High accuracy servo, various component selection

Hey all, I'm not sure if this is the right section to post this in but I'm new here so show some mercy! Mods, feel free to move it if it's wrong.

Anyways, I'm looking for some advice and guidance regarding component selection for my project, especially regarding servos. What i'm looking for is a servo with a very high degree of repetitive accuracy; I'm talking within a few degrees thousands of times. End game would be years of use without calibration. Without divulging more details, if my product ever made it to market (I know, a long shot but I can hope) and this servo was inaccurate, it could be life threatening. Would a secondary feedback system be necessary for this? I can also spend more than your average hobby servo, but I'd like to keep it as cheap as I can.

As for other components, I'm also looking for a small vacuum sensor, a cheap color display (preferably at least 3.5") , and some directional backlit buttons (the squishy rubber kind you see on appliances). I'm not in dire need of guidance and I think with enough googling I could figure it out, but I would definitely take suggestions from those more experienced than I. With so many electronics websites, it's hard to just jump in and start buying stuff since I don't really have the feel for the market that only comes with experience. However, the servo is a priority. Thanks!

it could be life threatening

I think you may find it difficult to get people to give you advice on selection of devices in life critical devices.

E.g there was a thread about heart attack detection, on this forum, where people explained why they could not give advice, because of the legal implications.

Well the device itself wont kill you. If you had someone who just absolutely wasn't paying attention though, they could hurt themselves using it if the servo wasn't in the right spot. It's not nearly as serious as a heart attack detector. Basically think of it as a knife. If I could tell you more I think you would understand. But I get your point. I'm not sure how legally sound this is, but I absolve anyone on here of any legal or moral responsibility. I just need an accurate servo :D. Besides, this thing is lightyears away from any sort of production. I'm trying to make a functional prototype that I can pitch to people, it would never actually see use.

precision servos exist for many applications but you get what you pay for. With servos, it is more about the manufacture of the motor itself I think than the feedback unless you are talking about absolute encoders. The servo itself and the absolute encoder are not the kind of things you can make yourself. If you have enough money you can get the precision you need. They are made by companies that make miniature insulin pumps or other medical related hardware or laser mirror positioning motors. http://www.physikinstrumente.com/en/pdf_extra/2009_PI_Fast_Steering_Mirror_Scanner_Catalog.pdf

The specs for this one say

Millisecond Response, Resolution to 0.5 ?rad

I'm no math genius but if a urad is 57.6 degrees /1,000,000 = 0.0000572957795 degrees, then 0.5 urad = 0.00002864788975 degrees I could be wrong but I read that as 28.6 udegrees. Is that precise enough for you ?

Is the critical position of your servo on the end points of the movement? Can you place a physical barrier past which the motor can't move?

Raschemmel , I wonder what's the price of those thing. They sure sound costly.

Hi, I think the OP posted:

I'm talking within a few degrees thousands of times.

I'm talking within a few degrees, thousands of times.

Not a few thousandths of a degree.

Tom......... :)

tboneh: if ... this servo was inaccurate, it could be life threatening. Would a secondary feedback system be necessary for this?

Absolutely essential - and the whole system would need to be designed so that all conceivable failure modes failed to a safe condition. Without that level of protection, the nominal repeatability / accuracy of the servo is the least of your problems since you have to assume that eventually one or more things [u]will[/u] go wrong (mechanical wear, electronic failure, firmware bug, wiring fault, user error etc) so that the servo position will go arbitrarily wrong.

There's nothing to stop you from putting together a proof of concept, and using it yourself if you want, but before you sell a product which has safety implications you have a massive amount of work to do and you should not be relying on advice from a public forum about how to do it - you would need expert qualified advice with appropriate indemnities.

Raschemmel, I appreciate the enthusiasm but those might be a bit TOO accurate :P. I'm guessing the price of those is probably in the hundreds. Now that it's starting to look like a secondary feedback system is needed, I think I can sacrifice some accuracy in the servo itself for a cheaper solution.

