35mm film driven repstrap

Viktor,

I have played around with OpenCV, which has some nice IR algorithms. Haar (used for general object reco eg face), Optical-flow (Lucas-Kanard), Hough (used for basic shape eg line or circle), and Kalman-filter (tracking motion)
The implementation of these libraries are totally in C and is INTEL specific

It would be interesting to attempt Hough with Kalman in this problem and see how far I get.....especially with @BeagleFury's suggestion of encoding spatial information within the data-train, Kalman should work, IN THEORY.

I think encoding spatial information within data-train will work better than my initial idea of attempting to extract camera calibration information....from video....

At any rate, the best is to put the video through the sausage machine, and see what comes out the other side....

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

Edited 1 time(s). Last edit at 05/11/2010 10:58AM by BalanceSeeker.

Here's the start for a 2D pattern. It would support a grid of up to 45x45 cells, where '0'-'9' and 'A' characters replaced by '.' or '|' based on whether the 11 bit grid cell index has a corresponding 0 or 1 bit:

|...|01|..
|...|23|.|
|....45||.
|678|9A|||

Alternatively, just print a grid of distinct QR Codes

ok i don't think 1mm lines a cm apart will work, the belt moves pretty fast they just become a blur:

click link to see footage of a 20mm (actually 30mm) jog at 2000feedrate
[www.marklagana.com]



I'm thinking maybe coloured dots spaced much further apart perhaps?

UNrelated question:

because my gear is larger in diameter to the reprap/makerbots, my movements are 1.5 times the distance than the Jog controls in replicator.

how can i go about changing the software to compensate for this?

I'm using macosx at the moment, would it be easier if i used Windows XP and used the reprap software?

Edited 2 time(s). Last edit at 05/19/2010 02:04PM by mlagana.

I would not use a camera here but a photo-diode and an LED shining through the film at close proximity.
They don´t integrate over time as much as a camera.

---

-------------------------------------------
* homeprototype free 3d design repository
* Blog
* Google+

Yes, not sure where the camera idea came in, seems like it'd needlessly complicate things. We don't need a 2D encoding scheme with the tiny amount of information we're encoding.

All that's really needed here is the aforementioned photodiode and LED. The 35mm film is wide enough that multiple tracks could easily be done with multiple photodiodes/LEDs.

I think that what's needed is 2-tracks, one with absolute positioning and one with a relative positioning like Grey code. It would be nice to get by with just absolute positioning, but seeing as how that requires quite a few more bits, it would wind up being lower-resolution. Adding the Grey code or whatever would allow higher-resolution positioning while maintaining a periodic absolute position check from the other track.

Reading back, this is essentially what BeagleFury posted (though he recommends 2 tracks of Grey code). So give him the credit.

plasmator Wrote:
-------------------------------------------------------
> Reading back, this is essentially what BeagleFury
> posted (though he recommends 2 tracks of Grey
> code). So give him the credit.

Yeah. Three tracks makes it brain dead simple to write the encoder logic - Two tracks for the 2 bit grey code counter (encoding rate + direction), plus one track for periodic absolute position.

If you're using any of the G-code firmwares, the feedrate is controlled by a steps-per-mm definition in a header file somewhere. Adjust this to match your motor steps-per-rev and your pulley size.

It's better to calculate the steps-per-mm value from the motors steps-per-rev and the pulley size, rather than trying to measure and adjust it. Assuming that you know the spacing of the holes on either side of the film and the number of teeth on the pulley, use that instead of pulley diameter, as it'll be more accurate.

Edited 1 time(s). Last edit at 05/20/2010 02:40PM by jgilmore.

---

--
I'm building it with Baling Wire

jgilmore Wrote:

> It's better to calculate the steps-per-mm value
> from the motors steps-per-rev and the pulley size,
> rather than trying to measure and adjust it.

Could you explain why it is better?
It´s not obvious as a meassurement-aproach
is more reliable. It can e.g. pause the printer
if something goes wrong and it misses steps.

---

-------------------------------------------
* homeprototype free 3d design repository
* Blog
* Google+

thanks jgilmore/guys


i'm using my optostops as well but i might track down a photo-diode and an LED i think we might even use them at work

i was excited to try the webcam because i've been wanting to learn motion tracking (for programming not just visual effects) for ages... hard to get into from scratch when you have no programming experience. just found this so hopefully it will help:

[www.societyofrobots.com]

My thoughts on using encoder vs. using open loop control:

Open loop
+ Lower cost
+ Lower complexity
- Lower reliability
- Lower precision
- No error/shutdown capability to prevent damage

Closed loop
- Higher cost
- Higher complexity
+ Higher reliability
+ Higher accuracy
+ Error/shutdown capability to prevent damage

For a standard RepRap, I think open loop is the marginal winner. However, as devices print more and more of it's own parts, I think the cons of closed loop may lower to the point that the higher reliability and accuracy make it worth it. At the very least, having encoder modules allows greater numbers of options for the types of motion drivers that can be used.


