CHEAP or EASY String Pots (source, or fabbed)

[Qbranch]

Use a retractable keyring holder and trim pot. Just super glue the trim pot's screw to the shaft of the keyring holder. Gives pretty good linearity even though the string winds up on the drum, but if that's a problem, you can make a lookup correction table or a correction function for it. Without correcting, yields about 1/4" or better resolution over about five feet.

This setup was used to provide feedback for all of this robot's actuators. I'm friends with the person who built the thing... so... pretty good bet this should work for you too.

-q

[IKE]

I have also often wanted to make one of these. You mention needing 25" of travel. What level of precision and accuracy are you expecting, wanting? If you are looking for something that is accurate within 0.5", that can allow for one level of construction. If you need something within 0.05" or tighter, your best bet may be to stick with a preofessionally made string pot.

Another key techincal point is response time. for instance, a tape measure takes a couple of second to wind up, where as a key/badge retractor may be significantly faster. The pro grade string pots I have worked with in general aren't very reliable past 20 Hz, or 0.05 seconds as you start to excite string modes.

Beside the key or badge retractor, you may want to look into a Yo-Yo ball.

[kmcclary]

Originally Posted by IKE

You mention needing 25" of travel. What level of precision and accuracy are you expecting, wanting? {also, what response time?} Beside the key or badge retractor, you may want to look into a Yo-Yo ball.

Not necessarily professional quality. I'm assuming in the 0.1"-0.5" resolution, not 0.05" or better. I'd prefer something in the 4-8 steps/inch (0.125"-0.25"/step) ballpark. Whatever is SIMPLY achievable.

Linearity is not needed, but REPEATABILITY is vital. Out and back to the same point should yield VERY close to the same value, whatever it may be. This allows for calibration tables to deal with the linearity challenge.

Response time? Fairly slow. The application is pneumatic cylinder extension monitoring, so as long as the data keeps up with the cylinder motion we're fine. Since we're limited to 24" strokes, a 25" max cord is reasonable.

Originally Posted by Qbranch

Use a retractable keyring holder and trim pot. Just super glue the trim pot's screw to the shaft of the keyring holder. {...}

We tried that at one point, but had both fab and repeatability challenges with it, because the reel used allowed for overwrapping.

I do like the PVC threaded end suggestion! I need to think on (or tinker with) that one.

Any other ideas, or working models? BTW... IF there's anyone that has some images of a working model, I'd LOVE to see them posted in the CD-Media area!

Thanks!

- Keith

[JesseK]

Last year our team brought up using an extending lanyard ($3) with a multi-turn pot ($3) for this purpose on our elevator. We tried it with a lanyard and with a $2 tape measure.

The main issue with the lanyard is that the internal rotary 'spring' doesn't have enough force to turn the potentiometer when the string is almost all the way retracted back in. Repeated wear only exacerbates the problem. With a tape measure, no one axle 'rotates'...the entire spring and tape measure rotates as you pull it out. Hence the reason tape measures are always smooth and slippery against metal.

[IKE]

Can you attach a chain to a structural member and then use a sprocket as a cheap rack and pinion system? If this doesn't make sense, I can try to re-explain it.

[kmcclary]

Originally Posted by IKE

Can you attach a chain to a structural member and then use a sprocket as a cheap rack and pinion system? If this doesn't make sense, I can try to re-explain it.

Yea, we thought of that, but now you're likely to require an additional drawer slide to assist alignment (for the chain with a fixed pot, or for the pot with a fixed chain), which is makes the total device heavy when compared to a cable pot or string pot.

We also looked at beaded chains wrapped around a drum with a spiral track groove and dimples. THAT looked promising, but it's still a complex fab for the drum. (Hmmm.. a weird thought. A wrapped ROLLER CHAIN on a "helical sprocket" that was clamped onto the pot shaft??? <shudder> Whoof... THAT would not only be heavy, it'd be a horrible thing to try to fab! <chuckle>)

Now a non-FIRST friend offered the pretty cool idea of coupling a cheap motor's shaft onto the pot end from a facing motor, and driving it with a tiny current limited supply. This makes it a "virtual spring". (In essence, stall the motor, but with a small enough current that it won't fry...) Now THAT was interesting!

A few problems with using that one in a FIRST contest: A) We're severely motor limited in this contest. We often only get ONE raw motor and even if we gut other assemblies we're losing useful motors for a trivial application that could probably be better done another way with some thought. B) The "motor-spring" -must- be driven by a Spike, Victor, or Jaguar without stuff between the motor and driver. If we drive it with one of the speed controllers at the lowest setting, it might not be so bad for ONE widget, EXECPT for -> C) The real killer with this method is the fact that all of the above drivers are disabled except when the round is active, creating a HUGE chance of it unwinding, and tangling. (DARN...)

So... back to researching a true "spring based method"...

More design thoughts? Comments? Experience?

- Keith

[EricVanWyk]

Originally Posted by kmcclary

Now a non-FIRST friend offered the pretty cool idea of coupling a cheap motor's shaft onto the pot end from a facing motor, and driving it with a tiny current limited supply. This makes it a "virtual spring". (In essence, stall the motor, but with a small enough current that it won't fry...) Now THAT was interesting!

Very cool idea. Current mode control (rather than voltage mode) has a lot of neat applications.

[kmcclary]

Now that we're all back, I'm reviving this thread from last December. We are still looking for a way to get our hands on (or make) some cheap string pots!

Originally Posted by JesseK

Last year our team brought up using an extending lanyard ($3) with a multi-turn pot ($3) for this purpose on our elevator. We tried it with a lanyard and with a $2 tape measure.

Interesting.

The spring system doesn't seem to be the biggest problem here. IF the pot has low enough friction, it looks as if keychain retrievers or autoclosing tape measures should be sufficient to keep it under tension.

The big problem seems to be fabbing a lanyard and drum rig that prevents string overlap. If an overlap occurs, you've completely lost data integrity and linearity.

How did you handle the overlap problem with your rig?

I've looked at creating a reel guide geared to the drum, but that's no longer a hardware store solution. (We're back to machining parts.)

I'd much rather find a Hardware Store Solution, if at all possible.

Any other suggestions, on how to make a String Pot with hardware store items, that will drive a standard 1/4" shaft multi-turn potentiometer? (OR, does anyone have a COTS source for CHEAP [say ~$20] 25" long string pots???)

- Keith

[Al Skierkiewicz]

Keith,
How about salvaging a print head position encoder from a dead printer?

[AdamHeard]

Originally Posted by kmcclary

The big problem seems to be fabbing a lanyard and drum rig that prevents string overlap. If an overlap occurs, you've completely lost data integrity and linearity.

I would say reduced instead of completely lost.

I recommend you try out the pot and lanyard reel and see how accurate it truly is.

What are you actually measuring, and what are you trying to determine based on these measurements? The magnitude of this inaccuracy may be rather insignificant compared to the magnitude of your final result.

I get in a similar mode where I'm so deadset in my view of the problem, that I am too quick to write off solutions that probably could work.

[Bob Steele]

How about using a spring tape measure and a simple photo resistor.
Paint the tape measure white...put your marks on it...use the photo resistor to "see the marks" ... count them.... process.
You will have to figure out how to tell the difference from going up and going down.... but that is doable too...

[kmcclary]

Originally Posted by AdamHeard

I would say reduced instead of completely lost. [...and what are you actually measuring?]

True, but from an instrumentationalist's perspective, if it won't give me the same value every time I return to the same point, I can't use the solution.

What is our project? This is a Fall R&D project. We wish to gain the capacity to accurately servo control a 24" sliding joint to say ~7-8 bits (or better) position resolution over its length (~4-10+ steps/inch) , with a cheap, easily made, compact feedback device.

Rack and pinion pots won't easily fit there, and optical slides could get dirty. Without a rotational joint present, cheap pots won't work either.

Something equivalent to a 25" string pot (into our Analog Inputs) seems ideal, but it:
A) Needs to be CHEAP to make (no $100 commercial string pots!), and
B) Preferably can be assembled MOSTLY with COTS hardware store items, with as little machining as possible.

Ideally, the sensor should be made with Hand Tools only, HT + Dremel Only, or at least as few specialized machine tools as possible. That makes the solution available to ALL teams, regardless of budget, or machine tool capacity.

We've tried dissecting and hacking on some key chain retrievers to add a 10-turn pot, but that still requires machining several parts. The basic KCR drums allow string overlap too, making it unsuitable for reliably hovering an axis, or even returning it to the same position.

I was hoping that someone here already solved this generic problem with COTS hardware store items. (Has anyone???)

BTW: Custom PCBs and soldered on arrays of lots of VERY cheap sensors are still included as a candidate here, as long as the total cost for the entire sensor array device would be in the $25 or better range.

Now, I like the idea about turning a measuring tape into an incremental encoder! That's the kind of divergent thinking I'm seeking here! However it would be NICE not to be forced to run the axis to one end at boot-up just to home/sync the position counter. Not all applications may easily allow us to do that. Something that gives us an absolute position at bootup is preferred.

How about something like:
A) Having a pinch roller pair and a 10-turn pot watch a COTS tape measure go in and out? (Friction/slip may be a problem. Can that be overcome?)
B) Punching/drilling holes in a COTS tape measure, and have a crude drum sprocket and 10-turn pot follow it? (No slip, but requires more fab time to mod the tape, and make the pin drum.)

Some linear devices:
C) Use a Mylar CNC 8-channel punched tape, punch codes into it, lay it out, and watch it optically. (Not all teams could easily replicate this one.)
D) Laser Print an encoder strip with multichannel absolute Grey code, watch it optically. (Cheap and doable by all teams, but may have mounting issues in some instances.)

Both C & D require more space, and may have dirt issues (which is why we keep finding ourselves gravitating back to string pots techniques)...

Thoughts?

(Hey Al - do you know of a cheap printer head monitoring system, that'll do 25"??? It has to be COTS, so we could use it in a contest.)

- Keith

[AdamHeard]

Originally Posted by kmcclary

What is our project? This is a Fall R&D project. We wish to gain the capacity to accurately servo control a 24" sliding joint to say ~7-8 bits (or better) position resolution over its length (~4-10+ steps/inch) , with a cheap, easily made, compact feedback device.

If this is a prototype for possible FIRST use, I would say that +/- 1.0 inches is sufficiently accurate. If your actual design during season requires more than that, it won't be very human friendly.

Now, purely for R&D it's a cool idea, but don't think that such accuracy is always required. Your manipulator should be designed to allow substantial amounts of inaccuracy.

[Alan Anderson]

Originally Posted by kmcclary

Rack and pinion pots won't easily fit there, and optical slides could get dirty. Without a rotational joint present, cheap pots won't work either.

Something equivalent to a 25" string pot (into our Analog Inputs) seems ideal, but it:
A) Needs to be CHEAP to make (no $100 commercial string pots!), and
B) Preferably can be assembled MOSTLY with COTS hardware store items, with as little machining as possible.

Try a "chain pot". Instead of string, use a length of #25 chain. Think of a flexible rack & pinion connected to the end of your slide, with the chain bending around a pot-equipped sprocket and having a tensioning spring at the far end.

With a large enough sprocket, you might even be able to get away with a single-turn pot (or a magnetic absolute position sensor).

What are you using to move the slide? If it's a chain, you're most of the way there already.

[kmcclary]

Originally Posted by AdamHeard

If this is a prototype for possible FIRST use, I would say that +/- 1.0 inches is sufficiently accurate.

I think you are missing the point here. This is not to solve a SPECIFIC design problem, it is For The Toolkit, as a Challenge... You can always DROP resolution for simpler needs, but it is much tougher task to INCREASE it.

Originally Posted by Alan Anderson

Try a "chain pot". Instead of string, use a length of #25 chain. Think of a flexible rack & pinion [...] What are you using to move the slide? If it's a chain, you're most of the way there already.

Oooooohhh... Now THAT is an excellent idea, with potential! We'll have to look at that one! (See... I have faith in creativity of this community!)

BTW... What moves the slide, and the load are both irrelevant at this time. The challenge goal, as stated, is to make/develop/find a CHEAP and SIMPLE way to SENSE POSITION of a servo controlled 24" slide, and if at all possible, with the above spec'd resolution (or the best we can do, for the given $20-$25 budget). We can choose the app and a specific actuator later.

Because: A closed loop system is first off limited by the resolution of the feedback sensor/system. That is why we are shooting for what some here seem to feel is "excessive resolution". Given a "decent" resolution sensor, we can do ANYTHING. Once one knows where the slide is, LATER experiments can then test how EITHER cylinders OR motors actually perform, to move it around. (Which works better may well depend on the final app, load, and mounting orientation.)

Heck, if we can't REACH 0.1" resolution with the sensor that may be fine, but that is not a reason to never even TRY for it. Who knows? If can only hit 1" resolution for $20 per axis we are still WAY better off than before we tried, and we STILL have a new trick up our sleeve. (But I would bet if the team tries, it CAN do better than 1" resolution for $20 in sensing cost [discounting labor of course]...)

Any more cool ideas on making a cheap and simple feedback sensor, for 2' slide joints, for about $20, and using primarily COTS hardware store hardware??

- Keith