Absolute Encoder Teardown

According to [Lee Teschler], the classic representation of encoders showing code rings is out of date. His post says that most industrial absolute encoders use a special magnetic sensor known as a Wiegand wire to control costs. To demonstrate he does a teardown of an encoder made by Nidec Avtron Automation, and if you’ve ever wondered what’s inside something like this, you enjoy the post.

This is a large industrial unit and when you open it up, you’ll get a surprise. Most of the inside is empty! There is a very small encoder inside. The main body protects the inside and holds the large bearings. The real encoder looks more like a toy car motor than anything else.

The inner can is nearly empty, too. But it does have the part we are interested in. There’s a Melexis Hall effect sensor The Weigand wire is a special magnetic wire with an outer sheath that is resistant to having its magnetic field reversed and an inner core that isn’t. Until an applied magnetic field reaches a certain strength, the wire will stay magnetized in one direction. When the field crosses the threshold, the entire wire changes magnetic polarity rapidly. The effect is independent of the rate of change of the applied magnetic field.

In other words, like old core memory, the wire has strong magnetic hysteresis. Between pulses from the Weigand wire and information from the Hall effect sensor, you can accurately determine the position of the shaft.

It is always amazing to us how many modern pieces of gear are now mostly empty with the size of the device being driven by physical constraints and not the electronics within.

22 thoughts on “Absolute Encoder Teardown

    1. The linked article says the encoder has a diametric magnet and magnetic rotational encoder sensor IC as well. The article doesn’t say precisely, but I assume the diametric magnet and encoder IC provide the 4096-resolution relative position with the Wiegand wire and sensor setting the zero position.

      1. The supplier page for the encoder claims “no batteries”, so I think that the Wieland wire is to provide power for multiturn counting when the encoder isn’t being externally powered.

      2. A diametrically magnetized disc magnet can be used instead of 2 bar magnets: this configuration (2 bars) improves single-turn accuracy. An IC (the Melexis sensor in this case) provides angular position with 4096 resolution over 360°. The same magnet is used for the multi-turn counter and it consists in:
        1) Wiegand wire
        2) Hall sensors (ELMOS IC)
        3) Wiegand counter (ELMOS IC)
        4) Non volatile FRAM (it should be a Ramtron IC)

        Every time a Wiegand pulse occurs (2 times per turn) the Wiegand counter IC powers up and turn on also hall sensor IC to evaluate the magnet pulse polarity and then if the magnet is turning clockwise or counter-clockwise, to increment or decrement the counter in FRAM.

        When the encoder is powered up the MCU has to determine withe the single-turn position if the multi-turn position is correct or not (i.e. a new Wiegand pulse is about to occur or there was a direction change), then the multi-turn position needs to be adjusted (+1 or -1 correction).

        In newer designs Hall sensors and Wiegand counter are integrated in the same IC.

        This technology is patented and requires licensing

  1. According to the Datasheet: “The AV30 measures the shaft rotation and position without the need for external power or internal batteries through it’s innovative Wiegend wire energy system.”

    Another quote: “By utilizing Wiegand wire energy harvesting technology combined with magnetic sensors, Avtron has created an absolute encoder design which requires no batteries, long-term capacitors, glass disks, or gears!”

    Seems kinda weird though since all the connection diagrams have it connected to power, ground, and data wires.

    1. The energy harvesting is so that it can maintain it’s internal position while unpowered. It still needs power when the machine is in use. If the shaft stops turning, and the encoder runs out of power, and stops transmitting, the machine is going to fault out because it thinks that the encoder’s broken.

    2. I might be real dense today, but both of those are like saying “Gasoline and water release a lot of energy when burned in a stochiometric ratio… and hence through the miracle of internal combustion Lindberg was able to cross the Atlantic” ummm there’s a lot missing in the middle there.

          1. That’s ‘coz gasoline is all like “In ‘ertford, ‘ereford and ‘ampshire, ‘urricanes ‘ardly ever ‘appen” and then they run off with the Os.

    3. Having built systems around industrial encoders I assume what they want to express is the fact that the encoder still knows its absolute position when rotated *without power being applied*. Because that’s what you expect from an absolute encoder and that’s what glass disc (=gray code) encoders are able to provide.

    4. I think that means that it’s able to track position (but not report it) on just the power generated by the motion itself.

      An absolute encoder EITHER needs enough bits to immediately know where it is if it powers up from zero and the application asks about it, OR needs backup power to continue counting relative pulses and mentally track its position in RAM and be able to report that when it powers up and the application asks about it.

      1. Yikes… As awesome as this wire thing is, I find it rather disconcerting to learn that absolute positioning encoders in huge machinery rely on batteries. Or, more importantly, that folk who rely on these sensors maybe completely unaware the batteries might run out one day. And,TBH, even though this system takes batteries out of the picture, I still don’t think I’d be comfortable around big equipment which relies on this.

        I really expected to find that this wire-sensor was used in multitudes around a magnetic sprocket, maybe with some sort of vernier scaling. Or used for quadrature in non-absolute systems. I’m actually rather disappointed it’s not really used as a “sensor” at all, but a generator… especially being that its use as a generator, here, seems rather hackish in a subsystem which itself is rather hackish for the sorts of systems that could literally take heads off. I love me some hacks, but not there.

        1. That goes for RAM, too… We’re seriously relying on something which may be prone to tin-whiskering or even just vibrational damage? But, they can’t be using volatile RAM, here, can they? So, how many writes does their Nonvolatile memory of choice guarantee? Has it even existed long-enough to be tested to that level?
          Good lord, it’s friggin’ terrifying to think how many engineers read these whitepapers and assume that the technology they’re purchasing is built *atop* long-reliable designs. How could they possibly know what they’re getting these days *without* tearing it apart?

        2. Fanuc CNC battery backed encoders are industry standard. 1 million pulse per rev.
          I’ve got machines that are 20 years old and still on original batteries. If they ever get low the machine will give you a few days notice to swap the batteries out.

          1. I had the opposite case, omron inverter and encoder, machine with 64 axes, battery down after 1.5 years machine did not say anything. After machine power down over the holydays – little big catastrophe

        3. Calm down. Most traditional designs of absolute multiturn encoders rely on gray code disks (or rings) and a mechanical gear between them. They are read either optically or inductively. No power source needed to maintain their absolute position.

          1. LOL, this does the exact opposite of calming my concerns in my last paragraph. That is in fact my point. Investing in a device like this, someone with this level of confidence in their design/functionality could easily assume a new product will live up to the long-term proven reliability of that “tradition.”

            And for the 20yr battery-backed comment… I guess it depends on whose hands might get crushed. Though, I’m having a really hard time believing a system like this would be intended for machines which are always only used-by and only ever in the presence of an operator that knows how to use it, what the error message means, not to ignore it, and more.
            A system like that surely would be much wiser to rely on a “homing” routine at each power-up (lathes, mills, huge plasma cutter beds), so has no particular need for an absolute encoder, nor tracking rotations across power cycles.
            The only systems I can imagine that would actually need a device like this are *huge* systems like cranes which will have *numerous* folk working in around and under it, none of whom are aware of how it works, nor should be expected to be. Their workflow being heavily reliant on its just working, as it always has, would be upset dramatically if the device someday lost track of its position and slammed into the end at full speed.

            What if the battery connector itself wobbles loose, what if the warning light goes ignored for so long that folk don’t even notice it anymore? Nerves Not At All calmed by these anecdotal “evidence” of reliability.

            Frankly, it seems to me, more evidentiary of /exactly/ my concern that even the folk making the decision to purchase these devices don’t really even know why they need them, as opposed to other options. Building a waterjet cutter? “well, the indusry standard is an absolute encoder on each axis” OK, but WHY? “Well, then you don’t have to wait half an hour for a homing-routine every morning!” OK, then… the answer is, it’s the lazy defacto standard, and it’s OK in that environment since no one is in the waterjet chamber. But now Bob is an “expert” on such systems and is tasked with cargo-container gantries where folk are walking under and between… And it’s no longer a matter of “needing” absolute-encoders to save a half-hour homing routine, but a matter of: at the end of the day the entire plant gets shut off, and moving that gantry through a homing routine in the morning is not an option because it’s still holding a container. Now actual position on powerup is essential. But Bob hadn’t really thought about *why* and just knows that it worked great on the waterjet machine. And that it’s even more necessary here. In BOTH cases, really, neither are necessary. It’d be easy to add inductive proximity sensors spaced all along each axis. “Homing” is merely a matter of moving slowly in the direction you intend to go, anyhow, until that next sensor is tripped. No batteries, no RAM. No question of gear-slippage on a multiturn encoder, so on and so forth.

            So, then, what *are* these encoders, or even the mechanical multiturn optical ones, *actually* needed for? Maybe axes which aren’t cartesian, “elbows” where external position sensors aren’t an easy option? Do you really want a huge machine like that to only know its position based on the viewpoint of the actuator? What if there’s a slow hydraulic leak? What if a driving gear loses a tooth?

            UGH! Have you ever looked at the steering system in your car?! “Industry standard” is “we must have God looking out for us since we don’t hear of 60MPH car-flips Every Day in Every State.”

  2. I struggled to understand how this works based on the original article, but these references made it much clearer for me:


    As I understand it, there’s a special hall effect sensor that can measure absolute rotation by itself, but only within a single turn. The Wiegand wire device is there to harvest power so that any complete rotations of the shaft can be tracked and stored in FRAM, even if the power is off.

    Combining the two results in a multi-turn absolute encoder.

    1. This exactly the point. The original article is incomplete or misleading, the above article adds more confusion. Happily your comment provides the correct references, thanks.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.