[kong]

Following on from this thread I have thinking about building a DRO based on rotary encoders and the software available over here.
The biggest problem I can see is how to mount the encoders to the axis (assuming 3 axis knee mill for now). I have read about using fishing line to rotate the encoder, and someone even used a mouse mounted to an axis and used the mouse' own encoders for the DRO.
What I have in mind would be similar, I am thinking along the lines of a rubber wheel attatched to the encoder, and having the movement of the tables turn the wheel, therefore rotating the encoder. Obviously there are some issues with swarf or the rubber wheel not gripping efectively (either not enough grip or too much to damage the encoder). But what do you guys think?

[lerman]

I'm working on a similar project using linear encoders (www.usdigital.com). The 500 CPI (2000 counts per inch) encoders are $4 per inch plus $28 (I think) for the head.

When I called usdigital to ask about the precision of the encoder strips, I was told that they are as exact as you could hope to measure them. Of course, the material you mount them on is going to change size with temperature, but that is well within the precision I'm interested in.

I've purchased a 7" strip that I will use for my prototype. So far, I've got most of the components machined, but am no very happy with the polypropylene (that I had lying around) for the slide. I'll probably order some delrin and try that.

For a display, I have a 4 digit up/down counter (K129) that I ordered from http://kitsrus.com. That was around $25. It uses an Atmel processor and I plan to write code for it that will convert it from an up/down counter to a quadrature counter. Four digits will be fine for my 7 inch axis. After I have it working, I'll build my own board with 5-1/2 digits.

The slide will be enclosed in a track made of 1/4 inch aluminum and travel on 1/2 inch stainless rod. It measures 2-1/2 by 1-1/2 by 10 inches (the track length plus 3). This is probably overkill for my 7 inch encoder, but I think will be very nice for the larger ones needed for the X and Y axes. (The linear encoders are available in lengths up to 34 inches.)

Ken

[H500]

I think a rubber wheel won't be accurate enough because the diameter would change with pressure. That will change your counts per inch.

Take a look at the rollers on a band saw. When clamped hard enough, the rollers will move with the blade without sliding.
It would be a simple matter to use the same principle to rotate a 3/8" shaft with a disk on it.

I still favor the use of an optical mouse sensor for the pickup because of the high resolution and the fact that the disk can be made of any non-mirror material. There is no need for a slot pattern.

Instead of using a PC, why not just use a microcontroller and LCD? A 2 line by 16 char is easily purchased for less than $15. I like the atmel AVR controllers. Free software, trivial circuit board and it can be programmed by hanging 4 wires off your parallel port.

[kong]

Thanks guys, since I already have a "spare" pc in the garage, it would be daft not to hook it up. Lcd's aren't that cheap over here
Secondly, I have already tossed out the rubber wheel idea, and am contemplating the use of the wire betwen bearings or a mini timing belt system. The encoders I have on hand are only 128 CPR, so if I use a pulley system such as the timing belt, I can gear up the encoders for a greater resolution - higher than if they were simply rolled across the surface of the table. I am learning/thinking a lot as I go through this one, keep the ideas coming!

[H500]

The steel cable would be my second choice. I prefer the pinch roller system because it is possible to produce a much higher contact force so that there is less chance of slipping.

128cpr seems a little low. It would need to be geared up at least 8 times to get any usedful readings. It would take several pulleys/precision bearings. Over here, the cheapest optical mouse goes for about $13.00. I don't think a pulley mechanism can be made for less.

[kong]

If I have a shaft diameter of 3.2mm, then the encoder shuld put out 128 counts over a circumference of around 10mm. Equates to 0.08mm resolution. If the program takes into account the quadrature reading, it brings it down to 0.02mm resolution. I can live with that. The mechanical advantage of using the two bearings bolted on the non-moving part of the table as seen over here should bring it down to 0.01mm resolution. I see what you are saying, but I have the bearings kicking around somewhere, and the encoders, and erm.....need to buy the cable only.
Now then, back to the pinch roller system, my mind is elsewhere at present, can you expand on how it would be setup?

[jimbo]

I dont see how one can increase resolution beyond that of what the encoder can do, ie. the 128 cpr.

Never mind. It was stupid question.

[Al_The_Man]

I dont see how one can increase resolution beyond that of what the encoder can do, ie. the 128 cpr.