Shpaget: Is the critical position of your servo on the end points of the movement? Can you place a physical barrier past which the motor can't move?

I'm going to put a table of sorts on top of the servo. The critical positions are several regular angles. Currently it needs to go to 8 positions, divided evenly, so 40, 80, 120, etc. I need a full range of motion, so physical stops are out of the question. I was thinking of gluing 8 bumps to the underside of a table that depress a single switch as it goes around. However, then you deal with the problem of mechanical wear. Do you guys think reed switches could give positional accuracy within a few degrees on say, a 3 inch diameter circle if calibrated correctly? Also open to other forms of feedback if anyone has a suggestion.

Oh and PeterH, I hear you. If it was ever made I'm guessing a lot of the components would change due to regulations and safety concerns. But right now I just need it to work, preferably for as little money as possible! Plus low cost aids in marketability.

Is the servo going to be driving your table directly? How big/heavy is this thing and what sort of loads will it have on it? I'm struggling to visualise the mechanical side, and without understanding that it's hard to advise how to drive it.

I'm talking within a few degrees, thousands of times.

Not a few thousandths of a degree.

Oh, my bad. Must be the A.D.D....

PeterH: Is the servo going to be driving your table directly? How big/heavy is this thing and what sort of loads will it have on it? I'm struggling to visualise the mechanical side, and without understanding that it's hard to advise how to drive it.

Yeah I understand, I'm not trying to be mysterious or anything, I just don't really want my idea on the internet for everyone to see. But I can describe the mechanical side. The table is indeed directly driven by the servo, but very light loads. It's a round plastic piece about 3 inches wide and an inch or so high, mostly hollow. The goal is to get said table to rotate to specific angles with an high degree of repeatability. Now that I realize secondary feedback is necessary, my question has shifted from not what type of servo to use but rather a feedback system that won't wear with use. My two thoughts so far are a single reed switch and a magnet array under the table, or maybe an optical encoder.

Perhaps a Hall sensors would be better than reed switches. Reeds bounce, could stick or not close properly. Stay away if you can.

If you can stay away from safety critical issues and long term durability, I would have thought that most servos would be able to achieve a resolution of around a degree. Digital servos are faster and better at holding position if you need to apply a sustained torque but I doubt you need that here from your description.

In my opinion the 'mark 2' version designed for long term use would be better using a geared DC motor with a rotary encoder on the final drive - this will eliminate problems due to wear on the pot and drive mechanisms. However it will be more expensive and harder to get working.

If the system is safety critical then you need to address the possible failure modes and resulting liability issues, and I'm not going to tell you how to do that.

Shpaget: Perhaps a Hall sensors would be better than reed switches. Reeds bounce, could stick or not close properly. Stay away if you can.

Bingo, those look perfect. That can also allow me to detect the area of most intense magnetic field so I can position my magnets right on top of it. Thanks!

PeterH, what do you mean by "mark 2"? What you described sounds good but I can't seem to find anything online about it. And believe me I understand all the safety stuff, it's something I've known about for a while so it's kinda beating a dead horse at this point. This is only a prototype and it's impossible for it to hurt anyone because it won't see use, and I really need to focus on getting it working before I should worry about liability. And if I can get a high degree of repeatability I can always program in safe states and failure detection later.

tboneh: PeterH, what do you mean by "mark 2"?

I mean that after you have the first prototype working to demonstrate the basic functionality using a standard servo, you can create a better solution based on a discrete DC motor/gearbox and rotary encoder that is better at meeting your durability requirements.

Oh alright I get you. I think for my first approach I’ll try the hall sensors and maybe some strong magnets embedded at 8 points on the underside of my table, since it’s simple and sure to work but with a little less sensitivity. If I get some funding then I can try the encoder and a DC motor. Anyone have any favorite servo brands? I need something very reliable and with just a reasonable amount of accuracy (since I’m using secondary feedback now).

Futaba and Graupner are good brands.