Ball bearing encoders - DIY?

[Mcgyver]

not sure if this belongs hear, but I'm exploring some closed loop ideas that would require encoders and came across these from Newall. I thought this encoder scheme was really neat - incredibly accurate, no moving parts etc.

http://www.newall.com/Newall_US/Img/pdf/688.pdf

They are around $500 each (say 12 or 24 ") so not completely unreasonable and lot less than ground ball screws. Nevertheless thrift is part of the home shop culture so I was wondering whether this learned group sees this as something that could be homebrewed? ideas?

Mike

[Evodyne]

Hi Mcgyver!

Thanks for sharing the info. What a simple idea! The good ones always are.

I've got a few optical linear encoders (RSR, Heinehain) that offer similar resolution, but would be darn hard to duplicate because of the intricate gratings/tolerances.

This however looks like it could be done. With 5mm ball sizes it's not microscopic-the required coils could likely be scavenged from common equipment. Driving the coils at 1-10 khz is no biggy either. A pair of amplifiers to buffer and magnify the induced voltages, some type of peak-reading circuity, two channels of A-D conversion, and a small microprocessor like a PIC, and it could be done at home.

Try moving the thread to the electronics forum, since the difficulty (or challenge) would be the electrical portion.

A thought: the guys who know (not me!) say your better off using a rotary encoder on the motor vs. a linear encoder on the slide unless you have no backlash. It prevents hunting. You say this "...is cheaper than ballscrews" as if you plan on using the encoder to achieve accurate positioning with a system that has backlash (as kind of a work-around). You're likely to run into problems. This has been addressed in other threads-do a little searching/reading.

Finally, if you want to experiment for well less than $500/unit, I have a few optical linear units I'm trying to sell-see the classifieds or contact me.

Lance

[Mcgyver]

k, Lance I was hoping to avoid going there as the underlying idea is kind of wacky and not fully developed, but you've forced my hand!

Rotary encoders are not applicable. The reason for linear encoders is I'm trying to figure out how to use hydraulics instead of ball screws to control a mill. I want all the benefits of precision balls screws without the 2k/axis price! So my idea is 1) small gear pumps are cheap. 2) cylinders are cheap. What’s not cheap (I believe) are the proportional electronic valves etc. So my idea is, because you can run a gear pump in two directions, why not hook up each side of the gear pump to an end of a double acting cylinder, thereby having one pump per axis with its own DC motor that is controlled (direction, speed and acceleration) via microcontrollers close looped with a linear encoder.

it avoids the bang bang of hydraulic systems (because you can control the acceleration of each cylinder), the high price components, resolution to tenths, solid etc etc. I know there’d have to be some pressure relief, snifter valves etc. As per my wacky idea of concrete (now crushed granite & epoxy) filled normalized fabricated superstructures, I table with the hopes of learning from the peeps here. I know little about hydraulics so there may catastrophic challenges to this idea, but I’m exploring!

My objective is an uncompromising open source cnc mill who’s price tag still qualifies it as a hobby endeavor and find myself having to think up some crazy stuff to solve that conflict!

[Evodyne]

Mike,

In no particular order:

I seem to remember one of the automation companies (Allen-Bradley I think) did something similar (your linear encoder feedback) with pneumatic cylinders to get accurate positioning despite the compressability of air.

I know a guy who has done granite-epoxy machines on commercial machines-micron level positioning of optics where vibration is a big no-no. I can PM you his email address if you want to talk to him.

How big of a machine are you thinking of building? I've managed to get good 5'-6' long, 1" to 1-1/4" diameter ground screws with bearings for $300 and under after lots of scrounging. That source has dried up, but there must be others selling similar stuff. What is the basis for your 2K/axis cost? What would cylinders cost?

Are you at the point where your thinking about a small, proof of concept rig?

Curious to hear your reply...

Lance

[Mcgyver]

There are actuators and solenoid valves etc for using hydraulics for linear motion but the electro hydraulic systems are expensive. My idea of a pump on each axis and eliminates the expensive valves. With air you don't the bi directional advantage of a pump and there is the compressibility complication. What I envision is kind of a fluid version of a capstan, like on a small surface grinder. There’s going to be almost no pressure as the pump is just pushing fluid back towards its input.

I've looked at various hydraulic linear actuators and there's some slick stuff. Earlier hydraulic linear motion had the problem of sudden starts and stops. Newer controls overcome that, but again are expensive (as far as I can tell, none of these companies post price lists. grrrr).

Off the shelf low volume pumps and cylinders might be $200 -300 per axis. By hydraulic standards, this will be an ultra light application.

If I put a new Newall with it, (and presuming the idea works) I've got an axis that will virtually never wear out, has zero backlash and is accurate to a few microns per meter for a fraction of the new cost of ground ball screws, precision angular bearings and servo.

The size of the mill is not finalized, maybe 24x12” of travel, maybe less. Sort of a large version of a bench top mill. Weight will be 1000-1500 lbs. It’s got be a solid performer in cutting ferrous metals. It's probably premature to put me in touch with your friend. While there is much to learn, I’m pretty sure concept for the superstructure will work, so I’ve compartmentalized that off to one side for the moment as I puzzle out the rest of it.

The 2k per axis cost came from something I read on a cnc site say that is what the ground balls screws + bearings cost. Ideally I don't want to scrounge. The project may be a collaborative effort or benefit others down the road so I’d like parts to be available. Also, being in Canada adds costs and hassles to buying from ebay.

I would like to prove the encoder/controller/hydraulic idea next. Then again, this concept is taking me into many new areas so I'm in not rush to start buying stuff. First experiment will probably be to develope a simple controller hooked up to a pc and see with what resolution I can move the cylinder, around, etc.

Hopefully a homebrew linear encoder is feasible which would lower the per axis cost further.