Apart from gearing, the leading and trailing edges of the quad count can be used to multiply the resolution by 4 times.
Al

[kong]

jimbo, by using pulleys and the cable system, you can double the length of the cable that passes the encoder for a given table movement. I didn't understand it myself until I did a google for "mechanical advantage" which will explain all about pulley configurations.

[jimbo]

Thanks for the responses. For a second I was thinking that gearing would somehow increase the encoder cpr. Cant change the counts per rev, so by adding gearing introduces more revolutions, and hence more counts for a given distance.

[Bubba]

Jimbo,
If you look at the crappy diagram at the bottom of my web page (The cable should show all across the top of the diagram), and follow how the cable is run, it may become semi clear.

Think of it as a block and tackle. By doing this repeatadly, it can dramaticly increase the length of the cable that runs over the the encoder shaft. You have to be sure to place the encoder shaft at the correct location (so the maximum cable goes over it) or you will not increase the advantage.

http://bellsouthpwp.net/A/r/Arts_home_page/DRO1/

[MrBean]

I've been reading this thread with great interest. I found some encoders in some printers I dismantled the other day and was wondering if they'd be of any use for a DRO project. I know next to nothing about encoders so I've attached a picture. The disk is about 40mm Diameter, 96 slots, (not many). If it's quadrature, does that make it effective 384 CPR? How can I tell if it's quadrature? Also, as mentioned in this thread, I could use gearing to increase the resolution.

Do these look worth experimenting with, or would they look better in the bin?

Regards Terry.....

[H500]

Bubba,

your site looks very interesting, but the text flows on top of each other, so I can't read it. Do you have any problem with chips getting under the cable?

[H500]

Terry,

In theory, you can get 384 counts per inch with the right software. If there are at least 4 wires on your sensors then its probably quadrature. If not, you can mount your own optical sensors on it. Just gut an old mechanical mouse.

[H500]

Kong,

I made a sketch of the pinch roller principle. Hopefully it makes sense.

The disk rolls when the table moves. The bottom pinches the rod against the disk shaft. Since it is made of steel, the force can be very high to prevent slippage. The rod can be quite thin, like 1/8" to reduce chance of picking up chips.

The linear resolution is multiplied by the ratio of the disk diameter to the shaft diameter.

For example if shaft is .5" and disk is 2", the linear resolution is multiplied 4X. With a 400cpi mouse, the resolution is 1600 cpi. The disk doesn't need any slots since an optical mouse reads off the surface texture.

[kong]

Thanks H, I like that. The rod (I am thinking silver steel or something ground accuractly) could be fixed to the mill table and would move, with the encoder fixed to the non-moving section. You have saved quite a few bearings (pulleys) already. The encoder could be enclosed with wipers or brushes to keep the rod clean.
Now the problems, with my current encoders with 3.2mm shaft, I will need to mount two bearings on the shaft to maintain rigidity. The shaft is very short and those bearing might not fit with enough clearance to run the rod against the encoder shaft. So......hang on, I am thinking of gearing here, but then you have backlash and/or added complexity which is no good. I will go look at the encoders and get back!
The only thing that worries me about the optical mice is how accurate are they to begin with? Are they capable of reading 0.1mm accurately, surely they are not a precision instrument capable of error free readings?

[kong]

Heres a thought, using a small diameter shaft supported between two bearings, along with the tensioning bearing below it, it should be possible to maintain a 0.02mm resolution. The only thing would be to bore the bearing bushing to recieve the encoders shaft. This way it will protect the encoder from the forces on the bearings. I also found some ground flat stock that should make a good guide rail.

[Bubba]

H,
I just looked at it in both Opera and IE and looks fine on my end?

Not sure what happened to you????

[H500]

Bubba, I viewed the site from work and it looks fine. I think I need to upgrade IE.

[H500]

Kong, it looks good. I like the way you used a single bearing on the bottom. That's 1 less than what I had in mind and makes it easy to apply the pinching force.

As for the optical encoders, I came across a university paper (can't find the URL now) where someone did some measurements to see how accurate they are. His conclusion was that they are indeed good for about 400 count per inch directly when mounted about 1mm from the surface.

So using the big disk/small shaft multiplication, it seems possible to get very high resolutions out of it. I believe the Agilent encoders are based on this principle.