Help with autocalibration using FSR sensors and Duet board

[shodapublicity]

I have been trying endless to no avail to level my Rostock Max v2. I can get prints that stay within the center to print really well but anything on the edges and it does not. Spend my whole entire day yesterday trying out various things to get my FSR sensors to trigger more evenly across all points but still nothing. Link below is an image with a spiral showing where I am having issues getting the bed leveled.

The area between the Z and Y towers, and basically anything in the center. They are too close to the nozzle so that is why you do not see any filament.
https://imgur.com/PoXXBAQ

I am running a Duet board running DC42 (RepRap fork), and using FSR sensors.

[Jimustanguitar]

I've had mixed results as well. The issue I'm seeing is directly inline with my X tower's FSR about 100mm from the center. It's a much larger usable area that I've ever been able to get out of my Rostock, but I still want to figure it out and to be able to print edge to edge on one of my machines.

[img]http://i.imgur.com/7V3Th3F.jpg[/img]

The brims and first layers are beautiful everywhere else.
[img]http://i.imgur.com/FrIPyHb.jpg[/img]


I'm going to investigate whether moving the sensor's wire, and then moving the sensor itself makes the symptom move. It's also the very first point that the routine probes, and I can see the hotend deflect a little bit, so it's pressing harder in that spot than the others. I'm going to try changing the probing routine to hit a different spot first to see if that makes any difference, too...

I'll share my findings later when I've poked it a little bit more.

[shodapublicity]

I noticed that my bed is more sensitive than others and I believe that is why I am running into this. I've tried various mounts, and techniques but yet still nothing.

I tried manually following along the hotend with the FSR but even that did not work...

[shodapublicity]

Kept at it and then tried printing something to test. Was paying attention to my other printer and this happened... Its not perfect, but it is a lot better than before. Victory!

[Jimustanguitar]

Interesting that you've got the same tight spot down by the X tower that I do... I'm suspicious that the first point probes heavier than the others. I'll try a different script this evening and report back.

[shodapublicity]

I watched as the skirt was printing and I noticed that my bed is not level at all. It varies off and on but due to the starting Z offset that I set it offset most places and allowed for printing. More work left to do, but for now I am happy.

I am going to order some Hall Effects for my endstops as I recently came across a post where a guy built his own custom delta but was running into issues and noticed that switching from mechanical to hall effect endstops halved his issues. His Z-probe was still not quite right but I am thinking that since my Z-probe seems to be working quite well now the hall effect endstops will solve the rest.

[Jimustanguitar]

I did change my bed.g file to try some different probe points, and I'm still having trouble. The file I tried did 0,0 first, then 8 points on a 250mm circle, and 6 points at 125mm. Same results, so it's not a math anomaly or fluke.

This is probably what's going on. I can see my hotend deflecting more in the point that's too tight, so I should play around with some H values.
http://reprap.org/wiki/Configuring_and_ ... the_hotend

How would one go about measuring those points? Seems like it would just be trial and error for now... Maybe I'll make an STL with a little circle at each of the probe points and then measure them with calipers...

[DeltaCon]

Although I do not run a duet, I think I can second Jim's experience where the hotend seems to need to push harder at some probing points than on others. My fsr's trigger nicely when tapped, but I notice that at the "between towers" points two of the three leds should trigger and for me, between the X and Z tower that does not always happen. Also probed at X0 Y0 all three should ideally trigger at once. The hotend also seems to strain just a bit more, but on the other hand the bowden tube and wiring can also pull more at the hotend on one point compared to other points. I still have to start real testing prints because *ahem* some other problems... But all this is proof that real great calibration can be made very difficult by things like the weight of the wiring, and probably also friction in the fsr mounts that is not equal at all temperatures (it can strain at 60 degrees while it is strainfree at 90 or the other way around). Having to countermeasure that by a new manual calibration method seems a bit tedious and probably not very repeatable.