[H500]

Interesting idea. The cheapest way to do a linear encoder would be to convert the linear motion into rotary motion.

I wonder if hydraulics cyls can be controlled with such accuracy. They tends to be very non linear at small displacements. Some time ago, I played with displacement feedback using servo valves and I did not get the impression that the system was capable of correcting small errors. It was very easy for the system to go into oscillations.

Have you looked at the rolled ball screws at McMaster Carr? They should cost less than $100 per axis. The backlash can be removed by using 2 nuts.

[lerman]

A few years (well, decades) ago, I worked on a control system for the retrofit of a copper rolling mill. The original machine used large screws with motors to control the thickness of the metal.

The retrofit used hydraulic cylinders controlled by servo valves (I believe they were made by Moog). It was fast and precise. But not cheap.

Ken

[Evodyne]

Mcgyver,

Hi! I dug through an older Grainger catalog (2000-2001, cat. no. 391) and found their smallest positive displacement gear pumps (Haldex/Barnes Hydraulics, page 3014). Curious as to what these would do in your setup I threw together an Excel spreadsheet. I saved it in '95 format to avoid compatibility problems.

Looks like a good size NEMA 34 200 steps/rev with about 300 oz-in of torque might work well with a 1.333" diameter piston...take a peek.

The pumps (in 2000-2001) ran from $114 to $145 U.S.

A thought: if the pump rotation translated directly to linear travel (no trapped air, etc.), and the motor was coupled to the pump, then a rotary encoder on the stepper (to indicate actual rotation vs. missed steps) might work as well as a linear encoder. If you knew you had enough motor torque, then just counting motor pulses might be enough.

Lance

[Mcgyver]

Lance, thanks for doing all the thinking!

The problem is that pumps are designed with slippage so rev counting wouldn't be accurate.....so you're not getting out designing the ball bearing encoder electronics! (kidding)

That was a great find on the Haldex pumps. That pump is currently $118 US from the Grainger web page. I got similar number to yours, thanks for the spread sheet – much better than what I did. I was using a 2” bore, 1” rod which gives 3.14 and 2.35 in area depending on the direction. I still like that pump as I’d want a slower speed than 5ipm. If you’d never use such a slow feed, you could put a larger pump on and have a faster rapid traverse.

On the non-linear nature, controlling the low volume pump speed rather than a valve that controls a couple of thousand psi should help (keeping my fingers crossed) . I’m hoping to avoid the most expensive items, as unlike the rolling mill, this hydraulic system will be extremely low pressure – I’m coining it a fluid closed capstan.

The electronics stuff I’ve got to work on is a pwm motor control (I think I know how to get started) and the linear encoders (just barely understand the principal), followed by figuring out how to make the two cooperate (starting to read about using PICs).

Mike

[Evodyne]

Mike,

Looking at the specs on the pump it's output is entirely linear. The positive displacement types shouldn't, if there good and tight, have any leakage.

What was encouraging was that the spreadsheet showed that the typical motors we see guys use (and therefore readily available) seem to mesh well with the pumps if the cylinder is on the order of 1.3" to 2" in diameter. So your idea looks even more promising.

I'd recommend the Geckodrives for steppers (vs. building)-I've got a few and I love them. As they are "choppers", they give good high speed performance. The G201's are $114 straight from Mariss. I'm using the same logic that you used on me concerning the screws (only fair, right? ). Requoting you, "The project may be a collaborative effort or benefit others down the road so I’d like parts to be available".

Like you I'd like to do stuff with the PICs. I just need to do it. Right now I'd like to do a little inertia dyno for servo and stepper motors.

Another thought (I'm full of them. ...or of something else!): have you seen the complete slide assemblies using rodless pneumatic cylinders? I think Festo is one company that makes them. I'm wondering if one could be used with hydraulic fluid if you stayed in the recommended pressure range. This would be an entire axis "out of the box" and ready to go.

That's all for now!

Lance

[Al_The_Man]

Ever heard of a Digital Hydraulic stepper cylinder?
http://www.victorycontrols.com/
I came accross this company a few years back and they have some interesting devices, including servo valves operated by stepper motors, Typical servo valves operate with a pilot operated valve, and contamination is a problem usually, these stepper operated valves operate on the main valve actuator and reduces this problem.
Al

[Mcgyver]

Al, that’s the perfect product for this app, If I survive their response to my price request maybe its the way to go - thanks.

Lance, I'm still inclined towards closed loop servers vs. rotary because I can't rely on the manufacture's claim that its linear. Why? A claim of linearity means the plot closely follows a straight line, but that is not say there isn't a variation. That variation may insignificant but I don't know that. I also can't imagine that the manufactures tests concentrated much on 100-200 rpm range.

Its not leakage that I'm worried about but slippage. This is designed into the pump for lubrication - i.e. the clearance between gears and the body of the pump. Even if it was a fraction of a percent it would quickly put you out several thou. Finally, slippage will change with pressure & viscosity (and maybe speed). Maybe not enough to dispute a linear plot, but maybe more than enough make a rotary encoder inaccurate.

No problem using someone else’s controllers. I also don't intend to make my own ball bearings or wire then again, originally I had a few friends interested in making it. if its 4 peeps x four axes the controller cost starts to add up.

I like the idea of pneumatic hardware used with hydraulics and am getting prices on Festo cylinders. Ultimately it doesn’t look beefy but maybe for an interim circuit board mill? I wonder if there are components in the air stuff that won't stand up to hydraulic fluid?

Apologies to the electronic experts, the thread's been mostly mechanical. but the hope is still to get some ideas on the how accurately measure the change in inductance to build a Newall type encoder. An encoder accurate within microns per meter has got to have some appeal to this crowd? Ideas?