Rotary encoders

[Andy from WorkshopshedParticipant @andyfromworkshopshed]

Ah yes, I've heard that can be done, understand now. Here's the link http://www.youtube.com/watch?v=bci7Gi05BNc

Edited By Andy from Workshopshed on 09/05/2013 17:17:45

[Andy from WorkshopshedParticipant @andyfromworkshopshed]

Ah yes, I've heard that can be done, understand now. Here's the link http://www.youtube.com/watch?v=bci7Gi05BNc

Edited By Andy from Workshopshed on 09/05/2013 17:17:45

[Sub MandrelParticipant @submandrel]

Thanks to a suggestions from Michael Gilligan I found THIS chip with a 30×30 pixel view and the ability to take and stor a whole frame (actually 2/3 of a frame). It uses SPI so very easy to interface with an AVR/Arduino or even a (spit) PIC/Picaxe!

Unfortunately I haven't fopund a source of the chips, and all the mice I've dissected so far have (even) low(er)-res chips that don't store a frame, just take sequential pixels.

I want to fit an 'eye' to my YETI!

Neil

[Sub MandrelParticipant @submandrel]

Thanks to a suggestions from Michael Gilligan I found THIS chip with a 30×30 pixel view and the ability to take and stor a whole frame (actually 2/3 of a frame). It uses SPI so very easy to interface with an AVR/Arduino or even a (spit) PIC/Picaxe!

Unfortunately I haven't fopund a source of the chips, and all the mice I've dissected so far have (even) low(er)-res chips that don't store a frame, just take sequential pixels.

I want to fit an 'eye' to my YETI!

Neil

[Sub MandrelParticipant @submandrel]

Thanks to a suggestions from Michael Gilligan I found THIS chip with a 30×30 pixel view and the ability to take and stor a whole frame (actually 2/3 of a frame). It uses SPI so very easy to interface with an AVR/Arduino or even a (spit) PIC/Picaxe!

Unfortunately I haven't fopund a source of the chips, and all the mice I've dissected so far have (even) low(er)-res chips that don't store a frame, just take sequential pixels.

I want to fit an 'eye' to my YETI!

Neil

[Michael GilliganParticipant @michaelgilligan61133]

Neil,

This German list might help you find a suitable sacrificial Mouse

and here is an interesting application

.

MichaelG.

Edited By Michael Gilligan on 10/05/2013 08:34:37

[Michael GilliganParticipant @michaelgilligan61133]

Neil,

This German list might help you find a suitable sacrificial Mouse

and here is an interesting application

.

MichaelG.

Edited By Michael Gilligan on 10/05/2013 08:34:37

[Michael GilliganParticipant @michaelgilligan61133]

Neil,

This German list might help you find a suitable sacrificial Mouse

and here is an interesting application

.

MichaelG.

Edited By Michael Gilligan on 10/05/2013 08:34:37

[Michael GilliganParticipant @michaelgilligan61133]

Neil,

I've just rediscovered this page

Lots of useful info.

MichaelG.

[Michael GilliganParticipant @michaelgilligan61133]

Neil,

I've just rediscovered this page

Lots of useful info.

MichaelG.

[Michael GilliganParticipant @michaelgilligan61133]

Neil,

I've just rediscovered this page

Lots of useful info.

MichaelG.

[Sub MandrelParticipant @submandrel]

Thanks Michael. No mouse is safe from me now…

Neil

[Sub MandrelParticipant @submandrel]

Thanks Michael. No mouse is safe from me now…

Neil

[Sub MandrelParticipant @submandrel]

Thanks Michael. No mouse is safe from me now…

Neil

[Andy from WorkshopshedParticipant @andyfromworkshopshed]

I gave a demo of my work in progress rotary encoder at the London Arduino meetup yesterday and there was a very relavent question.

"How accurate does it need to be"

Currently I've got it at 1024 postions per revolution, which gets me about 0.3 of a degree. I've seen commercial sensors that can go down to 0.1 degree and by using the magnetic chip I mention or if I can get the optical mouse solution to work then I should be able to match that.