[HumanLiberty]

shodapublicity - Congrats on solving this. I'v been having a very similar problem, and had pretty much given up on my FSRs and gone back to
turning bloody screws, and sliding around MFn paper. You've given me hope. Would you mind sharing the configurations you're using?

I'm still using the original RAMBo, and was working with RollieRowland's OPENDact for FSR calibration software, which I was never able to make
respond to the FSR signals.
I'd really appreciate any help/info.

[shodapublicity]

I did change my bed.g file to try some different probe points, and I'm still having trouble. The file I tried did 0,0 first, then 8 points on a 250mm circle, and 6 points at 125mm. Same results, so it's not a math anomaly or fluke.

This is probably what's going on. I can see my hotend deflecting more in the point that's too tight, so I should play around with some H values.
http://reprap.org/wiki/Configuring_and_ ... the_hotend

How would one go about measuring those points? Seems like it would just be trial and error for now... Maybe I'll make an STL with a little circle at each of the probe points and then measure them with calipers...

Have you seen this blogpost? I believe he frequents this forum so you might already know him. Anyhow, in his blog he describes how you go to each point and use the old "paper grip" test. You would go to each coordinate and lower the probe using the paper grip test and then use the reported Z offset as the H value in your bed.g. I have not yet tried this as his blog post around this step was confusing but after reading your post somehow it clicked and I finally understood what he meant, or at least I think i do.

[shodapublicity]

shodapublicity - Congrats on solving this. I'v been having a very similar problem, and had pretty much given up on my FSRs and gone back to
turning bloody screws, and sliding around MFn paper. You've given me hope. Would you mind sharing the configurations you're using?

I'm still using the original RAMBo, and was working with RollieRowland's OPENDact for FSR calibration software, which I was never able to make
respond to the FSR signals.
I'd really appreciate any help/info.

I am at work, but I'll try and remember to upload my configuration once I get home.

On a side note, my next plan is to do what I was talking about in my post above. About going to each probe point and using the paper grip method to modify the H value of each probe. If this does not work I plan on mounting the FSR directly under the wooden frame under the heated bed. Currently I have the FSR mounted on the edges in the spaces that the wooden piece under the heated bed has. My thoughts is that if I position it directly under it then the sensors will be a more sensitive since what I have noticed is that the less in sensitivity is due to the standoffs that are used to hold the bed down. They're too rigid so probe points that are close to that do not trigger as well as other places.

[shodapublicity]

Here are my configuration files. My config.g and bed.g files are attached.

[Jimustanguitar]

I did this process with the paper trick, and am very happy with the results.

If it helps anyone, I did also create an STL with a little 1 layer high (.3mm) disc at each of the 13 probe points so that you can print and measure them if you really want to.

http://www.thingiverse.com/thing:1598129

[shodapublicity]

Awesome, glad to hear that that worked out Jim. I am going to give it a shot too once I have some time. Question before that though, how were you able to go into negative 0 values? I've tried before and I cannot go pass 0?

I actually saw you post this on Thingiverse first and then came here to reply haha.

[Jimustanguitar]

I ran the auto calibration first, which makes the whole bed at 0 or above (due to deflection), plus the thickness of the paper, so I didn't run into the need to go into the negative Z's at all.

I documented the Z height for all 13 points that gave me a good feeling on a piece of paper, shifted all of the numbers so that the mode was 0 (no need to change the most frequent measurement, it's much easier to define 0 with it), and added in the appropriate "H values" to my bed.g script. Lather, rinse, repeat.

[Jimustanguitar]

A note to anyone doing the "paper drag" test. The height that you measure the paper at, whether positive or negative from Z0, should be subtracted from your current H value. The polarity seems backwards to intuition (for me at least), you're changing the overall Z height by that amount, not moving 0 by that amount. I.E. if your paper drags nicely at +0.02, you need to subtract .02 from your current H number.

A tight spot is fixed by decreasing your H value and shortening your resultant Z travel at that point, and a loose spot needs a higher number to flatten out.

Hope that makes sense and keeps someone else from doing the numbers backwards and chasing your tail trying to calibrate.