"Quick" calibration idea

Re: "Quick" calibration idea

Postby Niggle » Tue Feb 18, 2014 1:01 am

I've been thinking about this some more.

I suspect that we have at least two things to calibrate.

Checking that the bed is level is one thing. And it makes me ask 'Level with respect to what?'

Checking the length of a step in an arm is another. It really needs to be done for each arm independently. I was toying with the idea checking all of the arms at the same time, but I couldn't think of a way to measure the height of the effector after each step.

The best idea (at least it is my best, which isn't saying much) is to detach an arm from the effector and place it on a jig fitted with a microswitch. Run a firmware routine to count the steps between equidistant marks on the jig and save the results. Repeat for each arm in turn and then we should be able to correct some of the positioning errors.

Note that the processor only has one measurable reference point, the motor position when the end stop triggers. If that is not repeatable or is changed, the calibration is shot and would need to be redone.

Equally, assuming that the spool is the main source of errors, anything that caused it to shift on it's motor shaft would also ruin the calibration.

That's it for now.
Niggle
 
Posts: 84
Joined: Tue Feb 11, 2014 7:27 pm

Re: "Quick" calibration idea

Postby NeoTheFox » Wed Feb 26, 2014 7:21 am

I've got an idea based on this thread - what if we use linetracer for callibration? All it needs is a probe, simular to bed autolevel, and marks on basalt. It should seek for the marks (you can print the marks on a regular printer) and remember the points they are located in. You can harvest the sensor from an old mouse. Some instructions
The callibration table in my mind should look like lines of marks coming from shoulders attachments. Each hand scans own line a few times, making itself a table of errors, and after this the whole system centeres at the crossing point, probing z axis.
NeoTheFox
 
Posts: 92
Joined: Tue Jan 28, 2014 5:54 pm
Location: Moscow

Re: "Quick" calibration idea

Postby Nicholas Seward » Wed Feb 26, 2014 7:59 am

@NeoTheFox: I have looked at this before. I was initially very excited but I feel the need to share the drawbacks.

1) The resolutions isn't quite good enough. It would be better if we had similar optics on a webcam.
2) The mouse sensors are not an open platform and are not conducive to a final solution.
3) Not all mouses can do this.
4) Integration into a control scheme would be hard with current firmware and controllers.

None of this is blocking. I personally would like to do this with special optics on a webcam that has a razor thin depth of field. This will allow for x, y, and z probing. Again, none of this is blocking but would require substantial effort but could provide a cheap solution in the future.
Nicholas Seward
 
Posts: 738
Joined: Mon Nov 25, 2013 10:41 pm

Re: "Quick" calibration idea

Postby NeoTheFox » Wed Feb 26, 2014 9:34 am

Nicholas Seward wrote:@NeoTheFox: I have looked at this before. I was initially very excited but I feel the need to share the drawbacks.

1) The resolutions isn't quite good enough. It would be better if we had similar optics on a webcam.
2) The mouse sensors are not an open platform and are not conducive to a final solution.
3) Not all mouses can do this.
4) Integration into a control scheme would be hard with current firmware and controllers.

None of this is blocking. I personally would like to do this with special optics on a webcam that has a razor thin depth of field. This will allow for x, y, and z probing. Again, none of this is blocking but would require substantial effort but could provide a cheap solution in the future.


I can agree with this, however I am not shure the resolution will not be enouth.
We can make the task even simplier with sensors like this - they are open, and with tcrt5000 you can detect 0.2mm lines, and we can use more than one. All we need is precise periodic perforation on the line. It is pretty much like encoder for DC motors, exept not round, but linear.
This can be easily hooked up to arduino or any other board. And since we are not moving kinematics to firmware just yet it can be pretty easy to make a firmware just to scan theese, leaving all the processing to PC. Values can be stored by segmentize. I think adding callibration like this to existing firmware is simple, also it can be calibration-only firmware.

I can try to do this, but I am not really good at math. I do have some encoders, however, but I am not sure about the resolution.