So how would I calculate how accurate the table would need to be for say cutting an 8 tooth gear?

Edited By Andy from Workshopshed on 14/05/2013 12:13:39

[Michael GilliganParticipant @michaelgilligan61133]

Andy,

Many people are getting very satisfactory results using a 200 Step per Rev Stepper Motor directly coupled to the input shaft of a 60:1 or 90:1 Worm driven Rotary Table.

… Sounds like you are ahead of the game already. [assuming you use the Worm drive]

Please keep us informed of progress !!

MichaelG.

 

Edited for clarity.

Edited By Michael Gilligan on 14/05/2013 15:41:59

[Andy from WorkshopshedParticipant @andyfromworkshopshed]

Ah no, I don't have a worm drive but will likely need something similar for fine adjustments as it could be difficult to align the table/chuck accurately otherwise. Will re-read Ivan Law's book to see if there is anything about how accurate things need to be.

[Keith Wardill 1Participant @keithwardill1]

Not sure if its relevant to what you are doing, Andy, but many years ago we experimented with a digital camera made from a static memory chip at the place I worked then (I think we used a 2114 if memory serves).

The idea was that if the plastic covering the chip die on the integrated circuit was removed, then light could reach the 'cells' making up the memory. This was not too difficult to do, simply by rubbing the IC on 'wet & dry' paper, finishing off with metal polish. (probably skim it off with a miill and fine cuts these days).

The controlling processor wrote a logical '0' into a memory location, then read it back. If light hit the cell, then the charge (logic '0' ) leaked away, and it would read as a 1. Conversely if no light fell on it the charge was retained, and read as '0'. The memory is actually an array of these cells, so multiple 'pixels' were available, all individually addressable using the normal memory addressiing. Light could be focussed on the array with a simple lens, producing a 'digital eye'. We found we could distinguish numerals and letters written on cards – the processor also analysed the returned pattern read from the cells – but the experiment was terminated shortly after, and as far as I know, it just remained an interesting experiment.

 

Edited By wotsit on 14/05/2013 17:57:25

[MICHAEL WILLIAMSParticipant @michaelwilliams41215]

Hi Andy ,

If your dividing system has N steps available then the worst pitch error on PCD of anything being worked on is just PI x PCD / N .

For example if you had 1000 steps and were working on a gear with PCD of say 3 inches then :

Worst positional error is 3.14 x 3 / 1000 or about 9.4 / 1000 or approx 10 thou .

That's very poor accuracy . If a real gear was being worked on you would want accuracy of 2 thou at worst and 0.5 thou on a quality job – so you need to increase your step number by 10 to get close . This means more index steps or (very) accurate gearing or more sophisticated detector which allows accurate interploation .

There is of course more to it :

Random indexing has more potential position error than sequential indexing .

Error is influenced by method of detection . Hard to write in a few words but some systems have crisp sharp detects and some have woolly ones .

The number of index steps available is not the whole story – the indexing device itself may have pitching errors and these are just copied onto the work addition to the basic error .

There are favourable and unfavourable numbers of index steps for specific jobs . Ideally whatever you are indexing results in an exact number of index steps per move . In general it won't and you'll have situation where some index moves are different size to others . This problem shows up particularly with a very low number of index steps available – taken to extremes this is like trying to index 8 positions with a 9 step indexer

There's several more pages that could be written – please ask if you have specific questions .

Regards ,

Michael Williams .

[Sub MandrelParticipant @submandrel]

Hi Andy,

I cut all my gears using on my home made rotary table, which reads to 0.1 degree. I have worked out that this give errors within normal manufacturing tolerances assuming reaosnable accuracy of the worm and wheel and the cutter. Half a division (the maximum rounding error) is less than a thou for a 2" diameter gear. As all the gears I have made so far (9 teeth up to 63) have been OK, I assume there are no problems with this.

Wotsit, I have heard of making digital light detectors the way you suggest, but that's the first account I have come across of it actually being done.

