Question about precision and repeatability

Discussions relating to development and use of the OpenServo hardware.

Moderators: jharvey, Secondary Admin, Admins

engr.eric
Posts: 8
Joined: Mon Apr 06, 2009 9:53 pm

Post by engr.eric » Thu Apr 09, 2009 5:08 pm

So I was playing around with a dissected HS-311 and to do the TLE4966 encoder, I noticed that there was no good place to attach a magnet to a one of the high rpm gears. At least not without cutting some plastic on the housing.

This made me look at the AS50XX again and the drawings that jharvey provided showed a lot of promise. For the magnet, wouldn't it be as simple as getting a small neodymium and hot glueing to the output shaft where the pot sticks in. I don't think it would need to be d shaped

jharvey
co-admin
Posts: 362
Joined: Sun Mar 15, 2009 12:06 pm
Location: Maine USA
Contact:

Post by jharvey » Thu Apr 09, 2009 10:15 pm

D shaped is probably a good idea. Perhaps it could be hot glued, or perhaps it could fit with a mild interference fit. I remember stumbling across some "potable" magnetic material I think it was an epoxy with Alinco materials in it. If I can find that stuff again, and if it's not stupid expensive, I think I might gut a POT, and make a wax mold from the existing shaft. Then POT the mold, making an exact replacement of the shaft. Then with a car battery and some metal coat hanger, I think I get enough flux out of the bugger to produce a custom magnet.

Someone thought that fridge magnets are simply this powder suspended in plastic. Perhaps I can simply melt some of those flexible magnets to get my potting material. If only I had some nichrome wire sitting around.

Cliff
Posts: 150
Joined: Tue Jan 23, 2007 1:25 pm
Location: Saratoga, CA

Post by Cliff » Fri Apr 10, 2009 10:52 pm

Jared,

A 'D' shaped magnet will not work with the on axis AS50xx devices - it needs symmetrical N-S diametric fields to work. Also, Alinco magnets are not strong enough. The recommended magnet types are NdFeB ( Neodymium-Iron-Boron) or SmCo (Samarium-Cobalt) - both of which have seriously strong fields. In case you missed my earlier link to Austria Micro' magnet selection guide, here it is again: http://www.austriamicrosystems.com/eng/ ... load/13858

Cliff

jharvey
co-admin
Posts: 362
Joined: Sun Mar 15, 2009 12:06 pm
Location: Maine USA
Contact:

Post by jharvey » Sat Apr 11, 2009 9:13 am

That's interesting. I had only skimmed the PDF you had linked to, because it notes most of that content in the chip's data sheet. In the AS5043 sheet, section 15 Choosing the Proper Magnet, it notes "Typically the magnet should be 6mm in diameter and ≥2.5mm in height. Magnetic materials such as rare earth AlNiCo, SmCo5 or NdFeB are recommended."

I was looking at AlNiCo because it's fairly low cost. However, it's lower flux density may make it harder to work with. The recommended gap between the chip and the magnet ranges from .5mm (0.019) to 1.8mm (0.070in). Using AlNiCo may narrow that gap even more. Looks like NdFeB is the next lowest cost, and the magnets noted on their on-line store are NdFeB. I also see the PDF notes corrosion protection yes, where SmCo is noted as not required.

Is the corrosion an issue for us?

wurpfel
Posts: 25
Joined: Wed Dec 20, 2006 3:23 pm
Location: swiss

Post by wurpfel » Sat Apr 11, 2009 1:46 pm

hopefully not... its inside of the servo ;)



I use nickel-plated (thats the normal manufacturing prosess) when mount "outdoor" in the rudder-axe
not to old to play with scrap

engr.eric
Posts: 8
Joined: Mon Apr 06, 2009 9:53 pm

Post by engr.eric » Sun Apr 12, 2009 8:39 pm

About the AS50XX Devices,
I just sampled the AS5046 Device and I am going to prototype it with a bit-bang I2C interface on an avr. I still don't have any open-servo boards so, integrating my bit-bang I2C with the project later could be interesting and a little difficult.

There have also been talk of the different AS50XX devices. I'm not too particular of the resolution 10bit or 12bit is plenty for this project in my opinion. But the different interfaces on the devices will change things.

The reason I like the AS5046 specifically is because, well I'm comfortable with I2C and generic Two-Wire Serial connections, and I have written bit-bang code for these before, but some of the other AS50XX devices use an absolute SSI connection which I am unfamiliar with.

The other nice thing about AS5046 is the optional 10bit analog output, which could be a direct replacement of the potentiometer.

Pros:
-I2C, easy for me to write bitbang code.
-12bit max digital resolution, overkill but nice
-Pot replacement by 10bit ADC; which I think provides the fastest open-servo integration.

The only code required is an init section for the as5046 from open-servo and the rest would just be a pot replacement.

And after coming up which an easy installation method and retrofit for the current open-servo V3 then the same hardware without modification can be expanded by utilizing the I2C interface for reading the sensor values.

I am curious about what I may be overlooking from the other AS50XX devices?

Thanks!

jharvey
co-admin
Posts: 362
Joined: Sun Mar 15, 2009 12:06 pm
Location: Maine USA
Contact:

Post by jharvey » Sun Apr 12, 2009 8:52 pm

No objections about the 5046 here. I simply chose the 5043 because it was cheaper, and offered the analog. I hadn't looked into the digital side at all. I'll plan for the 5046.

Also your bit banging pins probably want to use AVR pins 9 and 10. That's (PCINT21/OC0B/T1) PD5 and (PCINT22/OC0A/AIN0) PD6. The most recent layouts from Cliff have pads for these pins. This will likely make it easier to attach wires to if connected to open servo boards.

jharvey
co-admin
Posts: 362
Joined: Sun Mar 15, 2009 12:06 pm
Location: Maine USA
Contact:

Post by jharvey » Wed Apr 15, 2009 3:18 am

How about a design something like this?

Image

Image

Theory being, your PCB outline is a washer. You mount the AS5046 chip over the center hole. The magnet is a standard rod with a notch cut making the D part. These parts replace the shaft and PCB from a standard POT reusing the POT's housing. The magnet shaft is inserted such that it touches the bottom of the AS5046 chip through the PCB. The PCB would maintain alignment. If more spacing is required, a hole punch and a small piece of plastic can easily increase the gap between the die, and the rod.

The key thing I don't like is that you don't have room for alignment. If you don't solder the chip within that +/- .01, you up a creak. However, it can't be that hard to maintain alignment of that one chip can it?

I also wonder if you don't hit that .01, what happens. Does it change from say 12 bit to 11 bit (noise floor increasing due to weak signals)? Does it get a dead spot when the flux is out side of that target area? Does it do something else? Perhaps it's linearity isn't just perfect? Hmmm.

engr.eric
Posts: 8
Joined: Mon Apr 06, 2009 9:53 pm

Post by engr.eric » Wed Apr 15, 2009 6:09 am

I was under the impression that the AS5046 was pretty good at compensating for misalignment.

Have you tried any tests with the ic yet? I have my samples in and I am bread boarding a small circuit. I'll have more news about what I find once my midterm is done this week

wurpfel
Posts: 25
Joined: Wed Dec 20, 2006 3:23 pm
Location: swiss

Post by wurpfel » Wed Apr 15, 2009 6:34 am

dont fool about misalingment: +/-0,5mm woulnt be a problem, you can even build a centerpush-function with the testpin ;)


I replaced all poties in my TXbox: very exited about precission and feeling..
not to old to play with scrap

jharvey
co-admin
Posts: 362
Joined: Sun Mar 15, 2009 12:06 pm
Location: Maine USA
Contact:

Post by jharvey » Wed Apr 15, 2009 11:03 am

My feel about it is that if you aren't in that perfect .01", you simply loose some of the linearity of the device. The readings that range from 0 to full scale would simply have a little wonder in it, instead of being a straight line. Effectively reducing the 12 bit accuracy to 11 or 10 bit. Either way, I suspect that is more than accurate enough for most applications.

Perhaps we can leave some vias that can allow for an alignment option if someone needs it. Or perhaps we should simply not worry about it, and if someone wants to keep extreme accuracy, they can re-spin the board to add an alignment option. I think I'll assume the chip will be placed accurate enough.

engr.eric do you know when you should receive the samples? Can you make a proto similar to what I've modeled above?

jharvey
co-admin
Posts: 362
Joined: Sun Mar 15, 2009 12:06 pm
Location: Maine USA
Contact:

Post by jharvey » Wed Apr 15, 2009 11:42 am

I just updated the above pictures. They now include an optional hole where a pin can be inserted to hold the shaft in place in an event when the shaft can't be held in by sandwiching the POT Encoder in the center of an assembly. This same hole in the shaft can also be used to prevent thrust loads from being absorbed by the AS5046 chip.

They also better show how the PCB is held in place.

Also here's a pic from the back with some parts as transparent.

Image

engr.eric
Posts: 8
Joined: Mon Apr 06, 2009 9:53 pm

Post by engr.eric » Wed Apr 15, 2009 4:30 pm

actually, I have the samples already. I was going to attach a proto board with the ic to my mini mill's xy table, attach the magnet to the spindle. turn on the mill at a low rpm and move the table around to see how much of a difference it makes =) I think I should get a pretty good idea from that, and Ill just graph the curves on excel

jharvey
co-admin
Posts: 362
Joined: Sun Mar 15, 2009 12:06 pm
Location: Maine USA
Contact:

Post by jharvey » Wed Apr 15, 2009 5:04 pm

Awesome, and keep me posted. You may also want to try the test with the magnet stationary, then see how the output is affected by moving it side to side. That test might produce some quick results. The spinning approach I suspect will require more work because it would produce an AC signal right.

Picture or video are might help :wink:

engr.eric
Posts: 8
Joined: Mon Apr 06, 2009 9:53 pm

Post by engr.eric » Wed Apr 15, 2009 5:54 pm

yeah, it would create some sinesoidal signal

Post Reply

Who is online

Users browsing this forum: No registered users and 0 guests