Source Sealed Potentiometers?

Reverse the set up a bit, put the magnet on the sensor and just leave the punched pattern on the table. That's essentially how gear tooth detecting hall sensors work, at least some of them incorporate the magnet into the sensor assembly. Building an array of them to achieve the desired resolution I'll leave to you ;)

Robert

[%X------------Robot Car Description----------%X]

A situation he was most careful to explain he wished to avoid.

I had been reading some other links thrown up by this thread which had sector encodings on a single track. This would require an, admittedly, small rotation to read the entire code with the minimal number of sensors (two adjacent patterns required to be certain). Of course, sensor rich solutions might be of interest to the OP if the individual sensors were cheap enough and met his environmental criteria.

One sensor that might fit some of his criteria would be this one:-

However, as it is only IP67, it may just fall a bit short of his intended usage environment. It may also be out of his budget.

They were mentioned here also.

The problem with magnetic fields is that they easily stray into zones where they are not needed. It doesn't take much to divert them. So, I do not see the magnetic encodings being able to be used in as fine a pattern as optical ones. Either the encoding of a distinct number of segments and doing finer position detection by other (absolute) means or coming up with a full magnetic encoding scheme that gives the desired 2 degree resolution, would seem to be a sensible direction for the OP to take.

It is probably the nature of our differences in experience that led to the differing take. Lately, I have only dipped into the newsgroups on a more eratic basis than usual due to being quite occupied with other things. However, now that the OP has explained a bit more about his application it seems that there are plenty of potentially good ideas coming forward and some lively discussion about the merits of each one.

I can imagine that there would be a few interesting applications for such a scheme with the environmental requirements as the OP stated. I hope he finds the most suitable solution for his needs and his wallet.

--
********************************************************************
Paul E. Bennett ....................
Forth based HIDECS Consultancy .....
Mob: +44 (0)7811-639972
Tel: +44 (0)1235-811095
Going Forth Safely ..... EBA. www.electric-boat-association.org.uk..
********************************************************************

--snip--

--snip--

Robert,

By "leave the punched pattern on the table" did you mean not use it? Or that I should attach my oh-so-laboriously punched metal strip to the rim of the platform ("table")?

I've run across descriptions of Hall Effect geartooth detectors, but I never took the time to look at the datasheets. I assumed that they were looking for a change in the magnetic field caused by the presence (or absence) of a tooth. It feels like one could build an absolute encoder with this type of detector by using an irregularly- toothed "gear", that is, by putting gear teeth at appropriate positions around the rim of John Mianowski's (the OP's) platform, but I'd have to do a bit of research on the detectors before I started building one.

Your posting started me thinking (not necessarily a good thing when I have RealWork(tm) to get done ). There are a number of fascinating things about the approach TAOS describes:

1) It's separable. On the one hand there's the pattern and variations on that, and on the other there are the sensors themselves. You could implement this absolute position detection scheme with an in-line 128-element optical sensor as TAOS did, but you could equally well use Hall Effect magnetic sensors, capacitive sensors, or even crusty ol' mechanical MicroSwitches.

Imagine, for example, JohnM's platform with a variable-depth cam pattern around the rim, and a set of those cam-driven MicroSwitches (the ones with with rollers at the end of the switching lever) arranged around the perimeter.

2) It's robust. If someone twirls the encoder while the encoder is powered down, the new resting position will be reported accurately when the system powers back up. 3) It's extensible, at least up to the limits of sensor resolution and placement. 256 divisions not enough? Just get narrower sensors and print more bits/bars. 4) Sensor placement is flexible. TAOS used a single in-line 128-sensor array with all the sensor elements clumped together, but one could also arrange individual sensors at roughly equidistant points around the perimeter of the platform or in other locations as long as one suitably adjusted the encoder pattern. This is a GoodThing(tm), since mounting six or eight of even the smallest MicroSwitches so that their rollers are only 1-2mm apart could get a bit tricky.

If I were going to implement an absolute position reporting system based on TAOS's approach, I'd want to pay particular attention to two things: "transition" points, and the properties of truncated MLSes.

Transition points, those "points between", are always tricky. One of the reason Gray Codes are so handy for absolute encoders is that even if the encoder wheel happens to be midway between "official" positions, at most one bit will be affected. The Maximal Length (bit-)Sequences used by TAOS, on the other hand, will almost always have more than one sensor changing value as the encoder moves from one "official" position and another, so multiple bits will be "uncertain" at the same time. TAOS's 128 sensors solve this problem with redundancy -- multiple sensors per encoder "bit" -- but implementations based on other sensors need to worry about what happens at transition points.

The other "open" topic is whether, if we need to know a rotation angle to within 2 degrees, we can find a 180-bit MLS pattern. It would be really _handy_ to have our encoder pattern "bits" sitting exactly at 2 degree points around the rim of the platform whose position we're monitoring; an angle reported out in increments of (360/180) degrees is much easier to work with than one reported in units of (360/256) degrees.

Unfortunately, the only references I've found so far assume that one only cares about MLS sequences with lengths of ((2^N)-1), that is, one less than a power of two. I'd be _really_ cautious about trusting that I could chop off (say) the last (255-180) or 75 bits from the end of a 255-bit MLS pattern, join the head to the tail, and still have an MLS pattern, but there might be another method for generating a 180-bit MLS.

Aw... you're _too_ kind!

Oh, and just to stir the pot a bit more, in my search for an array of Hall Effect sensors I ran across the Honeywell HMC1512 magnetic sensor:

Honeywell Linear / Angular / Rotary Displacement Sensors

hmc1501-1512.pdf

AppNote AN-211: APPLICATIONS OF MAGNETIC POSITION SENSORS

The HMC1512 reports the _angle_ direction of a magnetic field with a reported range of -90 to +90 degrees and a resolution of less than

5/100 of a degree! Not bad for a sensor chip that Digi-Key sells for $6.45, and a pair of them and a couple of magnets might be all that's needed to satisfy JohnM's requirements (assuming he can solder an SOIC-8 package ).

Ah, well... It's been fun, but I have to go do some RealWork(tm). Sigh.

Frank McKenney, McKenney Associates Richmond, Virginia / (804) 320-4887 Munged E-mail: frank uscore mckenney ayut minds pring dawt cahm (y'all)

-- The joy of research must be found in doing, for every other harvest is uncertain. -- Theobald Smith

--

The latter, sometimes English is just too flexible :)

I've usually used Allegro sensors (linear sensors to measure current) The one time I used a Honeywell sensor it had entirely too much drift. Linear hall sensors in general are rather inaccurate and need calibration before use. I think you'd need a rather gross pattern to use them.

Well, that's no fun!

Robert