Position Encoding.

It's always nice to reconsider old problems with the hardware of RECENT years, though. Instead of paying high prices for low-shaft-runout gizmos to get high resolution, you could make the rotary part a sin(k * x) + sin(k_2 * y) transparency, and make the sensor into an image sensor. Then a 2-d FFT will give you the direction of those oscillations with no dependence whatever on (for instance) shaft runout.

It's far from clear that the old engineering solution is always the best developed; we keep on changing the development environment. Pots, binary-code disks, Gray-code disks, ... what comes next?

an

approach to

for most

incremental

to use an

absolute

at

unfounded.

supply a

and sow the

here,

something

participating

day.

there

two of

Nothing to do with LVDTs or synchros. And, as usual, it never got built. At 200 pages, it probably collapsed of its own weight.

PADS will export/import *anything* as ASCII files. That can be very handy. I write PowerBasic programs to crunch netlists, and other guys here do Perl and Python ditto. When you are using 750-ball FPGAs, retyping the pinouts from the PCB layout into the FPGA suite gets tedious.

Sounds like a heap of traditional logic could have been replaced with an FPGA in 1998. Nowadays, even low end FPGAs do good stuff at 250 MHz clock rates.

John

Oh, he's hopeless. He just hangs out here to have targets for his pathetic, ponderous insults. He hasn't designed electronics in a decade or so, and most of the stuff he did before than was a failure. His Baxandall oscillator project has become a sick joke. What a useless drain on society!

What's cool is how quick and easy it is to design stuff these days. There are so many wonderful parts and tools around. I love FPGAs, because I can push huge amounts of complexity into a chip, and get on with the specifications and manuals and PCB design, and pipeline that with the FPGA design. Of course, I get other people to do the actual FPGA design, but that's wonderful too. If I need 24 digital filters, or a few dozen multipliers, or a programmable delay line, it's literally a few lines of VHDL.

John

You call it that, only because JL refuses to spoon feed you experienced information to fill your sock for the year.

Jamie

Oh really, you put yourself in yet, another trap state. We were not at any point talking about Q encoders.

You should be more diligent and do your home work before hitting the enter key. It only serves to expose yourself.

Jamie

The high-res encoders, as in 8 million counts/rev, pick up the "analog" sin/cos signals, digitize, and interpolate to a small fraction of the encoder disk pitch.

All sorts of stuff is possible if you can throw a massive amount of DSP and calibration at it.

John

If you don't mind being limited to shaft speeds less than about 2 rpm, you could probably do something like that. Image sensors are _slow_, remember. The best way to get rid of runout error is to use three moire' sensors oriented at 120 degrees to each other.

(It's an old saw of mine that a software guy's idea of how to do anything optical is to point a webcam at it and crunch the daylights out of the video stream. Works maybe 2% of the time.)

Cheers

Phil Hobbs

Sounds like a great idea for an encoder. A disk with some complex, goofy pattern and a 10-megapixel imaging chip. Software corrects for everything!

John

There are applications where that could work okay, but it the rotation would have to be very slow. You'd get aliasing at ridiculously low shaft speeds.

Cheers

Phil Hobbs

ple of an

approach to

pensive for most

incremental

self.

e able to use an

ut the absolute

raying at

ttack unfounded.

d to supply a

ncor and sow the

rtals here,

something

participating

e one day.

a lot there

only two of

s as if

I'm not

own

hat has

t have

so I'm

fits

t we

y as

tant as

our

et

papers

to

out

t of

e.

th

stem

ted, and

ler

lve

oo,

It's

at

ic

d to

ast

Obviously - but it was moderately complicated.

It would have been fine. The architecture wasn't so very different from the timing board in the Cambridge Instruments electron beam tester, which we did build, and did work - the circuit diagram for that was also heirachical, but we worked with A2 sheets (mostly printed out reduced on A3 sheets, though we did pen-plot the released version onto A2 drafting film). The board would have been a six layer triple extended Eurocard, wut a roughly 50% fill factor, which is easy with surface mount components. I don't think we would even have need to bury any of the buses.

As I've mentioned, the customer had to stop the project when he found out that he wasn't going to get any more graduate students to use it. It is theoretically possible that we couldn't have gotten it to work once we'd put it together, but if that had happened it would have been a first for me.

We could do that with the prehistoric Metheus UNIX-based system that we had at Cambridge Instruments - I only did it once or twice, and only fiddled with positions on the drawing, but the guys who were primarily busy with schematic creation and capture used it a lot more frequently.

The TTL side of the board was initially roughed out with more or less standard Fast TTL and ACT, but we eventually got the okay to use some specific Xilinx parts, which we were confident that we could program. I'd reworked the circuit diagrams around them about a year before the project got canned. They weren't fast enough for the ECL side of the board, and while other manufacturers claimed to have stuff that would have been quick enough and could drive transmission lines, we weren't exactly confident that they'd still be in business when we'd be given the okay to order parts.

The ECL was going to be clocked at 500MHz - the ECLinPS synchronous counters could then be clocked that fast, I could get a chemically thinned 500MHz crystal from at least two suppliers, and the MC100E195 had a guaranteed 2nsec or more delay range with roughly 20psec resolution. Today's MC100EP195 will do at least 2.2 to 12.2 nsec with roughly 10psec resolution, which would let you realise the architecture around a 100MHz clock if you felt like it, but I was designing around what I could get baack then.

