# How A Rotary Position Sensor Works

I just want to apologize for not being very clear in my question and I understand what you guys are saying. After reading the guidelines for this site I understand now that this was not the correct format or structure to ask this question, its not that this site can't give me the information I need, its just that the way I asked was downright terrible and wasn't facilitating any learning. This site is supposed to be a site of knowledge that could be used by all users in the community, not just for personal use and specific questions especially not for business matters. I have edited my question in what I believe to be a much more "relevant to others" format, that just seeks clarity in fundamental concepts of electricity and magnetism and how they could apply to sensing position/angular rotation of a physical object rather than clarity in the specific operation of a sensor that I haven't even given you any information on.

Unfortunately I was told explicitly that I cannot say what the function of the sensor is nor what it is called due to confidentiality agreements. So after re-reading my post and the helpful comments of the community I understand the issues with my post and have revised it accordingly.

I really appreciate all the feedback from you guys, this site is an incredible resource, filled with knowledge for all and is filled very intelligent people that are willing to help others. I would love feedback from you guys on the format of my newly edited question, I think I have a much better understanding of the purpose of this site and how to ask questions that not only help me figure out my own problems, but also benefit the community.

Just to clarify some points that I didn't explain in the original post:

• I am not asking anyone to figure out the issue we are having with this sensor, I haven't told you what it is or what its function is, nor any details as to the problem we are having with it. I would love to bring you guys into the problem and have in-depth discussions on possible root causes because I think it would be really cool to see the ideas you generate and I would definitely have a great time talking about it with you guys but again, due to confidentiality agreements I must keep the specific details of the sensor as vague as possible.

• It's not that I trust you guys more than the supplier of this part or that I trust the supplier more than you guys or that I don't trust the supplier at all, it's just that I am looking for a third party's explanation of some concepts the sensor uses just to confirm and clarify what the supplier is explaining to me. I am just simply having issues with some concepts they are explaining to me which could very well be completely valid, it's just that I am not grasping them. So I am looking for a third party that might be willing to help me. But if the supplier is just spoon-feeding me explanations and telling me to believe things that I don't completely understand I am not just going to believe them just because they say I should. So I apologize for not making these points clear in my original post.

Reformatted Question:

I work for a large car company in North America, and we have a sensor (which unfortunately I can't go into detail with due to confidentiality agreements) but essentially it is a rotary position sensor that uses two rotating magnets on different teethed gears (like the ones shown in Figure 1) to determine angular rotation of the big wheel (Wheel 1 in Figure 1) from some preset "zero" point. My team is trying to determine the root cause of an issue with this sensor so I talked to the supplier who explained to me how it works and I still don't completely understand how the sensor works which makes it hard to root cause, so I am hoping someone can give me some theories on how they think a hall effect sensor that uses two magneto resistive pairs that is setup like the one in Figure 1 would be able to tell the vehicle how many degrees the big wheel (Wheel number 1 in Figure 1) has moved.

                  Figure 1- The Image the Supplier Sent me


The supplier of the sensor explained how the sensor works and gave some speculation as to why the issue might be occurring. Regardless of what the sensor specifically does, the image and explanation above just give some context to my specific issue to better illustrate the concepts I am having issues with which is how could two magnets rotating on two different teethed gears be able to tell a computer (The Engine Control Module [ECM] on the vehicle) how far the big wheel (Wheel 1 in Figure 1) has moved.

Essentially the supplier told me that this sensor is using the hall effect to tell the vehicle how far the large wheel (Wheel 1 in Figure 1) rotates, my understanding of a hall effect sensor which is shown in Figure 2 (Which I guess I will put in the comments because I can only put two links in a post) was that the varying magnitude of the magnetic field strength which would come from the magnet physically moving around the big wheel (Wheel 1 in Figure 1) which would correspond to certain amount of degrees of rotation.