Neil

[Andy from WorkshopshedParticipant @andyfromworkshopshed]

Michael, Neil, thanks for the feedback, I'll look at the maths for my size of table which is likely to be in the 3in diameter range. I do like the potentiometer solution for it's electrical simplicity but I'm going to investigate the magnetic encoder chip technique. I'm not the first person to hook one of these things upto a microcontroller so I should be able to work it out. Here's a video of how they work, it's really quite clever, the one I'm looking at has a 14bit resolution which gives at best 0.05° accuracy, not bad for a £9 chip (magnet not included)

[Anonymous]

Andy,

Good luck with the AMS chip. I've used another one from the same family, and it worked well. The application was a focus control for broadcast cameras, so the absolute accuracy was less important than smoothness of operation.

Out of curiosity I've had a read of the AMS5048B datasheet, which claims to be a 14 bit device. As I'm sure you're aware there is a difference between resolution and accuracy. I note that the worst case rms noise is equivalent to nearly 3LSBs, so you will definitely need some external averaging, as mentioned in the datasheet. Assuming the noise is random that should be fairly simple. The datasheet also mentions external linearisation, which I suspect might be more of an issue. To me that implies you have a means of accurately measuring the position of the rotary table in order to correct for systematic errors in the magnetic encoder, which is a bit chicken and egg. The datasheet glibly mentions linearisation and averaging, but doesn't give any details, which is a tad annoying. Unfortunately it's a characteristic of germanic datasheets; logically organised and beautifully produced, but they don't give you much information on how to actually use the device.

As for the required accuracy of course it depends upon the intended use. My dividing head states a maximum error of 1min 30 secs, which I calculate to be 0.025°. I am going to have to cut the final drive gears for my traction engines using my rotary table as, at 14.8" OD, they're too big to fit under the cutter on the horizontal mill if I use the dividing head. If we assume the same error for the rotary table (probably a bit optimistic), 0.025° equates to an error of about 3 thou on a 14.8" circle. Not great, but acceptable for the final drive gears, particularly as the originals would probably have been 'as cast' and deeper than standard to account for the rear axle movement. I'm intending to cut the gears slightly deeper than standard, for what is known as a rattling good fit.

Regards,

Andrew

Addendum: I knew there was something else I was going to type, doooh. I see that the AMS device uses a CORDIC algorithm internally for rotation. Haven't seen that mentioned for many years, but a very clever algorithm, if you don't have a hardware multiplier to hand.

Edited By Andrew Johnston on 15/05/2013 11:50:41

[Andy from WorkshopshedParticipant @andyfromworkshopshed]

Averaging I think I can handle in code fairly easily. To be honest I'd not dug quite so deep into the data sheet as that, I was more concern with what external components were needed and how to communicate with the chip.

In theory compensating for non linearities should also be possible in code with but that would as you mention require some external system to calibrate it against or perhaps a second chip rotated with respect to the first. Perhaps I'll quiz AMS on it once I've got something to test against.

For the time being I'll just get it working. The averaging would be useful for my simpler technique as it will stop the display from flickering so I'll start on that first.

Thanks again for all the feedback and info.

[Michael GilliganParticipant @michaelgilligan61133]

Andy,

That's a very interesting little chip !!

Thanks for the link.

MichaelG.

.

P.S. Just found their page about the DemoKit

Edited By Michael Gilligan on 15/05/2013 14:27:13

[MICHAEL WILLIAMSParticipant @michaelwilliams41215]

(1) I think I'll give that chip a miss .

(2) There is a much better one in the Renishaw range . Expensive for the complete encoder but the sensor board and chip are available separately . Good explanation of how it works and of many other things to do with encoders ..

(3) Thought :

High resolution encoder discs made from CD's – whatever arrangements of codes you want programmed onto disc and read with a standard read head locked in one radial position ?

Standard read head dithering radially over small range and multiple tracks with identical coding ? .

Incremental , absolute or special purpose coding ?

Multiple different codings on different track sets on one disc – selected by setting read head radial position ?

Programmable encoder discs  – a new set of codes for each time of use – each optimised for a particular purpose ?

MikeW

Edited By MICHAEL WILLIAMS on 15/05/2013 19:40:05

[Author]

Posts