The Cambridge Instruments electron beam tester had used an 800MHz clock with Gigabit Logic's GaAs synchronous counters, which worked fine but the decision to go with that approach was pretty silly. I'd originally proposed the GaAs-based scheme as a way of making it obvious how stupid the marking insistence on 10psec granularity actually was, but management fell in love with the idea of developing a system that would be hard for the competition to catch up with.

-- Bill Sloman, Nijmegen

-- Bill Sloman, Nijmegen

Sixty posts further down the thread - post 173 on the goggle list at the moment, which is only 25 posts from the end, you confidently posted an erroneous statement on the subject dated Jan 1, 3:58am (Amsterdam time). You seem to have a short memory.

he

It's your own homework that seems to be defective. Try to remember whatever it was that you imagined a few hous earlier.

-- Bill Sloman, Nijmegen

to

or most

tal

use an

olute

unded.

y a

ow the

e,

ating

ere

of

and get

Happily I did get educated by real, knowledgeable, engineers once I got into industry, but I was a very quick study, and never needed to bullshit anybody. The real stuff was quite impressive enough. It's a pity Jamie is too dim to appreciate this.

-- Bill Sloman, Nijmegen

most

an

unfounded.

the

John Larkin posts lots of off-topic nonsense, and hates being told that he's posting nonsense, which he experiences as being isulted rather than being educated.

I have designed some electronics since I lost my last job, but I haven't put any of it together - it's an expensive hobby, and I rarely buy components until I'm pretty sure that I'm going to use them. That said, I've got 15 Xilinx XCR-3032XL-10PCG44C Coolrunner PLD's in my desk drawer which I'd really like to use ...

Not a lot of the stuff that I designed before then got into production

- some of it got canned before we'd even laid out the first boards, which was frustrating, but the Cambridge Instruments electron beam tester with the GaAs logic and the ECL static RAM was a realised as a working protoytpe some six months before the project was cancelled - there was no doubt that we could have sold a dozen or so machines, but that wasn't going to generate enough cash flow to keep the accountants happy, so they threw away a couple of million dollars worth of development effort.

I understand the sunk-cost-fallacy - it was going to cost another couple of million dollars to build that first production batch - but we had a machine which collected data a lot quicker than the competing machines and was consequently a lot easier to use. We'd have ended up selling quite a few more than twelve (at about half a million dollars each).

It was a rather larger scale development than John Larkin gets involved in, and he doesn't really appreciate what was going on.

It's certainly not making any progress at the moment.

I'm 69 and retired - much against my will. Everybody who is retired is a useless drain on society, which makes my situation just that bit more irritating.

Particularly if you design the kind of simple and undemanding stuff that John Larkin peddles

I had fun with ICT's PA7024 back in the 1990's. With only twenty cells you couldn't push that much complexity into a single chip, but you could do quite a bit. I got into using them because they were a plug- in replacement for a 22V10, and proved a life-saver when I had to rescue somebody else's design where they hadn't paid attention the

One shows up in my Peltier-based thermostat paper

The 32-cell Xilinx XCR-3032XL-10PCG44C promises to be quite a bit more useful, if I ever get my act sufficiently together to actually build something.

I wonder why? It's not that complicated - a bit messier than writing assembler, but less messy than actually hooking up actual digital chips.

If the VHDL is written by somebody who knows what they are doing - probably not John Larkin, from what he says about getting other people to do the actual FPGA design.

-- Bill Sloman, Nijmegen

Presumably Jamie has had a personal message from the diety involved, or imagines that he has had such a message. Jamie's imagination does seem to lead him astray rather frequently.

al

Jamie is letting his imagination run away with him. He's got the idea that it is "practical" to to use two separate tracks for a in-phase and quadrature clock generation, when most of us realised that you could do it with one a few years ago - back in 1992 in my case.

That kind of dogmatic ignorance can waste a great deal of money in development.

As if Jamie has any idea of what the sane and sensible people on this planet might be doing.

I'm afraid that it's your trousers that have burst into flame. It's not something that your imagaination would dream up, so you don't seem to have noticed yet, but even you will wake up eventually.

A slightly repetitive rhetorical device, but Jamie is the last person you'd expect to produce something balanced.

The vibrational and rotational states of the nitrosyl bromide molecule are quantised, which is why the infra-red spectrum consists of a series of absorbtion lines

If you write down the bond-strengths, atomics masses and the bond-to- bond interaction constants for the molecule you can set up the molecular partition function, and predict the infra-red spectrum as a function of these numbers - though in practice you use the infra-red spectrum to infer the bond-strengths and bond-to-bond interaction constants.

What's interesting is that if you want to work out the rate constant for the gas-phase reaction

you can work out the partition function for the transition state molecule O-N-Br-Br-N-O, using what you know about NOBr, NO and Br2, and that lets you generate quite a good theoretical prediction of the gas phase reaction rate.

If you were to read a copy of my Ph.D. thesis - available at the Melbourne University Library - you will find precisely this analysis.

The thesis is mostly about measuring the reaction rate, rather than predicting it, but that calculation gave a pretty clear indication of how the chemical reaction actually worked.

You'd be a bit too ignorant to appreciate the point, but it may entertain more sophisticated lurkers - though your name would probably be enough to put them off reading this far.

-- Bill Sloman, Nijmegen

You don't use a spell checker, do you?

Got to you didn't he? I can tell because you changed the subject line. Your pride is an open book.

Please start using a spell checker. Your occasional display of mental pyrotechnics needs it.

On the other hand we'd have missed out on Jamie being much more stupid than usual - which is to say being very stupid indeed.

He could be a republican presidential candidate ...

-- Bill Sloman, Nijmegen