So for example if the sensor saw X amount of magnetic field which would increase or decrease as the magnet moved around the big wheel (Wheel 1 in Figure 1) then that would correlate to X degrees of rotation for the big wheel(Wheel 1 in Figure 1). But in this sensor, based on the image given to me by the supplier (Shown in Figure 1) that is not what I think is happening in this sensor.

So I understand now that my original understanding was incorrect, you probably aren't going to have the magnet physically rotate around the big wheel (Wheel 1 in Figure 1), this wheel is pretty big relatively speaking [Can't give too much away ;)] so you would need a pretty strong magnet for the sensor to be able to detect the magnetic field if the magnet was all the way on the other side of the big wheel.

So I did some more research and came across another method for how a rotary position sensor could use magnets to determine the angular rotation of the big wheel.

Here is how I now think a rotary magnetic position sensor works:

         Figure 3- How I Now think a Hall Effect Position Sensor Works


I think that a single rotating magnet attached like the one in Figure 3 will produce magnetic flux which the sensor would somehow convert into sinusoidal voltages (Maybe one of you could explain how that works but that's for another day) which would correspond to a certain degree of rotation. But again, from the picture the supplier sent me, I don't think that is going on here either.

So I am just looking for some of your ideas on how a setup like the one in Figure 1 with two rotating magnets on two different teethed gears could tell a computer how far the big wheel has moved. I was thinking there could possibly be something similar to this setup used in another industry application that you guys might know about or if there is a general methodology for how setups like the one in Figure 1 work.

So hopefully I illustrated my question better, (I couldn't have made it much worse from the original lol) and I really appreciate all the help from this community.

• So you have a sensor which you can't tell us about, which the supplier explained to you how it works, but want us to explain it again to you because you trust us more to tell you about something we dont know what it is than the supplier of it – PlasmaHH Aug 1 '16 at 13:20
• You need to read up on the Hall Effect, which is sensitive to the orientation of the magnetic field. Your sensors contatain a pair of Hall Effect sensors at right angles, so the relative outputs can be used (within limits) to derive the orientation of the magnets. And the phase they're talking about is not an AC effect - it is simply the difference in angle of the two gears. You seem to have badly missed the point of the explanations you've received. – WhatRoughBeast Aug 1 '16 at 13:55
• As @PlasmaHH said, you provided very few information. Besides NDAs, you said you have an issue with your sensor, but you tell us nothing about it. How do you expect we can diagnose the cause of something you didn't even described? And if your intent is just to understand how the sensor works on the basis of basic principles, you should have posted the question in a different way. – Lorenzo Donati Aug 1 '16 at 14:21
• As it stands your question is not useful for other people browsing the site. It is like someone wanted to know how a bicycle works asking something like "I have a vehicle with a certain number of wheels (which I cannot disclose) that doesn't behave as I'd like. Can someone explain how it works, so that I can solve my issue?". Voting to close. – Lorenzo Donati Aug 1 '16 at 14:22
• BTW, just to avoid giving the impression we scare off newcomers, please read the guidelines of this site. This is not a general electronics forum, but a site which aims to build a high-quality knowledge base in Q&A format. Any question that is to survive on this site must be proven useful not only for the person asking it, but for future people browsing the site. – Lorenzo Donati Aug 1 '16 at 14:26

The following is a very likely correct explanation of what you dealing with. Even if the sensors used are not identical to the ones used for example the method is almost certainly what is used.

Note that the sensors are stated in the diagram's text to be "magneto resistive sensor pairs". MR sensors determine angle and NOT magnitude of magnetic field relative to a sensing axis. Answers and comments referring to Hall Effect sensors are incorrect as a Hall Sensor works on field strength and not field angle.

Read the following as slowly as needed and as many times as needed for it to make sense. Once the basic method is understood everything else 'drops into place'.

