Auto_Bed_Leveling: Enhanced G29 command

[Roxy]

Your bed is almost flat. It is much better than mine. I have a high spot in one place and a low spot in another place. The only issue I see with your bed is your right hand side looks a little bit low. Can you level that back left corner a little bit more? It should be OK where it is at, but everything else looks pretty good. It seems you should be able to raise the back left just .2 or .3 mm and have everything really good.

I don't know what is causing your bed leveling problems. It might be time to make a short video of it doing that same print you have up above so we can see how the z axis is compensating and what the plastic looks like as it comes out of the nozzle.

[Astek777]

Hi Roxy

Unfortunately the back left corner is pretty warped, and there isn't anymore adjustment available...I'm already using longer M3 screws.

I've uploaded a video as recommended. It does a 'G28', followed by 'G29 V4 T n5'.

https://youtu.be/WzPmOrYJEh8

Thanks again.

[Roxy]

Wow! Everything looks good but that one corner. It almost looks like you are too close to the glass for the first 3 legs of the print, but on the 4th leg it is too far away. I don't know what is going to fix it. But I would suggest slowing the Z axis down for the probing. Your M48 numbers seemed good, but that is faster than I would dare take measurements. It may be with more accurate measurements things get better.

The other thing I would try is another piece of glass. It maybe that piece is warped too much. Do you have a glass cutter? At the Dollar store you can usually get a cheesy picture frame with some thin glass for $1.00. It probably is worth a try just to see what happens. Also, what happens if you rotate the piece of glass, or maybe even flip it? Does the problem area move?

PS. That looks like a really nice printer!!!

[Astek777]

Hi Roxy

Thanks. I think the printer will be great if I can just get some reliability.

It does seem strange. When I made the bed really wonky it printed exactly the same as it is now, so it seems like it's always out by the same amount in the same places. I had tried printing on top of a 300mm square sheet of aluminium just resting on top of the glass (not clamped down). It's thick enough not to bow by itself, but this didn't seem to make any difference either. Exact same issues in the same places.

I'm in the UK, so not so easy to get cheap glass, but I'll have a hunt around and see what I can find.

There's nothing configuration wise that could be causing the issue. I don't have any X-offset, but do have 32mm of Y-offset. Everything's printing in the correct place though, and the probe points, as you can see from the video, are perfectly set around a 250mm square in the centre of my print bed.

Just to confirm how the topographic map works, what point does it use as it's base point to decide what is plus and what is minus? It can't be the default Z-home position or the centre probe point would always be zero. Is it just a mean value? Also, are the plus points the higher positions on the Z-Axis OR the points further away from the probe?

Whereabouts in the firmware is the Z-Axis speed set, or would that be in the print software (for me Simplify3D)?

[Roxy]

There's nothing configuration wise that could be causing the issue. I don't have any X-offset, but do have 32mm of Y-offset. Everything's printing in the correct place though, and the probe points, as you can see from the video, are perfectly set around a 250mm square in the centre of my print bed.

So, Like said, I don't know what is causing the problem. But one thing we could do is modify the code to probe more points in the bad corner. What that would do is weight that corner of the plane heavier when trying to do the least squares best fit. Or, a quicker thing to try would be to turn on the 3-Point bed leveling and see if you can adjust your 3 points to give a good leveling.

Just to confirm how the topographic map works, what point does it use as it's base point to decide what is plus and what is minus? It can't be the default Z-home position or the centre probe point would always be zero. Is it just a mean value? Also, are the plus points the higher positions on the Z-Axis OR the points further away from the probe?

The Topography report finds the mean of the measured points. And from that it displays how much plus or minus a given point is from the mean. The plus points are higher up and the minus points are lower down.

Whereabouts in the firmware is the Z-Axis speed set, or would that be in the print software (for me Simplify3D)?

It is in the Configuration.h file of Marlin. The probing is controlled from the Homing Feedrate. This is what I have in my printer. Probably, only the 3rd number needs to change. The numbers are for X, Y, Z and the extruder.

#define HOMING_FEEDRATE {50*60, 50*60, 4*60, 0} // set the homing speeds (mm/min)

[Astek777]

As recommended, I reduced my HOMING_FEEDRATE speed for the Z axis to 4x60 (previous 8x60). I also tried an 8x8 grid, just on the off chance but neither seems to have helped. I'm always amazed it's off by the same amount and in the same places every time...I would have thought scanning more points would affect something?!

Anyway, I thought I'd have a go at the simple 3 point plane. I'm getting some compile errors with this though. I commented out:

//#define AUTO_BED_LEVELING_GRID

...but then I got...

"AUTO_BED_LEVELING_GRID_POINTS not declared in this scope" referring to:

if ( code_seen('n') || code_seen('U') || code_seen('u') ) { n_points = code_value();
if (n_points<2 || n_points>AUTO_BED_LEVELING_GRID_POINTS ) {
SERIAL_PROTOCOLPGM("?Number of probed points not plausable.\n");
break;
}
}

...in marlin_Main.cpp. I tried declaring it, just to see if I could get past this, but then I received...

"plane_equation_coefficients was not declared in this scope" instead.

Is anybody actually using 3 point probing successfully?

Can you briefly explain the different between the two approaches? I'm aware it's obviously a simpler plane it makes, with just three points, but how does it differ from say doing a simply 2x2 grid which would cover the entire bed?

Also, as far as three-point probing goes, I'm not really sure where these points should be set. I've currently set them back-left, right-middle, centre-front, covering all of the main axes at least once, and including the point (back-left) which is the most out. Is that the best approach?

Thanks as always!