> It's better to calculate the steps-per-mm value from the motors steps-per-rev and the pulley size, rather than trying to measure and adjust it. Assuming that you know the spacing of the holes on either side of the film and the number of teeth on the pulley, use that instead of pulley diameter, as it'll be more accurate.

I think you meant to say it will be more precise... precision is not the same as accuracy.. I.E, it may be able to have a larger number of positions to which you can ask it to go, but not more accurate: when it indicates it is at position X, the margin of error of the physical position on the output device will be greater. Each step between microcontroller to mechanical output can incur error, including loose wiring, skipped steps, imprecise centering of gears, loose belts, backlash, artificial forces on the axes, etc.


In terms of encoder technology, I created a device costing ~US$2.00 consisting of a single bright LED and 6 reflective photodetectors that should be able to reliably give 3 tracks of encoder feedback. However, looking at the capabilities of a $2.00 ADNS-2610 chip, I'm wondering if it would be easier as this effectively gives you a 1500 fps 18x18 pixel camera.

I mean that it's easier and quicker to get an accurate value for steps per mm. The value isn't affected by backlash etc. It'll be the same value regardless, your final position will be less accurate, but the number of steps to move 1 mm will be the same. Backlash can to some extent be compensated for in software. Compensating for skipping would of course require closed-loop control, which the reprap hardware doesn't have, and which the reprap software doesn't support.

You can calulate the value.

Or you can guess the value, set it to something, and then tell the reprap to move 100mm, measure the actual distance moved, and adjust your value based on that. Rinse and repeat a couple times and you'll probably get the same value you could have calculated to begin with. Probably.

---

--
I'm building it with Baling Wire

jgilmore Wrote:
-------------------------------------------------------
> I mean that it's easier and quicker to get an
> accurate value for steps per mm.

Ah.. yep.. I misunderstood your statement. You're simply proposing one of many ways in which one can measure the open loop calibration, and indicate tolerances on the film strip and pulley keys will likely be higher than a manual measurement for a set number of steps taken by the motor.

Makes sense now.

hey folks;

I'm trying to run a nema23 of a gen3 (makerbot) stepper driver,

everything seems to work fine but after moving the bed around for a bit the motherboard and steppers turn off.

i've tried the other stepper drivers and the same thing happens.

any ideas why?

here are its specs of the motor:

Size: Nema 23 (Size 57)
Phase: 2 Phase
Step Angle: 1.8 Degrees
Voltage: 70V Max
Current: 2.8A
Resistance: 1.2ohm
Inductance: 4.7mH
Holding Torque: 269 Oz-in
Dual Shaft
Bipolar 4 Wire

oh and i just realised the stepper is getting pretty hot, is it possible i just need a fan to cool it down which will keep it running?

it's even worse with both x and y plugged in, as soon as i try the nema23 it turns off.. then if i turn it back on it will turn off straight away.

@mlagana,
You may be hitting the driver chip's current/thermal limit. Most of the H-bridge chips with integral drive transistors monitor the temperature of the chip internally, and shut down before they fry themselves. Try adjusting the current limits (down), and see if that keeps everything cooler and happier. Given your motor's (torque) specs, you really should be able to run at a fraction (namely half -- quarter) of the rated current and still drive your mechanism just fine.

---

Larry Pfeffer,

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

ohh i'm such an idiot i completely forgot about the boards onboard pot's and they were both set to full!!!! luckily everything is still working fine and i am jogging to my hearts content thankyou so much larry!

Sooo i have a 2d printer!!! and my Z axis + my extruder are already nearing completion!

i think it's safe to say that film WILL work: (also the pen was slightly loose)









i think i will call it, the filmstrap!

PS whoever was asking if you can use film from a stills camera, you can it's the same.

-mark

PS any mac users out there who have got cad.py to work? I've installed numpy and scipy but i'm getting "ImportError: No module named Image"

Edited 2 time(s). Last edit at 07/08/2010 04:23AM by mlagana.

Very cool. Nice work!

