35mm film driven repstrap

MarcusWolschon Wrote:
-------------------------------------------------------
> Any idea where one could obtain some scrap-film?
> I´d like to redesign a MiniMendel to use it and
> try this out.
>
> What do you use to glue it?


Buy a cheap disposable camera, they're $3-4 now

I was curious if the old 8mm cine film could handle the torque of the system?

not sure if stills 35mm film is the same as live action film - the picture is definitely not orientated the same way but i can't remember if the holes are the same... i'll check later and i have some super8 film i'll have a look at it... would be cool because it's so small but much less available

mlagana2 Wrote:
-------------------------------------------------------
> not sure if stills 35mm film is the same as live
> action film - the picture is definitely not
> orientated the same way but i can't remember if
> the holes are the same... i'll check later and i
> have some super8 film i'll have a look at it...
> would be cool because it's so small but much less
> available


There's loads of it on ebay

but it ain't free

True but its cheaper than buying belts

This is a really neat idea. It would probably be pretty straightforward to build a film exposure device that could expose a Grey code onto the film for use as optical encoder (as others have suggested). I could see using a low-power laser and geared stepper, scanning the laser across the film with a mirror. Could also print the code onto a piece of transparency and make a continuous loop, and run that on top of the film with a lamp behind it to expose the film.

With the low expense of B&W film processing chemicals and equipment, this could be a very accessible way to add some precision to the RepRap and other CNC devices.

plasmator Wrote:
-------------------------------------------------------
> This is a really neat idea. It would probably be
> pretty straightforward to build a film exposure
> device that could expose a Grey code onto the film
> for use as optical encoder (as others have
> suggested).

Note: Greycode is nice, but using 3 encoder tracks gives ability to track absolute motion in addition to relative motion. Two tracks are a 4 bit greycode counter. The third track encodes a sync bounded absolute position indicator. This allows an absolute position indication signal at frequent intervals, allowing the device to home itself while under motion. No more need to home back to the axis end-stop mid-print or between layers.

Such a device could be used to print on film this schema too.. Who's gonna volunteer to create such a device?

Greetings all,

I've been thinking of a variation on this overall idea, that may be lead to better closed-loop positioning for a variety of repraps, whether they use a film-strip as a drive element, or not.

Some background:
Before going over to primarily digital recording, the (professional) film industry used "magnetic film" to record sound. This was film stock (16 or 35 mm, with the holes for sprockets.) Having the same form factor, running at the same speed, was helpful for keeping sound synched up with the picture. (There were also variations, which put magnetic coatings on the edges of photographic film. The completely-coated magnetic film can still be bought, and I think a couple companies still make it (albeit in much longer reels than I need.)
Here are some links regarding the 16 mm version of magnetic recording film:

[www.motion.kodak.com]

[www.actioncamerasf.com]



[www.austriamicrosystems.com] makes magnetic encoder chips (both for linear and rotary applications), that use interpolation to get fine resolution from a fairly coarse (spatial) frequency on the "magnetic ruler." So far, they only sell "magnetic ruler" in frustratingly short lengths (100 mm), and the way they make it does not lend itself to longer continuous lengths. (I've inquired and persisted. They say they're working on it....)
For accurate measurements via these chips, the waveform must be sinusoidal, and or the correct frequency. (The chips can tolerate both a linear bias and a range of amplitudes, so long as the entire tape has the same bias and amplitude.)

However, it may be possible to make a vertical recording rig to put the right magnetic sinewave into such magnetic film -- or into the flexible magnetic material used for large, flat "refrigerator magnets." The latter material *is* available in long strips. I've already purchased some from
www.magnet4less.com It came pre-magnetized, but not with the right spatial frequency.

Such a magnetic linear encoder could be applied to a variety of repraps/repStraps. Possibly to machine tools in general, however the magnetism would probably attract chips/crud, requiring wipers, etc. to keep those from ruining measurements and/or the magnetic ruler.

I saw a recorder/duplicator for such magfilm locally, but for the price, I think I'll pass and perhaps try making a vertical (perpendicular to the tape) recording rig. For this application, it can be *slow*, so long as it is *very* steady.