What happens is that the two small gears have tooth counts varying by 1 (typically). For counts N and N+1, for every N tooth movement of the main gear, the N tooth gear rotates by 1 full turn and the N+1 tooth gear rotates by N/(N+1) of a turn. This allows the N+1 tooth gear to act as a "counter" of the number of times the N tooth gear has rotated. By adjusting the tooth counts of the large and small gears it can be arranged so that the N+1 tooth gear completes one less full revolution than the N tooth gear when the N tooth gear is NOT at its zero position. This allows the N+1 tooth gear to be a multi turn absolute counter of the N tooth gear's rotation for more than N_large/(n+1) revolutions of the N+1 tooth gear. The maximum number of turns of the large gear that can be accurately sensed depends on the mechanical construction accuracy, sensor resolution and properly chosen tooth ratios.

______________

The image below combines your original image with figure 1 from the NEWSCALE
Smart off-axis absolute position sensor solution document. PDF here. Your system may have devices corresponding to those shown here. In the newscale system they claim that the absolute angle of the magnet north-south axis can be determined with 14 bit accuracy.
They say -

• With a two-pole magnet, the sensor array can deliver absolute angular information. An example is the ams AS5048, which incorporates an array of Hall effect elements and integrated ADC. It provides absolute angular measurement over 360° with a resolution of 14 bits (16,384 positions per revolution).

Assume Large/small gear tooth count ~= 4:1 - more below.
The encoder from NEWSCALE claims 14 bit accuracy or 1:16384 resolution or 360/16384 = 0.021... of a degree. Even if it obtains 10 or 12 bits after backlash, noise, calibration and Murphy that's in the 0.1 to 0.4 degree resolution & maybe accuracy range for the small gear and 4 times better for the large gear. (I'll assume a working accuracy of 0.1 degree for the small gears in the material below.)

Now the stage 2 magic starts.
You have two small driven gears. They differ by one in the number of teeth.
For example, say the main gear A has 80 teeth, small gear B at lower leFt has 20 teeth and small gear C at lower right has 21 teeth.
Imagine that initially both small bear sensor SA and Sb are oriented at 0 degrees with say North pole up. Rotate main gear A through 90 degrees = 20 teeth. Gear B will go through one full revolution (20 teeth) and the 1/4 turn of A will be transmitted by the sensor Sa.
However, as C has 21 teeth it will have turned 20/21 of a turn so its encoder will output 20/21 x 360 = 342.857 degrees, or 17.142... degrees less than a full turn. Let's work with 17.1 degrees. From the Sb reading of 0 degrees and Sb reading of -17.1 degrees we know that the gear A has turned 1/4 turn.
Now advance A another 20 teeth = 1/4 turn so it has turned 1/2 turn.
Sb is again at 0 degrees BUT Sc is now at -2 x 17.1 = -34.2 degrees.
So, as we advance gear A by 90 degree steps Sb goes through 1 x 360 degree cycle and returns to zero, and Sc steps on -17.1 degrees per 1/4 turn of A. After 1 turn of A Sb = 0 and Sc = -68.4 degrees.
After 5 turns of A, Sb = 0 degrees and Sc = -342 degrees.
After 5.25 turns of A, Sb = 0, Sc = 359.1 degrees - so STILL not back at zero, and as A turns slightly more B is off zero as C passes through zero for the start of another 21 turns.

By careful selection of the teeth count on A, B, C the system can be used to determine the absolute position and number of whole turns of A from rest by simply reading the 14 bit position signal from the Sa & Sb sensors. The readings are absolute - no 'count' must be kept of transitions past a point or from zero. The system can be "immensely accurate" - it is likely that for rotations in the order of 10 to 20 turns of A the positional error would be governed by mechanical tolerances rather than electronic ones.

___________________________

What could go wrong?

Indeed. The $64 question. Or, the$N00,000 question in this case.
From the supplied information it is not clear what the problem is, but it should be 'easy enough' to determine if the system and equipment are working as intended and, if not, why.
sensor outputs should be well defined and completely repeatable.
Issues MIGHT arise with sensor noise, magnet attachment, demagnetisation, changes in magnet-sensor distance, digital noise, general electronic failures, ... . it SEEMS like an "easy enough" problem to troubleshoot - whatever it is.

