[turmite]

it is easy to contribute... actually do something!

vacpress, the only thing I can contribute is to encourage those that have been involved in this. Trust me, if I had the ability to give any kind of meaningful help on this it would have been done already.

Rockabye74, GrahamS mentioned the need for high resolution encoders. Could those also be made using some sort of optical sensing device at four different locations using four different rings of indexing marks for higher resolution?

Mike

[anoel]

What the Hell happened to the Digitizing arm discussion? It just died out... Did any of you continue to work on this thing?

Thought I'd just give this thread a bump as it was a pretty hot topic at one time.

[Graham S]

some basic software was done at the time and is ready for testing.

basically if you really want to do this you can, you just need to get the right encoders.

For even small arms you need very high resolution and this is not something you can hack. I have shown a backlash free gearing method but I have not tested it.

I personally don't want an arm as much as the cash for the encoders.

Graham

[HenryCNC]

Sorry about this being off-topic (a bit) but have any of you programming/math folks looked at making a vision based system? There are a few systems that triangulate a XYZ coordinate for each pixel in an image. The mechanics are very simple - all the work is in the software.

For DIY this is a good combination because the cost to build is less. There is no need for precision stuff. But (big but) somebody has to program the software.

Just my two cents, sorry for the tangent.

[MaxLogan]

HI all, my first post on this forum.......

Fantactic thread, I have been working on a digitizer for some time now. Much the same as you guys have here, came across the same problems and hurdles. I purchased 6 16bit optical absolute encoders (only way to do it with accuracy 65,536 ppr, also best to have absolute encoders for referencing each joint). They are not cheap, and to have an arm of any real world use you realy need 6 axis articluation, especialy if you intend to use it for digitizing car bodies as I do. The problem is callbration and error compensation, require a realy expensive bridge CMM to reduce errors in the cartisisan algorithims. Also need to compensate for temperature/expansion of the joints and acceleration of each joint during use. And the cummaltive error requires correct handling during different arm articulations.
So having spent so much time and money on the project I ended up purchasing a FARO 4' Arm, and coming up with efficient leap frogging techniques.

Im still developing the the arm thats for sure, great fun. Will it work? Hope so. LOL

I hope you guys keep going, it will be intersting to see where we end up. Analogue v's Digital.

[Spaceraver.DK]

BUMP..

Just wondering if the idea ever got realized.. some1 needs to post pics of theirs.. just curious

O/T.. Could it be possible to make a 3 way setup with a 3d mouse, natively supported you have X and Y from the mouseball right.. maybe using the wheel as Z?? im doing this as a low end idea... not very accurate, but if i only need 1/10 mm precision that would be enough??

[Graham S]

No, the high resolution encoders REQUIRED (even for 0.1mm accuracy) are too expensive for most to bother with.

A mouse is not suitable at all. You can prove this to yourself by drawing a few arms and moving one joint by the resolution you expect. You will be disappointed!

Graham

[turmite]

Hi Graham,

I was surfing today and found these. Are these good enough for resolution. They have some 10X this accurate but are $995.

Mike

[Graham S]

someone forgot the link

[High Seas]

Its a trick!! Keep us here waiting and Mike gets on another computer and buys them all up!

[turmite]

:idea: Hey! I have an idea, I'll add something to the post! (chair) I'll have to go back and get the link......sorry.

Mike

Low Cost - $12.50
Miniature Sin/Cos Encoder for high resolution
Sharp GP-IR14 Rotary Encoder,
192 Cycle,768 Quadrature Count (photo) $12.50
Outputs:
° SIN-(192 cycles/rev.) = 1 Vpp,
° COS- (192 cycles/rev.) = 1 Vpp,
° Index- (1/ pulse/rev. ) = TTL pulse
Resolution = 768 zero crossings, - with 20X interpolation - to 3,480 counts/rev.
Power: +/- 12VDC
Shaft: 4mm shaft, may be drilled out to 7 mm.
Quantities: 10,000 available from stock.
PRICES: $12.50/ea. 1000's in stock - Design this one in!

Please note: 192 cycles Sin/Cos = 768 zero crossings in quadrature.
Note - current interpolator circuits multiply resolutions from 5X zero crossings (3840 counts) to 256X (196,608counts). These resolutions related to a 5 pitch ballscrew translate to resolutions measured in millionths of an inch.

[turmite]

Here tis the link!

http://www.lasermotion.com/Servo.html

Mike

[Lston]

I have been in contact with Graham S and he sugessted I post this information here because he is too busy with work to continue with the project.

I fave found a UK company that sells a low cost 16 bit Absolute optical encoder.They do require a allignment tool and I don't know their price.
Graham S. says Farnell sells them for 116.86GBP plus VAT ea. but in quanities of 15 they are $51.34

http://www.ebv.com/en/websearch_resu...EAS-7000-1GSD0

If 3 people were to buy together that would be 5ea, just what is needed for ea digitizer and that would be only $256 per person.

Also I read in the forums that a optical mouse was considered for an encoder but it was concluded that they are not accurate enough.

Logitech says their new G5 laser mouse has 20X the resolution of older optical mice and up to 2000 dpi (800 CPI) (counts per inch)

It would serve as a incremental encoder so you would still need to zero it out to a "home position" when powered up, unlike a absolute encoder.

Mount a laser mouse sensor inside a joint that rotates around it as its pointed at the sidewall of that joint.
Then you could calculate the distance into degrees based on the circumference of the joint

EG: 3" I.D of joint X PI = approx. 9.42477" circumference
The specs say 800CPI (counts per inch)
So 1"/800CPI = 0.00125" for ea count?

Would one of them work?

The sensors are made by Agilent (specs on Page 14)
http://cp.literature.agilent.com/lit...989-3076EN.pdf

[unterhaus]

laser mouse isn't going to work because it is basically a camera taking pictures of your desk and figuring out how much you moved it. The $50 option sounds like a good one, otherwise go with ebay. Don't really see why you need an absolute encoder because the points are all relative to each other, but I could easily convince myself otherwise seeing as how the kinematics are so complex.

[tekno]

I havent seen much on ebay when it comes to encoders.

[Lston]

---------------------

[unterhaus]

The cursor moves on your computer screen when the mouse is moved reguardless of how.

that's a very good point, Sorry for my lack of imagination. I didn't think about putting a very small desk inside each joint. Would be interesting to see how small you could make the mouse guts. HP or Agilent started the whole optical mice revolution, don't know if they are still the only source of chips/sensors.

[vacpress]

zso. there was alot of interest and almosts in this thread... i wonder if it makes sense for the same crew of interested, but encoder-cash strapped inventors to switch to laser scanning? it seems harder at first, but there are lots of examples of working projects out there... i think i saw one that was made of.. like tape and cardboard and toilet paper tubes and stuff, but actually worked to some degree... you guys could do much better!

[Graham S]

I'm suprised there hasn't been more interest, we are talking a 16bit absolute encoder for hobby prices (at the 15 off price). That's over 65,000 points per revolution and not only that all the software has to do is ask the encoders where they are, no monitoring and no quadrature conversion and no noise issues. This is a proverbial piece of cake for a PIC, Rhino might even be kind enough to tell you what you need to send it for direct entry.

If you want to build a digitizing arm it will now come down to how well you can make the joints and how stable you can make it thermally.

This is asuming the encoder does what it says on the tin!

Graham

[turmite]

Hi Graham,

As for me, I have a probe and software coming that will do 20,000 points per hr. I also know that Art is planning something similar for the Mach3/G100 plugin system.

Mike