How would you use still film, is it strong enough to just tape the ends together? I just generally don't know how film splicing is done.

you just make your join with sticky tape and make sure that the join is placed inbetween the two points where the bed is attatched to the film, where it won't have to go around the gear

i have already made an STL of a film-gear which is about the size of the mendel gear, got it printed out at shapeways and it seems to work although for some reason the hole in the center didn't print out so i need access to a lathe to make it and test it.

you can see it on my shapeways profile along with some regular film projector sized ones i made, with fuller bodies for better grip with the grub screw.

i'll put them on thingiverse but i kinda want to perfect them first with the hole plus maybe a trapped nut space for the grub screw and maybe increase the size of the teeth by .3 of a mm. or if anyone's interested i can put them up now

Edited 1 time(s). Last edit at 09/19/2010 02:14AM by mlagana.

now for configuration.... don't hold your breath!

Haha, that looks pretty badass. Hope it works for you!

Hey all.

i have calibrated my axes, now i am just unsure about my extruder...

I am using wades geared extruder but i have hooked up makerbots kysan DC motor to it. This is because i can't work out how to change the firmware to run a nema17, i would use reprap as host but i have also have a makerbot heated bed so not sure how easy that would be to run on reprap.

i'm tempted just to give it whirl on default settings and see how I go but i thought i'de put it to you people...

can't find anything in the machine.xml about it, so i assume i control it through the 'Motor Speed (PWM)' in replicators control panel.

what i can't find info on is how many revs per min a nema17 would typically do as it extrudes on wades or a pinchwheel. Or should i just experiment?

i tried extruding (without the heater on) and it's very slow. The DC motor i'm using is "2RPM" i guess that means max... so i guess it won't work because wades spreadsheet specifies 10RPM.

Might try the DC motor on the non geared pinchwheel.

my printer is finished and calibrated i'm just having software problems..

i use a mac, but i want to use wades extruder as pictured above. replicator on a mac is set to run a DC motor but i'm trying to use a stepper motor. i couldn't install the reprap software on windows XP (running on bootcamp on my macpro), so then i tried to install linux and it didn't work and also killed my windows hard drive, can't find my XP installer, so then i tried to use linux on my laptop (ubuntu) but i can't get it running i don't think i installed java properly or something..

so today i made a makeshift extruder out of a copper offcut to use with a DC motor but it wasn't powerful enough. i printed something though buy aiding the motor by forcing the filiment down with my hands.



now i'm trying to get my hands on a PC because i have wades geared extruder and want to use it.. unless anyone can help me out with either:

-the GCODE for running a nema17 for extrusion (so i can use replicator)
-help on running reprap on ubuntu

Edited 2 time(s). Last edit at 07/07/2010 12:37PM by mlagana.

A common problem under Ubuntu with anything java is the default version of Java that comes in Ubuntu doesn't quite cut it in some cases.

You probably need to:

Install the package sun-java6-jre or mebbe sun-java6-jdk if you want to develop stuff. This is in the partner repository, which you will need to enable as it is not one of the defaults.
Run 'sudo update-alternatives --config java' and choose the number that corresponds with "/usr/lib/jvm/java-6-sun/jre/bin/java".

This second part is vital, as it sets up which version of java to use by default.

PS: I've not run the reprap software in this environment yet, but I've used a lot of java stuff. I got bitten by this recently after an upgrade removed the Sun version and installed the default one on me. Things looked like they worked, but stuff like file opens and the like just failed silently.

Cefiar: Can we get a few details, so that this strange "Java on Ubuntu" issue can be duplicated and investigated, please?

I'm running the RepRap host software on Ubuntu 10.04 Desktop amd64 under openjdk, and it can open files just fine. I created my own wrapper script to use the (64bit) Ubuntu java libraries, not the 32bit ones that come included with the reprap-mendel-*.zip file, but that is the only change I had to make to get this working.

What exact version of Ubuntu are you running, what exact version of openjdk, and do you have a small test java program that demonstrates the failure to open files, please? Also, if you reported the issue to Ubuntu already, what LaunchPad bug number has been assigned to this issue?

---

---

Jonathan

ahh thanks for the help cefiar and jonothan but i found my XP installer yesterday and have finally got reprap running... couldn't get there with ubuntu.

OK i have reprap software running on windows XP service pack2 (on bootcamp on macpro) and it doesn't seem to be connected to my repstrap. i've installed the FTDI driver and my machine is working fine under the windows version of replicator. I've changed the reprap settings to my port (com3) and changed CommsDebug to true.

The only other glitch i'm having is that when i upload the firmware (on windows arduino), the extruder firmware uploads fine, the FiveD_GCode_Interpreter also uploads but with this message:

avrdude: stk500_getsync(): not in sync: resp=0x00
avrdude: stk500_disable(): protocol error, expect=0x14, resp=0x51

also i get this in repraps console:

comms: G-code: N0 T0 *26 dequeued and sent [0.000s/-1279213586624ms]
comms: G-code: N0 G1 F1500.0 *116 dequeued and sent [13.516s/13516ms]


stuck