Source Sealed Potentiometers?

there

30

I could mount the sensors fixed, & the "sensed" material on the rotating platform.

No, I think you got it the 1st time. I don't have the room to connect a resolver so that 1 turn of the platform = 1 turn of the resolver shaft; I have to gear it down & can make 5:1 work. Trouble is, there would be 5 points in the turntable's rotation where the resolver produces the exact same output, & I need a way of knowing which of them is correct.

MIght be something there - I'll have to get some & see what I can make happen.

Not just miss a step, but know exactly where the turntable is pointed on power-up.

Yeah, but I expect that more often than not, when it powers up, it's not going to be triggering any switch, so I won't know anything about the initial position.

JM

there

30

I could mount the sensors fixed, & the "sensed" material on the rotating platform.

No, I think you got it the 1st time. I don't have the room to connect a resolver so that 1 turn of the platform = 1 turn of the resolver shaft; I have to gear it down & can make 5:1 work. Trouble is, there would be 5 points in the turntable's rotation where the resolver produces the exact same output, & I need a way of knowing which of them is correct.

MIght be something there - I'll have to get some & see what I can make happen.

Not just miss a step, but know exactly where the turntable is pointed on power-up.

Yeah, but I expect that more often than not, when it powers up, it's not going to be triggering any switch, so I won't know anything about the initial position.

JM

That's how I figure it, too. The only real trouble is that most solutions are getting a bit complex. My original intent was to just use a 5-turn potentiometer, but the need to immerse in water makes that tough. 5-turn pots that can survive immersion seem to be pretty expensive, from what I can find, at least. Best I've been able to find is about US$80.

Thanks to all who have responded! It's been an interesting discussion

- well beyond what I originally expected!

JM

to

there

to 30

I think I lost part of the picture here. If your "resolver wheel" makes five turns for the one full turn of the platform, doesn't _every_ reported resolver output represent five possibple platform positions?

But... why do you need a separate resolver? If you were using, for example, optoreflector sensors, why not just paint a black and white "absolute position" pattern on the bottom of the _platform_? It would cut down on the number of moving parts.

Great. The AllegroMicro has a few tutorials as well, and more material can be found scattered around the 'web.

That, too. I hadn't thought about that part. Let us know what you finally come up with. We'll be more than happy to steal your solution and an use it in our own projects.

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

-- "We are what we repeatedly do. Excellence, then, is not an art, but a habit." -- Aristotle

--

up to

there

to 30

that

Exactly, & I need to narrow it down to just 1.

That was part of somebody's proposed solution, to figure out which of the 5 possible positions is actual.

I've done that & it works very well in prototype. I'm not confident, though, that it will work reliably in the messy environment that the final version will be going into. The need for frequent maintenance/cleaning is something that I'd like to avoid if I can.

Sure. You're all more than welcome to it. Whatever "the solution" ends up being, probably won't be for awhile (it's hobby-related, subject to the parameters of funding & time that all hobbies must exist within - nothing planned to ever be sold).

Thanks!

JM

--snip--

Ah. Sorry, thought that was you -- I got lost in the quoting.

So the idea was to divide up the job? A 1/5 absolute encoder on the platform _combined_ with (say) a 1/36 absolute encoder on the secondary "resolver"? Makes sense.

You mentioned the possibility of the platform getting "dunked", which suggests outdoors: river water, pond water, rainwater. That kind of exposure to the elements suggests that the platform and its drive mechanism will need periodic maintenance, and any type of resolver or encoder that is mechanically connected to that platform or its drive will need at least as much maintenance (gears need greasing from time to time ). Given that, will adding mechanical, optical, or magnetic sensors onto the platform require more?

I dimly recall that the platform is, what, three inches in diameter? Putting an optical pattern on its uderside would mean... 180 divisions, so eight binary sensor "bits" to track -- an 180x8 grid, with fairly closely-spaced "pickups".

--snip--

It's too bad someone in Congress didn't sneak an "earmark" onto one of those last-minute bills for an American Hobby Reaserch and Development Grant Fund. Maybe next year...

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

-- The mere formulation of a problem is far more often essential than its solution, which may be merely a matter of mathematical or experimental skill. To raise new questions, new possibilities, to regard old problems from a new angle requires creative imagination and and marks real advances in science. -- Albert Einstein

--

After my last posting I took a walk and thought a bit about encoders and encoder patterns. "Absolute" optical encoder wheel patterns have been around for years; the ones I've seen are basically XxY binary arrays drawn as concentric circular segments, with the bit patterns generally based on Grey Codes or something similar.

If I take such an XxY array and string it out as a X*Y-length binary pattern "track", and if I use a second track as an "index marker" every Y "bits", I've transformed the M-track absolute position pattern into a two-track pattern.

Could I do it all with one track? What would I need?

- A "magic" bit pattern P of length N. - A set of sensors -- M of them. - P chosen so that every M-length "wrapped" substring of P was unique.

Well, I can see one trivial solution:

M=N and P=100...000: The position of the '1' bit will be different for each "substring" (remember, wrapping is allowed).

If I'll settle for only every Mth substring being unique, then I can use the "ravelled" absolute position pattern described above.

But there're these two _teeny_ details: For a given N, I need to find M, preferably as small as possible, and I need to find a suitable P. Clumsily put into math-ish notation, I could rephrase this as:

Find a bitstring P of length N [p1,p2,p3,...,pN] and number M

Something like this?

Robert

Which refers to:

Jon

Selection of the most appropriate materials that will survive the immersion in the intended environment is part of the design process. I know of one person who uses a reinforced nylon gear train for the final gearing of an electric boat propulsion system. No greasing required. I am sure that the OP has that anglecovered though.

--
********************************************************************
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..
********************************************************************

...

Rotation speed will be very low, & the rotating assembly is supported on a bearing, so loading on the gears is pretty low. Most plastic gears are self-lubricating & maintenance is expected to be little more than cleaning. Gears in need of maintenance typically start getting a little noisy but still function for a good while. Optical sensors, OTOH, tend to fail completely & suddenly - such as from dirt on a sensor lens or a white reflective surface. Magnetic, I would expect little trouble from.

Yes.

Or around the edge, where I have about 0.5" to work with. This is where I painted my black/white stripes on the 2nd prototype (1/2 black,

1/2 white to tell direction to "home"; narrow white on black to indicate "home"; stepper to rotate, count/accumulate steps to determine position) - which worked just fine, except that I'm not 100% satisfied that it can handle the dirty environment or continue to keep track of position when/if an external force rotates the turntable.

They do, & they'll get my vote! Like everybody else, I'm opposed to all pork-barrel spending except that which directly benefits me! ;-)

I hadn't seen that quote before, but I love it!!

JM

The essential benefit of absolute encoders is the fact that you can read position at any time without any rotation of the encoder. Any serialization of the track data such as you have proposed requires some turning of the encoder to read position. One benefit you can achieve using the plan you suggest is that you can achieve a synchronization at many places in a single rotation. A standard incremental encoder provides synchronization at just once place in a rotation. So your plan may be an improvement over an incremental encoder, but is it no absolute encoder.

Robert Scott Ypsilanti, Michigan

to

That depends on what you use for a sensor. The link I originally posted was from Taos who were proposing using one of their linear photo sensor arrays to read enough of the pattern at once to get an absolute position.

Whether replacing multiple patterns with multiple sensors is an advance is another question.

Even using a single sensor though I can see applications where a little jitter at startup to determine the startup position might be an acceptable trade-off for not needing to move to a home position.

Robert

Yes, although they seem to have done a much better job on the problem than I did.

For those who haven't already downloaded this file, it's an App Note published by TAOS (of color sensor fame) entitled "Design Considerations for an Absolute Steering Angle Encoder Demo". The AN describes a method for using a maximal-length sequence code to generate an optical encoder pattern which can be used with a linear array sensor to determine the encoder wheel's position.

The authors have started with the same problem I described a few postings back, but they've clearly thought through not only the mathematics, but the real-world considerations of such an approach as well... such as the problem of ensuring continuous readout values during transitions, and the "bar" width vs. the sensor size and separation between sensors.

I'm still working my way through some of the paper's calculations, and I think I'm going to need to dig into the datasheet on TAOS's TSL1401R sensor before I can get much further.

Anyway, it's nice to see their work, and I found some other interesting App Notes on the TAOS site.

Thanks for the pointer, Robert.

Frank

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

-- "I've wrestled with reality for thirty-five years, and I'm happy to state that I've finally won out over it." -- Elwood P. Dowd in "Harvey"

Thanks, Jon. This paper -- "Position Detection Using Maximal Length Sequences" -- describes the basic concept behind the encoder pattern and its use. They're a good combination. I missed it when I went through the TAOS site, so I appreciate your pointing it out.

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

-- The aim of science is to seek the simplest explanations of complex facts. Seek simplicity and distrust it. -- A. N. Whitehead.

--

Robert,

_Any_ serialization? Or just my two-track approach?

Robert,

Third attempt (first two deleted before posting ): You're right. The simple-minded two-track pattern I described does provide for absolute positions... but only at the "index marker" points on the second track. A "true" (as in, "usable" ) absolute position pattern would report positions at _every_ point of the encoder's movement. It's true that the absolute position encoder wheels have I've seen have their "blind spots", but they're relatively tiny transitions, where mine is full of "holes".

My only excuse for not thinking that example through more carefully is that it wasn't my intended end-point. I was throwing it out as a visual transition point to my real question, which was whether _any_ single-track pattern could be an absolute position pattern.

I then answered with a "yes" (one of the results of talking to one's self too often) and I described a method for doing this with N sensors when N unique positions were neeeded. This is... um, a "sensor intensive" approach , so the next question was whether one could do it better. Since N positions need to be known, the lower bound would be ceiling(Log2(N)) sensors -- you have to be able to _report_ N unique values -- but how close to this can one theoretically come?

The main TAOS paper uses (if I read it correctly) 14 sensor bits to track 256 positions. That's an M

--snip--

Less trouble in terms of getting silt deposits, but it feels like adding lots of tiny magnets (or the equivalent of gear teeth) could be more work than simply printing an optical-encoder pattern.

Well, optical or magnetic, you might check the parallel thread in this newsgroup titled "Minimal encoder patterns" regarding doing absolute position sensing with a one-track pattern -- sort of like a linear or one-dimensional barcode vs. a complex X-Y pattern like a 2-D barcode.

Thank you. I try to only steal from the best.

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

-- Never worry about theory as long as the machinery does what it's supposed to do. -- R. A. Heinlein

--

Yes, I unabashedly admit it.

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

I recall that the OP wanted to be certain of the turntable position without the need of any prior rotation. A serial encoding track would not fit this bill.

--
********************************************************************
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..
********************************************************************