Heuristic (AI) calibration for delta printers on Smoothie

[mhackney]

Thanks for the explanation! I have yet to have SA make it through without stalling and I am still getting hangs after reconciling the config file. I'm working on a mini kossel but I used a BA caliper and a granite slab to assemble it as precisely as possible. Given the results of the simple G32 it seems pretty good - of course tower lean is not used I can measure the distance between the towers with the BA caliper and they are within +/1 .002" all around top to bottom.

(BA = big ass)

I am working on the new FSR mounts for the Rostock. I actually redesigned the kossel mount so it will now work on the kossel or the rostock. I attached the STL for the long FSR version here. I have another option for the short FSRs, I'll upload that if anyone needs it.

FSR Holder-V2.stl

(146.86 KiB) Downloaded 766 times

[Polygonhell]

Did you mention earlier you had ethernet enabled?
If it's a memory problem, you might want to try disabling Ethernet.
It takes a ton of RAM to enable Ethernet on the LPC17xx, the receive side of the hardware alone requires you have 4 buffers of MTU size, and if the library they are likely using supports anything other than the most simplistic interface, you end up with a buffer per connection that you copy the received data into before it's consumed. It can easily consume 25% of the total RAM just to enable it.

[mhackney]

Yes, I do have ethernet enabled. But, I don't have any other way to talk to the smoothie. On OS X, Smoothie USB support sucks.

[Nylocke]

Really? I was able to just plug in the board and it worked from what I remember. Hell, one of the developers uses OS X I think. They were complaining about problems with Windows support..

[mhackney]

This board is haunted. Maybe one of the to new ones I get tomorrow will be better.

With USB, OS X always auto mounts the SD card. I've searched for ways to disable this feature and can find none except a $99 piece of software. If the SD card is mounted, the machine crashes into the endstops when homed and all sorts of other bad motion happens. I have to remember to manually unmount the SD card before doing anything. It's a PIA!

[Nylocke]

http://smoothieware.org/configuration-options

search for "msd"

[mhackney]

It does not work on OS X.

[Nylocke]

Did you use the right binary? it only works with a certain build.

[mhackney]

Yes I did. But on top of that, I also need to build the binary - like 626Pilot's build and some code I am working on. I have not figured out how to disable it in the build system.

[Nylocke]

So you did use the right binary, or you have to figure out how to build a binary with your stuff based on Pilot's branch with the disable enabled? If its the former, thats really confusing, since I read on the website that its recommended that if you're using a Mac you disable the SD card with the msd_disable setting and build, as OS X likes to read the SD card occasionally, resulting in print stuttering.

Also, check it out!
<---

[mhackney]

1001! awesome, another Jeddi joins the ranks.

I am pretty sure I downloaded the correct binary that has this feature. What's confusing is why is there a different binary for this when it can be enabled via the config? I'll give it another shot tomorrow morning. I can ask wofmanjim what's required to build with this option.

[Nylocke]

I thought that was weird too. Probably the smoothie dev team trying to keep Smoothieware as tiny as possible, you've seen how diligent they can be with your temp switch module.

[mhackney]

Got it from wolfmanjm:

make DISABLEMSD=1

I asked why it is separate and apparently it is complicated. If it is not defined on windows bad things happen so they make it a separate binary. At least I know how to build it now.

[Nylocke]

Let me know if it works with Pilots branch, when (or if) I make the switch to smoothie that will be useful

[626Pilot]

You can disable MSD in the config.

https://github.com/626Pilot/Smoothiewar ... lta/config

Veeeeeery last line.

I do everything in Linux, where it doesn't auto-mount everything. For some reason, negotiation between the board and a host computer takes a good while! That will definitely hang the firmware until it's done, and if the printer is moving during that time, you're out of luck.

What I usually do is turn on the printer, connect in Rep Host, mount the SD card, wait 20 seconds for it to finish mounting. If I want to print from it, I save directly to it in KISSlicer, then wait however long it takes to slowly sync the data to the card. When the Smoothie lights start flashing rapidly again, I run sync, in case any metadata is left over in the cache. Then, it will print from the card without issue.

Another thing you can do is put the gcode on the SD card from an SD card reader. That's the "official" way to do it.

[mhackney]

I didn't compile your branch with make DISABLEMSD=1 though so that config won't/shouldn't work, right?

[Nylocke]

I tried that on both the master and Pilot's branch, it has to have that feature enabled in the binary for the config change to work.

[626Pilot]

BA caliper and a granite slab to assemble it as precisely as possible

I noticed that when I wash the borosilicate glass plate and put it back on, unless it goes back in exactly the same orientation as before, the calibration is thrown off a little. This tells me that the plate is really not all that flat, as I observe that the first layer has areas that will not adhere at a height of 0.1 because the nozzle is too far from the surface. I can compensate with the "bed roughness" feature in KISSlicer, and that helps the first layer, but it's still going to throw the calibration off. I have been thinking it would be better to get a machined piece of metal or something that was guaranteed to be way flatter, like maybe 5-10 microns or less, lay that on top of the glass so it "averages out" the surface, and then use that for the probing instead. Is granite the best super-flat surface you can get for a reasonable price?

[mhackney]

Granite is the standard for machinists and I think it is the best alternative due to low thermal expansion. It is not all that cheap for a quality block - there are standards for the flatness. I got a big 2'x3'x4" block from Enco a few years ago and it has paid for itself many times with assemblies and accurate measurements.

I've measured my borosilicate on the granite and it is flat to my measurement tolerances. That said, adding PEI changes that completely. I put an arrow on the glass sandwiched with the PEI so I can always point that at a mark I made near the Z tower. But even then, replacing the bed does sometimes need a tweak afterwards.

[mhackney]

I just verified that 626Pilot's HC firmware does not have the mad_disable functionality (or if it does, it doesn't work as I suspect from my previous experience with the Git release). I am going to do a make DISABLEMSD=1 now and test again.

[mhackney]

Did a 'make DISABLEMSD=1 clean all' on 626Pilot's branch, flashed, set the config to disable SD mount and reset. NOW I don't have the SD card mounting! So, I can disable ethernet, connect with USB and see if that frees enough memory to run the annealing.

[mhackney]

Ah yes, one step forward, fall on face...

So, I *thought* that I could use config-set to change the value of msd_disable so I could reset and have the SD mount / dismount at my command. That way I can mount the card to edit the config file, etc. But apparently you can't set msd_disable or network.enable with config-set. So I am now going to have to remove the SD card and read/write from my Mac as I have no other way to copy the config to the board. This is a major PIA.

[Nylocke]

Kinda makes you wish the smoothies had some sort of EEPROM or something to make things easier.

[626Pilot]

What would go into EEPROM, Smoothie saves in a file called config-override. It's basically the same thing.

[Nylocke]

I was going to say "but you could adjust the config file easier" until I remembered the smoothie commands adjust as much or more of the config settings compared to EEPROM on Repetier.