I'd be interested in any comments/thoughts/opinions on this approach,

---

Larry Pfeffer,

My blog about building repstrap Cerberus:
[repstrap-cerberus.blogspot.com]

out of all the possibly sensors/encoders, LED, laser, camera, magnetic, (probably much more) maybe we should just choose the cheapest and work from there?

@mlagana,

Rotary encoders (whether optical or magnetic) measure rotation of an intermediate shaft, not the X,Y,Z motion; as such, they don't measure backlash. Of the others, there's been much talk, but (to my knowledge) no clear, "this one works to this resolution, and it costs this much."
So, I see no great harm in experimenting with magnetics.

---

Larry Pfeffer,

My blog about building repstrap Cerberus:
[repstrap-cerberus.blogspot.com]

cool i'm behind you, it's different but we use magnetic encoders that read a bit of metallic tape on the film at work and they never fail. Although in this scenario the 'magnetic ruler' means the film is actually magnetic and the chip not? email me if you want some film/help/want to split costs of ordering things for experimenting.

also it would be good if the chip could plug into the existing stepper drivers.

as discussed previously it would be good if we could automate the process ourselves... maybe using the CNC as a plotter we could make the lines across the film with a silver pen... the ink could possibly be reflective enough for a lazer pointer?

someones bound to make a simple breakthrough here

PS i'm so glad you're all into the film as belt thing... in my real life situation everyone has been trying to talk me out of it from day one saying just use timing belts that's what they are made for... they don't get it! (but truth is i work so much harder when people tell me i am unable to do something or that it won't work.)

... optical mices (commonly a camera with 19x19 pixels) have some thousand DPI, with a webcam and reversed objective you can resolve pixel-positions until 2 microns, whats nearly 13000 DPI ...

With two crosshatch-films i printed with a laser-printer with 50-microns "...-line-spacing-line-...", moving against each other and resolving Moirée-patterns with a 'naked' (without any optics) camera-sensor i've got resoultions down to 0.2 microns - but you need an image-recognition software ...

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

VDX Wrote:
-------------------------------------------------------
> ... optical mices (commonly a camera with 19x19
> pixels) have some thousand DPI, with a webcam and
> reversed objective you can resolve pixel-positions
> until 2 microns, whats nearly 13000 DPI ...
>
> With two crosshatch-films i printed with a
> laser-printer with 50-microns
> "...-line-spacing-line-...", moving against each
> other and resolving Moirée-patterns with a
> 'naked' (without any optics) camera-sensor i've
> got resoultions down to 0.2 microns - but you need
> an image-recognition software ...

With optical mice, the DPI is relative precision, not absolute, if I recall correctly. Any idea on how much repeatable absolute positioning they can guarantee? -- how much drift do they have? What techniques exist to limit or eliminate it?

Several people have suggested using optical mice as an easy to integrate feedback mechanism.

... in a German robotic-forum around the ct-bot we developed some tools for transferring the mice-sensor camera-output to a PC, so it could be handled like a normal video-stream.

Essentielly this was a camera with 19x19 pixels and the FOV of the optical footprint.

With image-recognition running through the videostream you can identify lines and shapes, so with a sort of optical coding/encoding with parallel segments of lines you can detect the absolute position of the sensor too.

There are some other camera based OS-solutions with image recognition for robotics, so there could be some more relating stuff ...

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

here is a good PDF that shows how to hack a mouse to use it as a sensor including some programming but for the Athena chip/environment

[www.kronosrobotics.com]

heres a good mouse hack for an arduino

[www.bidouille.org]

ok this is the best tute for using a mouse with arduino, he actually achieves coordinates where as above was a hack for using it as a camera

[www.martijnthe.nl]

... we had a special PCB for the mice-sensor - here the images with "Maus-Tuning..."

And here in this post (in German) some images of the input ...

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

ok i had a night off work and an old mouse here so:










i printed a barcode type thing and taped it onto the film. The serial monitor in arduino picks up it's coordinates except the x axis didn't seem to go up at quite the same rate, (when just using it as a mouse for instance) the y goes up into the thousands and the x the tens,

also sometimes the coorddinates would just increase even though the bed was staying still, i think because i didn't align it really at all i just blue tacked it down

victor those PCBs are genius

... sometimes it's really tricky getting perfect visibility - it's a simple camera with all pros and flaws like FOV, defocussing, illumination and such.

