Which shaft encoder to use?

[Jeff Waegelin]

I'm currently designing a holonomic drive system, and I want to use shaft encoders to measure wheel speeds. What kind of encoders should I use for this purpose? I tried looking in the DigiKey catalog, but I don't know much of anything about sensors (I'm an ME student), so I couldn't decide which type of encoder I should use.

Also, how does one mount an encoder on a shaft? I know some people use gearing for potentiometers, to increase resolution, but would you do the same thing to an encoder (if it were running at too many RPMs, perhaps)? The shaft I'd be measuring is 1/2", and spinning at 150-470 RPMs, depending on which gear the transmission is in.

Anyways, I'd appreciate any help I can get on either question (which to buy, and how to mount). Thanks!

[Dave Flowerday]

Originally Posted by Jeff Waegelin

What kind of encoders should I use for this purpose?

This year we used 4 Bourns quadrature encoders on our robot and were very pleased with them. The model number is ENA1J-B28-L00064 and they are available from Digikey. They give 64 pulses per revolution (per channel). So, if you decode the quadrature output and count both rising and falling edges on both channels, you get 256 ticks per revolution (but that's tough to do using the RC hardware). We used them geared 1:1 on our drive wheels and the winches for our lift height and tilt. They come in variations with other counts per revolution as well. They're quite pricey at $50 each but they are high quality and durable. These encoders were originally recommended to us by the Thunderchickens and as far as I know they were quite happy with them as well.

Also, make sure to get an optical encoder rather than a mechanical one, and don't get any with detents (detents are spots where the encoder "clicks" into position, often used with knobs on stereo controls or whatnot).

[sanddrag]

A good way to couple it to the shaft is to drill and tap a hole in the shaft for a socket head cap screw, something like #10 would be good if your shaft it big enough. Then, heat shrink the head of the screw to the shaft of the encoder leaving a little space in between. The ridges on the head of the screw and the surface area of the relatively long shaft encoder shaft will prevent slippage. Now you have a very cheap flexible coupling. I would recommend mounting the encoder with a rather thin piece of sheet metal so that has some flex too. Rigidly mounting/coupling encoders can cause them to wear and/or break.

[Andrew Rudolph]

Originally Posted by sanddrag

A good way to couple it to the shaft is to drill and tap a hole in the shaft for a socket head cap screw, something like #10 would be good if your shaft it big enough. Then, heat shrink the head of the screw to the shaft of the encoder leaving a little space in between. The ridges on the head of the screw and the surface area of the relatively long shaft encoder shaft will prevent slippage. Now you have a very cheap flexible coupling. I would recommend mounting the encoder with a rather thin piece of sheet metal so that has some flex too. Rigidly mounting/coupling encoders can cause them to wear and/or break.

We did this in 2004, and had some problems with it. The biggest being since your putting the wheels out at angles on the edges it leaves little room to mount the encoder, if you cant get the encoder inline with the shaft then it doesnt work very well, also the tubing we used to make the couplers would load sometimes when we changed directions. Also, since they were mounted outside the wheel they were very prone to breaking in a hit, even with lexan convering.

This year we devised a much better plan that gave us no issues except when the encoders were shimmed wrong and hit the rollers on the wheel and snapped the encoder shaft. We bought small timing belt and pullies from sdp-si and mounted the encoder to the side of the gearbox. You can see it in this picture, near the middle.
I dont remeber any of the info on the encoders, they were quadrature and cost like 25 dollars each i think.

[Stephen Kowski]

Originally Posted by sanddrag

Now you have a very cheap flexible coupling. I would recommend mounting the encoder with a rather thin piece of sheet metal so that has some flex too.

If you are going to use an actual coupling to isolate the encoders try the "Helical Beam Couplings" from mcmaster. may cost a little more (about $20 a pop), but will isolate your encoders from flexing and vibrations. That way if you get an expensive encoder you can be assured it won't be destroyed or give bad readings.

[Gdeaver]

I'll throw this out as something completely different. A good encoder will probably cost 50 to 60$. The problem with encoders is that they require the use of interrupts in programming. Two encoders isn't too bad. 4 gets a little tighter to implement. Add a couple timer routines and things get complicated. I've just started playing around with a hacked optical mouse as a alternative to an encoder. Basically an optical mouse can track movement of a surface. The Aligent 2610 chip that I'm using can track up to 10 inches per second, however I've found its better to keep the disk that the mouse is sensing to 6-8 inches per second max. Using 2 wire serial communications the mouse returns the number of counts since the last request. The 2610 resolution is 400 dpi. That's a high resolution non-contact tachometer for about 12$ and no interrupts to program and debug. The chip does the counting for you, just poll it often enough that the registers don't overflow. I've only just started playing but it looks good. There were some threads on here about optical mice. Note that the Aligent mice chips are optimized for an almond colored texture Formica surface. The mouse lens needs to be 2mm from the surface. At higher velocities it's critical. I'm looking to use it to track and log valve linkage movement.