[Roxy]

As recommended, I reduced my HOMING_FEEDRATE speed for the Z axis to 4x60 (previous 8x60). I also tried an 8x8 grid, just on the off chance but neither seems to have helped. I'm always amazed it's off by the same amount and in the same places every time...I would have thought scanning more points would affect something?!

Anyway, I thought I'd have a go at the simple 3 point plane. I'm getting some compile errors with this though. I commented out:

//#define AUTO_BED_LEVELING_GRID

...but then I got...

Ooooops.... The BeckDac fork has a bug I forgot about. It doesn't have those lines turned off if 3-Point is enabled.
Put some comment markers around that block... And it will compile and work. Sorrry about that!

/*
if ( code_seen('n') || code_seen('U') || code_seen('u') ) {    n_points = code_value();
    if (n_points<2 || n_points>AUTO_BED_LEVELING_GRID_POINTS ) {
        SERIAL_PROTOCOLPGM("?Number of probed points not plausable.\n");
        break;
    }
}
*/

Is anybody actually using 3 point probing successfully?

Yes. Lot's of people prefer it because it is much quicker. The problem is, your mechanics have to be pretty good. Any way, you will get to pick where your 3 points are. If the first set of points you pick doesn't work well, try a 2nd and 3rd set. It is very possible you will be able to get that to work well for you.

Can you briefly explain the different between the two approaches? I'm aware it's obviously a simpler plane it makes, with just three points, but how does it differ from say doing a simply 2x2 grid which would cover the entire bed?

The 3-Point code uses the fact that 3 non-linear points define a plane. So it uses those 3 points to figure out how the bed is aligned. The Grid code does a least square error calculation to find the plane that 'sort of' goes through all the points with the least amount of error. The grid code has to do way more calculations to come up with its number.

Also, as far as three-point probing goes, I'm not really sure where these points should be set. I've currently set them back-left, right-middle, centre-front, covering all of the main axes at least once, and including the point (back-left) which is the most out. Is that the best approach?

Thanks as always!

Yeah... There are differences of opinion. But doing a front and back corner on one side, and then doing the middle on the other side seems to get consistently good results. When I first did 3-Point leveling, I picked 3 corners, and that worked well as long as my part did not go off towards the un-probed corner. If it did, that part of the print did not work very well usually.

[AlexD]

Hi Roxy,

I have read through a pretty decent chunk of this thread. I have implemented ABL on my printer but as others, am experiencing some problems with it. I have attempted to implement the latest version of your Marlin Fork to try and get a typography report. However the grid was far too small for my bed, (450 x 450) and i was unable to change the amount of measurements taken. Are you able to shed some light on this for me? Or can i easily add your adapted ABL code into my existing version of Marlin?

If you get a chance to write back it would be greatly appreciated, i havn't had a successful print bigger than 1" squared in weeks and its killing me and motivation to get this printer setup right! Just for reference i have a home built gMax 1.5 Printer.

EDIT: I somehow managed to miss the first post of this thread (went straight to page 2?!) i will get your code into my firmware and report back.

Thanks

Alex

[Astek777]

OK, commented out that code group as suggested, and another three lines further on that also stopped it from compiling:

free(plane_equation_coefficients);
free(eqnAVector);
free(eqnBVector);

I assumed as these were being 'freed' that would be OK?

3 point plane didn't make any difference either. Tried three corners and rotating this round four times so all combinations tried...still roughly the same, although this time the front-left was a little too low and actually slightly scrapped the aluminium with the nozzle head...d'oh! Luckily the print bed is sprung at the moment so it has a bit of give, but I'm hoping I can fix it solid once the levelling is sorted.

I've ordered a replacement sheet of glass to try.

Obviously still trying to get to grips with how this all works, but, assuming the glass is perfectly flat (or close enough!), scanning a simple 2x2 grid (as close as possible to the 4 corners), should produce a perfectly flat plane, even if this plane is angled...is that correct?

If I'm still having the same problems, what does that mean? What else can cause this sort of thing? The measurements which are being taken are very accurate and repeatable (proved by the tested M48s).

Just a few thoughts which I wonder could have any impact...

* Is there anything that could be being loading from the EPROM memory that could be distorting the map?

* When the G29 is being completed I get 'Z endstop hit'...could this be stopping figures lower than the G28 Z-home position from being correctly registered?

* Is there anything else that could have warped that would cause this?

* If I've got software max-stop points set and the nozzle position moves outside of these during probing, which mine does as the probe is 32mm in front (Y axis) of the extruder and the probe scans right to the back corners so the extruder is just about off the table at this point, would that cause any issues?

* If one of the offsets is set to '0' (my X offset is), is there anywhere in the code where this would read as empty and load some default value or cause a 'division by zero' or similar?

As always, appreciate your on-going support...after 67 pages I suspect you're getting a bit fed-up with it by now, so thank you.

[Astek777]

Hi Alex

Sometimes it's easier if you attach your configuration.h file so it can be checked out, but the first page of the forum had lots of helpful information about how to use the G29 command. Basically though:

'G29' <- will use the settings from your configuration.h file
'G29 V4' <- turns on debugging/verbose at it's highest level (doing this from memory...pretty sure that's right).
'G29 T' <- will show you a topographical map upon completion.
'G29 nX' <- replace X with the number of points you want to scan (maximum of what you have set in your configuration.h file). So 'G29 n4' would scan a 4x4 square uses the boundaries set in your configuration.h file.

You can also use a combination, so 'G29 V4 T n5' will scan a 5x5 square, provide full debug information and show you the topographical map upon completion.

Remember, if your printer uses EPROM memory you sometimes have to run M502, followed by M500, when you've updated your firmware. This loads default settings from your configuration (M502) and saves them back to your EPROM memory to be automatically loaded in future (M500). This might be why the Z-offset you were changing wasn't updating...I had that. :-)