So we used the mice-sensor only for the relative translation-measurement and for exact positioning we used triangulation sensors against walls and aditional triangulation width beacons in bigger mazes.

A cheap web-cam with 720x562 pixels (or even 320x200pixels) is much more capable for complex absolute-positioning with coded 2D-patterns ...

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

mlagana2 Wrote:
-------------------------------------------------------
> victor those PCBs are genius

... it's a kit with the mice-sensor and two reflex-sensors for detecting line-borders - AFAIK the price was something around 19 Euros ... but not problem to make an own PCB.

I'll check if the sensors are still available - i should have two: one in the bot, another spare somewhere in the boxes ...

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

i have a usb webcam here.

so it would be usb camera footage going through motion detection algorithms (in what environment i'm not sure), coordinates sent to arduino...

i wonder if i would be better off restarting in windows XP to get a hold of motion tracking software. seem to remember nothing for macosx

or could this all be done in processing or something?

today i was able to mould and cast extruder parts which i guess i will put on ebay.

@mlagana2,

if you need help with motion tracking through, I can offer some assistance.
if you want, you can send me sample video footage, and I can see what can be done.


I am tracking this thread, or you can email me:
moveaftermove _at_ yahoo _dot_ com

Marius Botha
Pretoria, South Africa
[mariushermanbotha.wordpress.com]

... i made some image-recognition and motion-tracking with the free DIAS - maybe it's of interest for other IR-systems too ...

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

ok i'll give it a go with my webcam, what would you guys suggest i print and stick onto the belt? In that test i just went into microsoft word and typed a whole bunch of IIIIIIIIIII's...

maybe a 1mm line 1cm apart?

mlagana2 Wrote:
-------------------------------------------------------
> ok i'll give it a go with my webcam, what would
> you guys suggest i print and stick onto the belt?
> In that test i just went into microsoft word and
> typed a whole bunch of IIIIIIIIIII's...
>
> maybe a 1mm line 1cm apart?

An algorithmic pattern would be best, in my opinion... because it gives you both motion as well as homing information.

Something like:

|...|..|..|...|..|.||...|..||.|...|..||||...|.||..|...|.||.||...|.|||.|...|.|||||...||.|..|...||.|.||...||.||.|...||.||||...||||..|...||||.||...|||||.|...||||||

This would allow both motion as well as position to be detected. (If you look, the above pattern is just "|...|01|23" repeated over and over, with 0123 replaced with a binary representation of a number between 0 and 15.

Similar patterns in 2D could be used to detect position, orientation, and change in position and orientation.

[Updated] Also, I assume that you could, worst case, use a FPGA chip to output data to a microcontroller if you had to include it on an embedded system. Using the host processor would probably work as long as it wasn't needed for fine detail closed loop control.

Edited 1 time(s). Last edit at 05/11/2010 10:46AM by BeagleFury.

@Viktor,

DIAS looks very interesting.... It seems to be a high-level scripting language.
The literature references are mostly in german, and the only english language reference seems to be broken
I downloaded, and installed the DIAS application, but this is completely in german, and the help-file is no longer supported on my windows-vista laptop
Some of the DIAS applications seems very interesting to me, particularly the 3d reconstruction (moire).....
As you can probably tell, I have never played with DIAS before, which seems very powerful to create image-processing software, quickly.

@mlagana2,
1mm vertical, spaced at 1cm intervals should be fine......

Regards
Marius Botha
Pretoria, South Africa
[mariushermanbotha.wordpress.com]

Hi Marius,

... DIAS is a week-job to learn, but then you have a really sophisticated basis for IR - i was busy then with closed-loop nanopositioning with some crosshatched patterns (50microns lines, same in gaps) with Moirée-displacement 900:1 ... here a film-translation of 100 nanometers resulted in a pattern-translation of 90 microns, what was pretty easy to detect/measure with the camera-sensor

If you can't find another free IR-IDE in English, i'll try to find some time for reactivating my DIAS-brain and testing with pattern-recognition and -displacement ...

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]