EDIT: it is even possible to use color-code with something like this.
NeoTheFox
 
Posts: 92
Joined: Tue Jan 28, 2014 5:54 pm
Location: Moscow

Re: "Quick" calibration idea

Postby tbcarlson » Mon Mar 17, 2014 5:36 pm

I'm Tim. I've been lurking here for a few weeks, and my 15 year old son and I have been bitten by the GUS bug! I've printed a few motor and slave arms, and have most of the hub printed. My son is going to take on the assembly challenge, and is starting to solder a RAMPS board already!

Regarding calibration, I had an idea this morning that I haven't seen mentioned here, or on the RepRapWiki.

What if we were to attach a small, cheap "keychain" tape measure to the bottom of each motor arm, and attach the end of the tape to the far end of each slave arm. Both attachements would have to allow for rotation in the same axis as the arm bearings. We would then have 3 tape measures that would show the relative length of each arm. Now use the guts of 3 optical mice to measure the movement of each tape. The pre-printed patterns on the tape may give enough contrast for the mice to work, or a speckle pattern could be painted on the surface.

The tapes could be attached permanently, or just for a calibration run. If attached permanently, I don't think they would interfere with any printed object, as they would be quite a bit higher than the print head. If the position of the 3 mice could be fed to a computer, we could automate the calibration. The nice part of this scheme is that we could calibrate throughout the whole build volume, not just at the bed level. With the right electronics, this could eventually allow a closed loop system.

With a 1600 DPI mouse, the resolution would be 16 microns on each arm. I know at least some mice claim to be able to track several m/s, so the speeds at which we are operating should be no problem. But what about repeatability? If a mouse "misses steps" in normal use, it probably isn't noticed. But in this application, missed steps would make this idea useless. Does anyone know how repeatable optical or laser mice are?

Thoughts?
tbcarlson
 
Posts: 6
Joined: Fri Mar 14, 2014 6:46 am
Location: Colorado

Re: "Quick" calibration idea

Postby MrFaul » Mon Mar 17, 2014 9:37 pm

Hm the basic idea is really good and I like it. But my personal experience with mice on different surfaces leads me to the conclusion that they are skipping quit often at least the cheap one. And I don't like the idea to disassemble gaming hardware which for fills the necessary precision.
Hm thinking of gaming hardware mounting a Sixsense STEM could be fun but its tolerance is in mm so not really usefull.
MrFaul
 
Posts: 31
Joined: Fri Feb 14, 2014 8:17 am
Location: DE Wuppertal

Re: "Quick" calibration idea

Postby tbcarlson » Tue Mar 18, 2014 1:18 am

MrFaul wrote:Hm the basic idea is really good and I like it. But my personal experience with mice on different surfaces leads me to the conclusion that they are skipping quit often at least the cheap one. And I don't like the idea to disassemble gaming hardware which for fills the necessary precision.

I think 1600 DPI mice can be found that are very inexpensive (this $6 one on Ebay, for example). I would also rather order parts and build from scratch, but don't think you could get the parts and build a board for less than just hacking a cheap mouse.

If this were to be used just for calibration (not a closed loop), a little bit of skipping may not be fatal. Since the computer would be in charge of data collection, it could easily average multiple extensions to reduce the effect of random skips. And calibration could be done slowly to reduce likelihood of skipping.

An alternative to the optical mouse might be a sensor like this one at DigiKey, which actually gives absolute position, and is less than $10. The trade off is that now you've got to turn the linear motion of the tape (or string, etc) into rotational motion, which introduces more potential for error. If you try to use the tape measure's internal spool to rotate the required magnet, the effective radius is changing as the tape is pulled out (which could be accounted for), but the spool may not have a consistent radius, may wobble, etc. We would have to calibrate the calibrator! :)

I wish I had time to hack an old mouse and see how repeatably it could measure an extending tape. But building GUS will have to come first!
tbcarlson
 
Posts: 6
Joined: Fri Mar 14, 2014 6:46 am
Location: Colorado

Previous

Return to GUS Simpson

Who is online

Users browsing this forum: No registered users and 1 guest

cron