# Using a potentiometer to replace an absolute optical encoder

I am trying to replace a rotary absolute optical encoder (single turn). The resolution of the encoder is 13 bits (8191). The reason for the replacement is to reduce the cost.

My approach is to get a single-turn (electrically) infinite turn (mechanically) precision potentiometer. Use a 16-bit ADC (neglect the last three bits due to noise).

Questions:

1. Type of sensor element: As I have read, conductive plastic is the best for such an application. Am I right?
2. Linearity: The least I found was 0.15% for a wirewound potentiometer (which I do not know how much error in degrees it has and how it will affect my result)
3. Electrical travel: most potentiometer's electrical travel is a bit under 360 degrees (320 to 358). For example: beisensors. So in the last 2 degrees will the system be blind and does not know where it is? Is this common? Does it have a solution?

Please correct my information if I am wrong. Do you think these problems are limiting factors in potentiometers, and they cannot be used to the specific task that I need?

• A pot cannot have 360 degrees of travel unless it is a multi turn type. Commented Sep 10, 2016 at 9:18
• @Andy aka do you mean electrically or mechanically ? and why ? Commented Sep 10, 2016 at 9:29
• Because there has to be a gap between the ends of the resistance track. You would need two pots ganged onto the same shaft 180° out of phase. Commented Sep 10, 2016 at 9:33
• @Transistor , ahh okkk , it makes sense now , nice solution 2 out of phase , using each in its sweet stop . can you comment about linearity relation with error in angle ? Commented Sep 10, 2016 at 9:42
• @ElectronS - have you considered a magnetic rotary encoder? Full 360 degrees, some have similar resolution, and you don't have pot wiper problems. Example:AS5048 Magnetic Rotary Encoder from digikey. Commented Sep 11, 2016 at 3:18

All poteniometers are < 360° because there has to be a gap between the ends of the resistance track. You would need two pots ganged onto the same shaft 180° out of phase.

You got me interested, however, and a search for "360 degree angle sensor potentiometer" threw up some interesting products from Novatechnik. See rotary shaft type and the AW360-ZE where they seem to have this figured out.

Figure 1. Extract from the AW360-ZE datasheet.

Their "potentiometer" takes a Uref reference signal and gives a DC output proportional to Uref x angle.

They don't go into specifics of how they work but here are my thoughts:

simulate this circuit – Schematic created using CircuitLab

Figure 2. (a) A possible method of fabrication. A circular resistance track with connections at 0° and 180° would give a triangle wave output when rotating continuously in one direction. By adding 4-quadrant opto-detection it would be possible to figure out which quadrant the wiper is in and use that in the Logik circuit. (b) gives what might be a more reliable method. By simultaneously switching SW1 and SW2 they could use one switch position to get an actual voltage reading and the other to determine the quadrant.

• Using the scheme of Figure 2b with the switches and wiper in the position shown Vout = V+/4. That means is in the bottom half of the pot and it's either 135° from the 12 o'clock position.
• Throw the switches and Vout is again V+/4. That means it's 135° from the 3 o'clock position.
• Only one point satisfies both conditions.

• wooww , great answer and interesting potentiometer , however the price seems obsurd (more than the absolute encoder i want to replace): impexronshop.com/products/… Commented Sep 10, 2016 at 11:10
• the lowest absolute linearity is 0.035% ( i checked their website this is lowest ) , which according to Andy aka answer means 0.126 degree , which is quite good ,BUT if i assume no noise or other measuring errors , the output will be +/- 0.126degree from the actual position. WHICH IS nowhere near the OPTICAL absolute encoder (zero linearity) resolution : 360/8191= 0.04 degree . It looks like they are in different leagues :(( Commented Sep 10, 2016 at 11:15
• There are such things as pots with two wipers that turn indefinitely. I used a few in a simulated aircraft cockpit once. They are cheap alternatives to rotary encoders, but have low accuracy, linearity, and repeatability. It also takes some cleverness in the firmware to determine the angle from the two voltage measurement. Commented Sep 10, 2016 at 11:32

Let's start with the resolution of the rotary encoder. 13 Bits, provides a rotary resolution of 8192 angular positions (not 8191 - a minor point for sure). This means the rotary encoder, if properly utilized) can resolve 0.044 degrees of rotation of its shaft. In more conventional angular measurement terms that's 2.6 arc minutes of angular displacement. ("arc minutes" is a term used to describe angular displacement to distinguish it from the units of common "minutes" of time.) There is no potentiometer on the face of Earth which can match, meet, or exceed this resolution.

The best potentiometers available for this type of application are called "instrumentation potentiometers" or "precision potentiometers". They are a class quite apart from the type we buy from DigiKey and Mouser. They are typically custom made to order and to specification. There are not many manufactured anymore due to the encroachment of rotary encoders into the applications precision pots were traditionally used in.

There were, in fact, 360 degree (and wider) precision pots. But they were only meant to run in one direction, not in continuous rotary rotation. They used a spiral track with a spring-loaded wiper. A sort of mini multi-turn pot with a fractional second spiral. I digress.

Precision pots were "precision" for several reasons. First, they utilized a knife-edge (wedge-shaped) wiper which provided the fine resolution along the main resistive element. Second, the resistive element was very fine grain, homogeneous material. Indeed, plastic type resistive elements were the best available towards the end of the heyday of the precision pots. Third, the resistance element was mechanically ground to the level of specified perfection in a closed loop fixture. This fixture measured the resistive division ratio between the elements end points and the wiper pick-off point for a given point of angular displacement around its arc, and ground away part of the element automatically to achieve a specified ratio at that point. This operation guaranteed the linearity of the pot. The more points you specified, the more expensive the pot. (And they could cost several hundred dollars apiece back in the 70's and 80's when I worked with these.) You could also specify non-linear "tapers" of various contours thanks to this grinding process. Fourth, the whole assembly was made with great care & precision - special bearings for the shaft, machined (not stamped or molded) case, carefully machined shaft, polished (not sharpened) knife-edge for the wiper, total support for the resistance element, etc., etc... Again, I digress.

APPROACH: All wrong. 16 bits, or 13 bits, of ADC resolution will not have one "bit" of effect on the resolution of any ordinary (non-instrumentation) potentiometer's inherent ability to resolve small angular differences. Ordinary pots are simply not made to provide this type of resolution. The shafts are very loosely mounted, so there is a lot of slop and hysteresis with relationship to the wiper and the resistive element. And the element itself is very non-linear in these pots. You will need an angular displacement jig to test your candidate pot to see if you can achieve the required resolution with your candidate pots. There fixtures can be purchased from machine shop tool suppliers (even Harbor Freight sold them and maybe still do). Jig up your candidate pot to this fixture and exercise it using a simple voltage divider hook-up with a precision DVM in voltage mode. I doubt you will get better than 1 degree of useable resolution. "Useable" meaning repeatable as you move the wiper away from the specific angular position and then back again after various trial displacements in both directions (CW & CCW).

PLASTIC ELEMENT - Yes, this is likely the best path to follow. Wire-wound types have inherent cogging due to the wire ridges.

LINEARITY - Linearity is not resolution! Plus, there are many definitions of linearity and an equal number of ways to apply the quoted linearity specifications. There are many fine write-ups on the differences between resolution, precision and linearity available on the Internet. Linearity is probably important in your application. Nevertheless, it is quite a different parameter from resolution. Not to mention there are all sorts of specmanship tricks played when stating the linearity of devices such as potentiometers.

FULL-360-ROTATION - This will be a real challenge. If you are reduced to mechanically altering a stock potentiometer to achieve this characteristic, you will probably be better served directing your efforts to modifying an available rotary encoder to fit into your application.

If you search on Ebay for "Instrumentation Potentiometer", you will find a few examples of the type of pot I described above, mixed in with many more conventional "adjustment" pots. The instrumentation type were generally made by small companies with now-unfamiliar names. If you decide to use a potentiometer, get one of these.

• thank you for the thorough explanation , background , and the conclusions you have made . You saved me a lot of time and corrected my misunderstanding . It looks like the optical encoder is the cheapest option in this case and there is no free lunch with sensors and precision instruments . best regards :) Commented Sep 10, 2016 at 12:55

There are potentiometers intended for this kind of application. However, you aren't going to get the angle to 13 bits of accuracy. No way.

These types of pots can turn infinitely, and have two wipers. The resistance element doesn't go the full 360°, just like in a normal pot. By looking at the signal from both wipers with clever firmware, you get get angular position.

The raw linearity is not very good. You can fix this to some extent with calibration at a few known angles. However, you're not going to get a result good to 13 bits. The unrepeatability of the wiper contact and mechanical backlash probably prevent that anyway, but the non-linearity will certainly swamp it. It will also change with wear.

These kinds of pots are useful for things like cheap user inputs that need to be turnable indefinitely, not for accurate angular measurements.

• have you checked Novatechnik's AW360-ZE mentioned above? it's certainly not built for cheap user input?
– fhlb
Commented Sep 10, 2016 at 11:37
• @fhlb it is around 430euro (equal or greater than the price of the optical encoder) as i mentioned before ,link: impexronshop.com/products/… i donot know why such price how would use them instead of an encoder . by the way the life time is 50 million life as the datasheet states. Commented Sep 10, 2016 at 12:05

You haven't told us your specific task, so we have no way of knowing the suitability. You may put a ton of time into dealing with what you think of as the blind spot, and you might land on a software solution. You say your encoder was single turn, so I suspect you can put the blind spot of the pot in the region where the encoder couldn't reach without problem.

My biggest concern would be lifetime. Pots are only good for so many rotations. Optical encoders are often contactless, with very long lives. Make good and sure that your pot has a long enough lifetime. If you can't find a lifetime rating, find another pot.

If you know the expected lifetime, you probably won't care abut the material. Wire wound pots will have tiny flat spots, but Will probably be linear enough for your use (which I don't know).

the least i found is 0.15% for a wirewound pot( which i donot know how much error in degrees it will affect my result)

A linearity of (+/-) 0.15% means that if the angle of the wiper were exactly at (say) 50% of the mechanical travel (say 180 degrees for a perfect but impossible single turn pot) the electrical potential divider value isn't exactly 0.500000000 but can vary either side by up to 0.0015 making the real error range at 50% travel between 0.4985 and 0.5015.

By my reckoning, 0.4985 translates to 179.46 degrees or an error of about 0.5 degrees.

As i have read conductive plastic is the best for such application. right ?

I believe this to be correct.