High resolution, low cost, quadrature encoders

I can't tell you, off-hand. But I did work on 10000 count/rev optical quadrature encoders for medical infusion pumps (actually writing the software for testing them in real-time at up to 300 rpm while adjusting and aligning them against their cams and then tightening them and gluing them once they met spec.) These included the homing pulse (and more, actually.) So that may be one market area to examine for sources of these units. It wasn't _more than_ 10000 counts, though. So if you are looking for better, I haven't experienced it.

Jon

steps

A question if I may. Why so many steps? Has it to do with how fine the infusion adjustment needs to be?

Jason

I can't say for sure. I was worried about the details of monitoring both the A and B quadrature inputs at rates of 50000 counts/second or roughly every 20us -- but since there could be sloppy +/- 25% variations in A or B timing, I had to be good enough to sample at rates better than 5us or so to be sure to catch everything. Reality was even better than that, luckily.

But I was mired in the details and didn't ask the big questions like "why?" I merely assumed that they had a reason for it.

Jon

The size of the encoder sets limits. There used to be a company near me in the UK (Gaebridge), who made the encoder disks, and would even do 'custom' units. On the 50mm model, they offered up to 36000ppr 'off the shelf', and could go even higher in a larger casing model. The company died a couple of years ago, but models of similar resolution are available from encoders-uk. Two things limit the resolution. The disk size (obvious), and the support accuracy of the shaft. If (for instance), you have a 50mm encoder disk, running full quadrature decoding on a 36000ppr disk (giving potentially 144000 positions), The steps at the edge of the disk, correspond to just 1um, and if the shaft moves laterally by this amount, the count will change. Hence the high accuracy units have very tight specifications on the shaft loadings, and use very expensive bearing designs... Why worry about the source being in Australia?. Most companies will ship anywhere, and may even have agents. Search for the product first, then worry about where it is.

Best Wishes

"Roger Hamlett" schreef in bericht news:Z%3Re.1734$ snipped-for-privacy@newsfe4-win.ntli.net...

Then there are absolute sin/cosine wave encoders, no steps are lost, thanks to their absolute nature.

Here's a nice encoder:

Looks nice, but it really worries me when I see a spec for something like resolution quoted to 4 significant digits accuracy. Implies that the writer knows nothing about what figures mean.

Agilent offers an analog quadrature sensor that can give resolutions in the range of 10k counts/rev. They will require a microprocessor with a pair of A/D converters in addition to a standard quadrature counter system running off a pair of comparators looking at the analog signals. By using the analog outputs, you can interpolate the position to get at least 64 "fractional" parts of a digital count.

If that is more effort than it is worth, the Cannon encoder that another poster mentioned work well. They are what we used to proof our interpolation scheme. They offer resolutions lower than the monster that was mentioned.

Good Luck, Bob

In this case, it may make sense. Say you have an encoder with 99 steps per revolution. That would be ( 360 * 60 / 99 ) = 218.181818181818 arcminutes per step. Who wants to quote resolution as a fraction?

There are 1,296,000 arc seconds per revolution.

The encoder resolves 230,000,000 counts per revolution.

Accumulative accuracy is +/- 1 arc second or +/- 177.46913 counts.

Resolution of 0.005625 arc seconds or 0.998263 counts.

Don't get me wrong. This encoder is a very impressive piece of engineering.

It's just that if you are going to spec a resolution of one count in 230 million in arc seconds to 6 decimal places you should at least get the math right.

The correct answer is 1,296,000 / 230,000,000 or 81 / 14375 exactly, or 0.0056347826086956521739130434782609 approximately.

Check out

But only up ro 12bit 406 bits res.

These guys sell 10,000 count encoders online (in Australia) not sure if you'd consider them cheap though: