5
\$\begingroup\$

The question

I am looking for an absolute linear position sensor that can span a distance of 40 cm with ~5 mm accuracy. My primary concerns are reliability and cost at a production scale of ~10,000 units. Ideally I can find a vendor that provides an off the shelf sensor but modifying an existing design and printing up circuit boards is not out of the question. I need recommendations on which technology to be looking at (LVDT's, optical encoders, magnetic encoders, etc).

The search so far

I have heard that potentiometers and mechanical solutions in general are less reliable than contactless solutions like LVRTs and encoders. Mostly I have focused the search on encoders because I am a bit more familiar with them. Most encoders I find are accurate to within a micron and cost more than I am hoping to pay for them (example). It seems to me like what I need is the insides of a cheap digital caliper just blown up a little (to 40 cm long) but I can't seem to find a vendor that builds and sells all the digital caliper innards.

Things I have already read:

Questions like this one and this one are similar to mine but both have different accuracy requirements and neither were asking about production at scale. Answers like "buy an optical mouse and take it apart" worked for them but they don't for me. I have also looked at comparisons of position sensing technology (example) but they always seem to focus on solutions with significantly higher resolution than I need.

Gory details in case they matter or your are curious

  • Mechanical constraints: hardware can't extend more than ~5 centimeters past the stroke length of the sensor.
  • Definition of reliability: At least 20-30k cycles before failure.
  • Sample Rate: shouldn't be an issue, as low as 10sps would be fine
  • Enviromental factors: A sheltered open air environment so temperature can swing ~0-100 degrees C and condensation can exist
\$\endgroup\$
  • \$\begingroup\$ I think you need to expand what the use is i.e. explain what the machine is doing \$\endgroup\$ – Paul Sullivan May 6 '14 at 18:04
  • \$\begingroup\$ @PaulSullivan You mean physical configuration? I have a rail and a slider. I need to know where the slider is on the rail. Other than that I can set up the mechanics pretty much however I need to as long as the sensor works. \$\endgroup\$ – Hammer May 6 '14 at 18:30
  • \$\begingroup\$ What moves the slider? \$\endgroup\$ – Tut May 6 '14 at 18:42
  • \$\begingroup\$ With 10k quantity you could most likely get tooling to extend digital calipers. You take the internals and extend the slide part as far as you need. They can skip @ high velocities though. \$\endgroup\$ – HL-SDK May 6 '14 at 19:37
  • \$\begingroup\$ @Tut a person moves the slider \$\endgroup\$ – Hammer May 6 '14 at 20:53
2
\$\begingroup\$

If you're really glued to the idea of an absolute measurement, I'd look into going the theory of operation of an electronic caliper.

How does an electronic caliper work?

As a word of warning, these are inexpensive because they're commodity items at this point. It might get pretty expensive per unit.

| improve this answer | |
\$\endgroup\$
2
\$\begingroup\$

You could also use a capacitive sensor. Have a floating contact that travels over long thin conductive triangle. At one end there would be hardly any capacitance and at the other the surface area would be maximized increasing capacitance.

| improve this answer | |
\$\endgroup\$
  • 1
    \$\begingroup\$ 40KHZ Ultrasonics have a wavelength of 8cm... unless you can work out where in the wave you are (and ignoring echoes) you are going to have resolution problems \$\endgroup\$ – Paul Sullivan May 6 '14 at 20:04
  • \$\begingroup\$ Resolution of 1 mm so that's not the issue. Minimum distance is the issue actually. Cost is another one: maxbotix.com/Ultrasonic_Sensors/MB1013.htm I'll remove the ultrasonic part of my answer. \$\endgroup\$ – horta May 6 '14 at 20:57
  • \$\begingroup\$ sorry... not 8cm... it is 8.6mm resolution brainfart \$\endgroup\$ – Paul Sullivan May 6 '14 at 21:29
  • \$\begingroup\$ Your idea of measuring the capacitance of a triangle is not one I have heard of before, it sounds intriguing. Do you know of any applications of that technique? \$\endgroup\$ – Hammer May 6 '14 at 21:32
  • \$\begingroup\$ I do not. It's basically a variable capacitor tho. If you try this method, make sure you use a high enough frequency as the capacitance won't be much. An alternate would be trace pads at 5 mm intervals. Then just scan through the pads and see which one(s) have higher capacitance. \$\endgroup\$ – horta May 6 '14 at 21:43
1
\$\begingroup\$

You don't say if the position is required to be nonvolatile. If not, you could make a belt/gear combo that drives an optical encoder. With 1 inch dia gears, your 5 cm figure seems eminently doable. This would require a zeroing cycle every time power is turned on, though, which is why I ask about volatitity.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ I have definitely thought about that. I am just worried about having a mechanical belt and gear. Have you ever done something like that and had success? It seems that there is a trend towards minimizing moving parts when going for reliability... \$\endgroup\$ – Hammer May 6 '14 at 20:56
  • 1
    \$\begingroup\$ I've never used it a sensor, but the technique is commonly used as a driver for x-y tables, where the encoder is replaced by a stepper. Seems to work fine. And certainly I wouldn't worry about reliability if all you're worried about is 30k cycles. "Gear" isn't the precise word for it - try timing belt and pulley. See, for instance bbman.com for a wide variety of possibilities. \$\endgroup\$ – WhatRoughBeast May 6 '14 at 23:21
1
\$\begingroup\$

Skin a cat... But anyway... 2 cents

Encoders: Use gears (you can get reasonabaly cheap rack and pinion gear sets) BUT you are limited by lengths of racks

Absolute vs Incremental: Both can be bought cheaply (I have some from China which are 600 pulse/revolution @ $20) which when tied to gears are accurate enough. I would suggest you can get even cheaper and combined with gearing will be accurate enough.

Crackpot idea (come on real engineers) - Use copper tube and a brush connection, measure resistance with biasing resistor, low current/voltage and ADC...

Another idea is to have holes in/under the shaft at known distances that can be counted by means of i.e. LED + LDR combo...

| improve this answer | |
\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.