Touting for business is generally frowned on on this site. Si I'll just note that if after reading the above it seems 'correct' and you want someone to investigate it, please see my profile page for contact details :-).

A few minor liberties have been taken with numerical aspects which do not affect the gist of the answer and which improve readability.

_________________________

References:

TE - the 600 pound gorilla Magnetic sensors

Zetex application note - good Basic Introduction to the use of Magnetoresistive Sensors - Howard Mason, September 2003

Wikipedia Magnetoresistance

Honeywell

KMA36 rotational and luinear MR sensor - pops up all over.

Many images - each links to a related webpage.

You seem to be thinking of the whole system as if it is rapidly spinning. Generally steering systems don't do that, most people older than about 5 don't sit there spinning the wheel as fast as they can while driving down a road.

The sensor as described is intended for reading something that is moving slowly enough that to the electronics it is effectively stationary. Once you think of it in that way the explanation of how it works is perfectly reasonable.

You may have to break with tradition and actually fixing a problem under warranty for a change.

• and yet I can sense rotor position and velocity quite happily to 20,000rpm – JonRB Aug 1 '16 at 13:45
• 20,000 rpm is 3 milliseconds per revolution or 8us per degree. That's not that fast. – Andrew Aug 1 '16 at 14:37
• Faster than twitching a steering wheel – JonRB Aug 1 '16 at 14:38

If the magnets are just spinning and not physically moving closer or farther from the sensor then how is the sesnor getting different values for magnetic flux density?

Of course they are moving closer or farther away (as far as the sensor is concerned). A rotating magnet produces a field (to a static sensor) that reverses polarity every rotation. This is easily detected and even measured as a linear output proportional to instantaneous rotation angle.

Regards the uniqueness of the two signals over the whole range of rotation of the steering wheel consider it over just one rotation and - draw dots on the little gears and see how they align/misalign as the steering wheel is turned. I'm not going to do this but I urge you to. Ultimately upon several revolutions of the steering wheel they will re-align but the idea is that within the limited revs of the steering wheel they should never realign thus, every turn of the wheel generates a unique pair of signals.

Think about it - there are maybe 40 teeth on one and therefore 41 teeth on the other.

• Re "of course they are ..." -> probably not. It's liable to be a GMR sensor that determines angle of magnetic field raher than field strength. Magnet axis rotates over face of sensor. See my answer - and refs which I'm about to add. – Russell McMahon Aug 1 '16 at 15:52
• @RussellMcMahon At 90 degrees the field will have dropped to zero so I disagree with you in the absense of any specific field sensing methodology provided by the OP hnce why I voted to close the question. – Andy aka Aug 1 '16 at 15:57
• the specific sensor technology used is stated in his diagram. It says "... two magneto resistive pairs placed in line with the magnets ...". It's actually FAR more detailed than I'd have expected. | Your comment about field dropping to zero may make sense with respect to some classes of sensors arranged in some manner, but assuming tight technology bounds and using the assumption as a reason to close an answer is 'a bit sad'. Even if he had had maximal information available he would not have been able to describe the system perfectly. That's what engineers are for - or should be :-) – Russell McMahon Aug 1 '16 at 16:11
• @RussellMcMahon I shall remain satisfied in my level of sadness LOL – Andy aka Aug 1 '16 at 16:16
• His question explicitly identified the sensor type as "magneto resistive". Do you know what a magneto resistive sensor is, what it does, how it works? Links on my answer cover this. You say you voted to close the question because you thought the system would not work with the sensor type you assumed. Olin would say the diagram is too small and he couldn't be bothered reading it. Is that your position? An occasional [tm] bit of backing up when things prove not to be quite as they seemed does less harm than peristence when required information is found to already exist. (Sometimes :-) ). – Russell McMahon Aug 1 '16 at 16:38