Magnetic Rotary Position Sensors

Hello Andy,

I haven't used these sensors, but what is meant by linearisation is that you turn the object whose rotation you're trying to measure one full revolution while measuring the rotation with another device (higher accuracy absolute position encoder, or use a transmission to measure revolutions on another axis).

Ideally, every LSB of the output is exactly 360/(2^14) degrees. Due to sensor errors, or misplacement of your magnet it could be that in one place the LSB size is a bit larger than in other places, giving 360 degrees as a sum of all bits, but differing a little over the rotation (read in on INL and DNL on Wikipedia)

To linearize, you'd need to map the true rotary position to the measured rotary position, for example using look up tables.

Victor,

thanks for the clarification that seems straight forward enough. I'll have to work out how I might go about getting more accurate positioning/measurement. The joy of having a microcontroller in the mix is that I can upload new firmware so to speak late in the project.

Hopefully my primitive test rig will be sufficient to test out the electronics and software, I was a bit limited on non magnetic materials so wood and plastic had to do.

Cheers,

Andy

Bit more progress on the test rig

http://www.workshopshed.com/2013/06/magnetic-sensor-test-rig.html

The ams.com website is down currently so I couldn't see the datasheet, but that's really impressive if they can achieve this accuracy, and if the resolution is high. Glad you're investigating!!! I wonder if that means people don't need to spend a fortune on optical encoders for some high-accuracy applications. For a tool application, it would be interesting to know if motors/cables nearby or any other magnetic fields affect it. Have you noticed any fluctuations in your measurements?

EDIT: just saw the Farnell link. That's very low cost! But, I was unsure, the datasheet on the Farnell site said +-0.5 degree accuracy, is this different to the manufacturer site? I guess with gearing and keeping a count of revolutions, that is still pretty good. But then there is the issue of eliminating backlash I guess :-(

I remember reading some micrometers don't use optical encoders either, but use some accurately spaced parallel PCB tracks, and capacitance. Just as a slight digression,  it doesn't help your use-case directly, but the FRDM-KL25Z boards have a lot of capacitive sensor inputs. I wonder if they can be re-tasked to some measurement purpose. That would be pretty amazing.

shabaz,

the Farnell link is for an earlier board with a lower resolution, it works the same way though. The issues with the higher rez magnetic boards is noise induced in the Hall sensors and non linearity hence the first topic.

I'm working on this in my spare time so I've only just wired up the board. Need to get the I2C communication working next.

However I will see if it is affected by motors and large chunks of steel etc.

The encoders still have their purpose; firstly they go to even higher resolution (2000 pulses per revolution no exception in robotics). Secondly, resolution is not the only concern:

• Accuracy is not 14-bit, as you have 2.7 bit rms noise. Of course you can average, but that will cost speed
• Sampling rate is 11kHz, so in very fast rotating axes the bandwith of measurement is impaired; not so with encoders

Nevertheless, magnetic rotational encoders have a very wide range of applications, and it's cool to measure just by adding a magnet!

A company that has made a very high accuracy motor driver for us told me that they're primarily using IC's from IC-Haus: http://www.ichaus.de/keyword/Magnetic%20Encoder%20iCs.

This company is also interesting: http://www.rls.si/en/am512b-9-bit-rotary-magnetic-sensor-chip--15882#ElementMediaList40407

Would using a metal rig be that bad? I'd guess that linearizing / field mapping would remove the problems you'd get from that. Will you  be moving to metal later on? I'm curious what the effects would be.

I will be moving to metal, the table that I plan to attach this to will be mostly aluminium with a steel spindle.

I don't plan to rotate the table quickly, so the averaging is not too much of an issue. One thing I did realise just yesterday is that I need to be careful about the averaging and crossing zero. So I'll factor that into the algorithm.

I've see the Renishaw chips before but not the IC-Haus ones, thanks for the links.

The Magnet and Hall sensor device arrangement has been used for retrospective feed back on servos like in the open servo project.

In the case of Servos the feedback was normally a standard resistor which obviously got worn out using the hall sensors for feedback got rid of this problem.It may well be worth looking at what they did since it's open source.

Hi John, thanks for the suggestion, 

There's a few interesting links I found via the "Open Servo Project" and there's lots of useful advice to read from there too.

http://hackedgadgets.com/2008/11/19/open-stepper-project/

http://www.jaredharvey.com/

The Open Server forums do discuss hall sensors in detail and also the chips I've been looking at here.  They even have a project called the "Open Encoder" which looks at these chips in detail so many thanks for the mentioning that project

http://www.openservo.com/forums/viewtopic.php?t=766

http://www.openservo.com/forums/viewtopic.php?t=940

Cheers

Andy