axis as it moves long the X axis, measure the height of the bottom of the fan duct above the nozzle. To do this, first lower the nozzle until it just grips a sheet of paper. Then see how much you have to raise the head from that position so that you can just get the shank of a 3mm drill between the bed and the bottom of the heatsink duct. Do this at P1, at P2, and halfway in between.

3. One thing that would definitely cause the sensor to vary its trigger height (relative to the tip of the nozzle) as the head moves along the X axis is if the head rotates a little about the Y axis as it moves along X. This doesn't sound likely to me, but could perhaps happen if the X carriage isn't a good fit on the bearing, or if the head of the X idler screw catches on the extruder motor. To check this, you can repeat step (2) above, but this time after lowering the nozzle to grip the sheet of paper, see how much you need to raise it to just get the drill shank under the edge of the sensor board.

4. Check that the IR emitters and phototransistor are at the correct angles, as shown here [miscsolutions.files.wordpress.com]. Also look at them from the side to make sure they are at right angles to the board.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

I'll get back to your bullets later. Just wanted to quickly run a separate question. Is this issue in the print caused by a bad bed level?
As you can see, one of the edges of the star is slightly curved in the bottom. This edge happen to be closest to P0.




Edited 1 time(s). Last edit at 01/01/2015 06:49AM by bitfield.

That curvature indicates that the print initially stuck to the bed in that corner but lost adhesion there after some layers were printed. The lack of continued adhesion may or may not be connected with bed levelling.

Edited 1 time(s). Last edit at 01/01/2015 07:53AM by dc42.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Yes, that photo shows a classic case of warping. Besides incorrect Z zero height, another cause is the backwash of the fan cooling the fan side of the print too much - there are several designs for fan deflectors to prevent that. Bed and hotend temperatures also play a part and should be tweaked for best results, and also try to eliminate draghts over the bed while printing. ABS is far more prone to warping than PLA.

Dave

I have now ordered the aluminium x arm kit from ddmetalproducts and will put this issue on hold until I have replaced the acrylic parts. Keeping my fingers cross.

Alright, I've been having my hands busy with other stuff lately but tonight I finally finished off the installation of the aluminium arm and just ran the automatic bed level. Unfortunately, the issue is still present. These are the values:

Point - Z value after G30 - Z value after lowering until gripping a sheet of paper
P0 - 1.35 - +0.25
P1 - 1.35 - +0.05
P2 - 1.35 - -0.25
P3 - 1.35 - -0.25
P4 - 1.35 - +-0

What's my next step to try to get to the bottom with this?

My best guess is that it's caused by the varying angle of the Bowden tube producing more or less sideways force on the head, which affects the head tilt, and this is turn is causing the probe threshold to vary. But I am surprised the variation is so high. You could test this theory by removing the Bowden tube and measuring the trigger heights again. Alternatively, hang the extruder drive from a different slot on the X arm, and measure again.

If you can't eliminate the variation in trigger height, then if you upgrade your firmware to my latest 1.09e version, and use a bed.g file instead of the M557 commands, you can enter a height correction for each probe point. This is what I think you need in your bed.g file:

M561
G30 P0 X70 Y20 Z-100000 H0.25
G30 P1 X70 Y160 Z-100000 H0.05
G30 P2 X200 Y160 Z-100000 H-0.25
G30 P3 X200 Y20 Z-100000 H-0.25
G30 P4 X100 Y100 Z-100000 H0 S0

The H parameter on the G30 commands specifies the trigger height adjustment at each point. Adjust the XY coordinates of the probe points if necessary. If you have a bed.g file in the /sys folder on the SD card, then it will automatically be executed when you run G32, and the probe points set by M557 will be ignored.

Edited 1 time(s). Last edit at 08/11/2015 07:07PM by dc42.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
dc42
My best guess is that it's caused by the varying angle of the Bowden tube producing more or less sideways force on the head, which affects the head tilt, and this is turn is causing the probe threshold to vary. But I am surprised the variation is so high. You could test this theory by removing the Bowden tube and measuring the trigger heights again. Alternatively, hang the extruder drive from a different slot on the X arm, and measure again.

I'm not sure I do the measuring correctly. Do I need to do G32 again after moving the extruder drive to another position? Because G30 doesn't really seem to do anything with my probe as suggested earlier in this thread. At least it stops at the same Z for all 5 points. It's always the same Z as I set with G31 during calibration. I moved the extruder to the far end of the X arm and re-run G32 but I didn't see any difference. Still -0.25.

Quote
dc42
If you can't eliminate the variation in trigger height, then if you upgrade your firmware to my latest 1.09e version, and use a bed.g file instead of the M557 commands, you can enter a height correction for each probe point. This is what I think you need in your bed.g file:

M561
G30 P0 X70 Y20 Z-100000 H0.25
G30 P1 X70 Y160 Z-100000 H0.05
G30 P2 X200 Y160 Z-100000 H-0.25
G30 P3 X200 Y20 Z-100000 H-0.25
G30 P4 X100 Y100 Z-100000 H0 S0

The H parameter on the G30 commands specifies the trigger height adjustment at each point. Adjust the XY coordinates of the probe points if necessary. If you have a bed.g file in the /sys folder on the SD card, then it will automatically be executed when you run G32, and the probe points set by M557 will be ignored.

Thanks, I will proceed with upgrading the firmware to latest. Current version is 0.78h-dc42 so I guess it's pretty old as well.

Edited 2 time(s). Last edit at 08/12/2015 06:38AM by bitfield.

Stupid question. But latest Arduino doesn't seem to have the bossac binary any more? Where do I find it?
edit/ Nevermind, found it on sourceforge.
However, I do get Auto scan for devices failed, after doing Erase / Reset.

edit2/ Nevermind again. Switched to Windows and it worked.

Edited 2 time(s). Last edit at 08/12/2015 07:23AM by bitfield.