[Jeff Waegelin]

Originally Posted by Dave Flowerday

This year we used 4 Bourns quadrature encoders on our robot and were very pleased with them. The model number is ENA1J-B28-L00064 and they are available from Digikey. They give 64 pulses per revolution (per channel). So, if you decode the quadrature output and count both rising and falling edges on both channels, you get 256 ticks per revolution (but that's tough to do using the RC hardware). We used them geared 1:1 on our drive wheels and the winches for our lift height and tilt. They come in variations with other counts per revolution as well. They're quite pricey at $50 each but they are high quality and durable. These encoders were originally recommended to us by the Thunderchickens and as far as I know they were quite happy with them as well.

Also, make sure to get an optical encoder rather than a mechanical one, and don't get any with detents (detents are spots where the encoder "clicks" into position, often used with knobs on stereo controls or whatnot).

I don't have a huge budget for sensors for my summer project. Is there something less expensive you (or someone else) could recommend? Or will I be disappointed with the performance from lesser components?

Also, what did you use to attach the encoders with your 1:1 gearing?

[Al Skierkiewicz]

Originally Posted by Jeff Waegelin

I don't have a huge budget for sensors for my summer project. Is there something less expensive you (or someone else) could recommend? Or will I be disappointed with the performance from lesser components?

Also, what did you use to attach the encoders with your 1:1 gearing?

Jeff,
The encoders look exactly like pots and have all the standard hardware and mounting bushing attached. The shaft is 1/4" dia. and about 1/2" long so a shaft coupler works well. If you are looking for something cheap and dirty, Digikey does sell some optical couplers that you could incorporate into a wheel encoder of your own. All you need to do is design a interupting wheel that can be attached to whatever shaft you are trying to count and position the opto coupler so that the wheel breaks the light path. This is the same design that a computer mouse uses. Open one up and you will get an idea how it all goes together, keeping in mind that the mouse is a higher resolution than you will need. Reflective opto couplers also work well. We have used the Banner sensors looking at alternating white and black patches on our wheels for many years.

[Jeff Waegelin]

Originally Posted by Al Skierkiewicz

We have used the Banner sensors looking at alternating white and black patches on our wheels for many years.

How has your experience with the Banner sensors been? Do you have any advice on how to make them work most effectively? They seem like a pretty cheap and simple solution, especially considering the massive quantities we have from past kits. Given the cost differential compared to encoders (free, vs. $40-$50 apiece), would you recommend I try the Banner sensors for my summer project?

[Dave Flowerday]

Originally Posted by Jeff Waegelin

How has your experience with the Banner sensors been? Do you have any advice on how to make them work most effectively? They seem like a pretty cheap and simple solution, especially considering the massive quantities we have from past kits. Given the cost differential compared to encoders (free, vs. $40-$50 apiece), would you recommend I try the Banner sensors for my summer project?

The Banner sensors worked great, however there were some reasons we abandoned them. First, setting up an encoder based on Banner sensors takes up a lot of space (if you need to be able to measure direction as well as rotation, then you'll need 2 sensors per wheel). It's also important to note that different versions of the Banner sensors were included in the kit each year. There's 2 types: one that is designed for detecting retro-reflective tape, and the other which works for things like line following (looking at white tape, much closer to the sensor). To be honest though I don't think we had much trouble using either model with our wheel encoders. If you're using 2 sensors per wheel (to be able to determine direction), then you need to mount them carefully as the square waves generated by each encoder need to be 90 degrees out of phase. We used a pinwheel of alternating black and aluminum "pie slices" with the Banner sensor mounted about an inch away. Here's the best picture I could find that shows our Banner sensor wheel from 2003 and 2004:

[Al Skierkiewicz]

Thanks for the pic, Dave.
Jeff, it is easy to see from the picture the rough resolution of this type of sensor. The wheel is about 6" dia.

[Andrew Rudolph]

I know that our programmers found that with holonomic drive the higher resolution you have the better things work. And it takes quite a while to get 4 encoders running and not crashing the R/C

[Jeff Waegelin]

I was looking at the Digikey catalog, and I found something that might work. It's made by Grayhill, and is part number 61C11-01-08-02, found on page 1262 of the Digikey catalog. Would these encoders work for my purpose? Bear in mind that this is a summer prototype project, so it doesn't necessarily have to go through a full season of use and abuse. If they aren't up to it, I can spend the extra money on the Bourns encoders, but I'd prefer not to, if don't need to.

[Mark McLeod]

Originally Posted by Jeff Waegelin

... Grayhill, and is part number 61C11-01-08-02, found on page 1262 of the Digikey catalog. Would these encoders work for my purpose?

It's rated for human hand turning speed. You can't spin the encoder shaft very fast.
So the question becomes how fast (rpm) do you plan on gearing it?

[Al Skierkiewicz]

Jeff,
Go to the Grayhill site to be sure, but many of these encoders are not continuous rotation. The flag for me was the spec that an operational cycle was a full rotation through all positions and a return to zero. Since one application is for frequency set on a radio this may be only a 250-270 degree rotation with 32 detents. When checking devices be sure to find the spec on the maximum allowable rotational speed as well.