Position Encoding.

You are correct, I slipped a digit. It should be x^4 + x + 1 (and the first time I typed that it had x^5 again!!!).

I have written mod-2 polynomial arithmetic packages, including Sieves of Eratosthenes to find primitive polynomials, at least four times that I can distinctly remember, and probably more often than that.

I may even have one or two of them around, still.

Why give Slowman the time of the day? He's vociferously clueless. ...Jim Thompson

If you're implementing one of those in logic, you probably want to count zeros, and when you get to 2^N-1 of them, then hold off the LFSR for one cycle and insert a zero into the stream (and if you're paranoid about cosmic rays zeroing out your LFSR, insert the right pattern into it at that point, to boot).

That takes less logic than a ginormous zero-detector, or at least it does in an FPGA.

the

can

te

if

F.

ve

th

of

Now that's irony for you.

-- Bill Sloman, Nijmegen

Tim,

You're just having way too much fun!

Fred

Maybe that's why I don't like these cross-posted threads. Most of us don't know how to filter what we're reading because we don't know those other folks.

Fred

It would be fairly easy to do with a microprocessor, too -- they generally come with oodles of ROM these days. My two biggest concerns are that I would much rather do it with something off the shelf, and the TAOS parts look like they are not in any standard package, and may be a bit of a production headache.

Thanks Syd. That would be perfect -- except that I failed to mention that the shaft I need to get the position of goes all the way through the machine.

My bad -- I certainly should have said.

Well, theoretical fun. I really want to find a solution; the off the shelf ones I've found are spendy, and I'd like to not have a "roll your own" one unless there's not a good alternative.

pdf

I was answering the "Have you seen this, has it been done, how do you do it?" part of your post.

A COTS version would just be called an "absolute encoder." Pretty pricey. So, I thought the how-to info might be handy. They even give codewheel-printing PostScript source code.

usdigital.com is a decent place for encoder-shopping, both parts and complete units. TAOS and Avagotech.com make the optical sensors.

-- Cheers, James Arthur

And I do appreciate the effort!

I know. You can get incremental encoder chips from Avago, with directions on making the encoder strip or wheel or whatever (at least you could -- I'm not sure you still can). But I didn't see anything on the Avago web site about chips for absolute encoders. Maybe I'll look again...

I'll look. I was hoping for something between raw components and encoders that are stout enough for a gorilla to climb on -- I guess it'll be roll-my-own.

Well, I tried...

(snip, I wrote)

The design I remember needs a clock faster than the fastest possible actual transition. (Not counting going back and forth, bouncing across one edge.) I suppose many processors are fast enough to do an input loop fast enough to do that, but it is much easier with a clock and FF's in an FPGA.

The tiniest FPGA should do it if the BRAM is big enough.

-- glen

We boogered one of our VME waveform generators to simulate motion dynamics from a 8 million count encoder. I asked the customer what it was for and got the usual "we'd have to kill you if we told you" answer.

John

the

y a

he

ght

ve

Impressive. And, did they kill you?

Grins, James

(snip)

(snip)

This has complications if the bits don't change at the same time, and they never do. You need some way to know where the bits are centered under the detectors.

One way is to have a set of detectors half way in between, or somewhere around the track such that at least one changes state half way between the spot where the detector bits change. You then need to debounce that signal appropriately, latch the detector bits, and apply them to the look-up table. You can also compare to the previous bits to detect transitions.

You do have to take care of very fast bounce noise on the data lines if they are sitting right on a transition.

Reminds me, there was discussion in the FPGA groups some time ago about an effect which is presumably true for all asynchronous SRAMs, but seems to have been mostly noticed in FPGAs. If you have the wrong kind of noise on the address inputs of an SRAM in read mode, it is possible to change the stored bits. Consider that the address bits select which cell to connect to the read lines, and switching fast enough could cause two cells to be selected, during the switch time, long enough or cummulatively enough to flip the cell. That could happen for a look-up table on asynchronous optical encoder data.

-- glen

d,

DSP,

er.

e/

t any

and

se

uy

ve

There are degrees of difficulty here; quadrature detectors can be made to work without any optics, while a detector capable of extracting single-bit information from a pseudo-random binary sequence encoded strip probably needs a lens to focus the light source onto the strip sufficiently tightly that the illuminated spot is not bigger than the narrowest stripe on the strip, and may need a second lens to concentrate the light that got through onto the detector.

An eight-point detector ideally needs eight closely spaced detectors looking at a well focussed image of the - moving - encoded strip, with a light source that provides relatively uniform illumination at all eight detectors.

The mechanical design gets bulkier - and more difficult to get aligned and keep aligned - as the optics get more demanding. Tim claims to be looking for something a gorilla can put together.

-- Bill Sloman, Nijmegen

SP,

r.

any

nd

ttdesign.com

How about a Potentiometer with continual 360 degrees rotation or in your optical vein a disk with a gradient of clear to dark in 360 degrees. Intensity =3D position.

regards, al

If it simplifies the construction, the detector can sook at every nth line rather than adjacent ones. Tom wanted a device a gorilla could sit on.

Jerry