MLX90217 reliability problems
Alex.Stanfield
Posts: 198
Hi, I'm using two MLX90217 to form a quadrature encoder against a steel gear (modulus 4, 10 teeth, ~48mm diameter). I usually get it working ok on the test bench but I'm having problems when mounting the whole assembly (which has iron supports). The problem is that it randomly generates pulses in "the other direction" i.e: the gear moves in one direction and so do the quadrature pulses, but then the pulses get out of sync creating a sequence like if the motor was moving in the opposite direction.
Has anyone used these sensors with regular gear teeth (instead of square teeth I mean)
Thanks in advance for any tip
Alex
Has anyone used these sensors with regular gear teeth (instead of square teeth I mean)
Thanks in advance for any tip
Alex
Comments
I'm using a strong rare earth magnet 8mm x 1mm.
Has someone made a quadrature encoder sensing directly on the gears? On the picture the MLX90217 go one on each end of the 27,5mm dimension line giving 1 and 3/4 teeth separation.
Thanks in advance for any experience you may share.
Alex
Yes, I'm wondering the same. On the test bench I hardly had any errors. I'm not able to get rid of the iron (for security reasons) so what should I do to re-gain control of this?
The leads I'm following on this problem are the following:
- Miss-alignment (i.e: incorrect distance) between the sensors affecting the overlapping of the quadrature signals.
- tooth shape that makes the sensing more difficult
- incorrect gap between sensor and tooth
- magnetic flux too strong
- magnetic hysteresis on the gear teeth affecting the sensing
- supporting iron of the assembly affecting readings
Alex
Thanks for the tip, maybe a set screw will do fine and it's easy to assemble.
Alex
As "data trumps guessing, always" I decided to get a linear hall sensor and get some hard measures on the subject. Turns out that I had some points way out of specs.
- The bias magnet was over 400mT which is the max allowed on the MLX90217 face
- The air gap between tooth and sensor is way beyond the values i saw in Melexis' forum and testing showed that at that distance was very hard to get the min 10mT required by the MLX90217 to detect the edges.
I had the sensors behind a plastic "fence" to protect and support the sensors, I'm changing that to put them in front of the fence and close enough to the teeth also with less bias flux.
Will update after new tests, stay tuned.
Alex
The solution came around two things
- Use the correct bias magnet.
- You should have around 1mm of air gap between the teeth and the face of the MLX90217
I finally got the chips in front of the plastic support with 1mm gap and haven't had any more problems so far. Between the "fence" and the gear in the following diagram.This magnet http://dx.com/p/super-strong-rare-earth-re-magnets-8mm-20-pack-4248 has a little above 100mT which is perfect for the MLX90217. You can use two if you need for spacing and it will work ok
The closer the better. This is crucial to get the reliability for a quadrature sensor so don't be fooled into thinking that beacuse it works on a 3/4mm gap it's ok.
PD: Be wary however that sensing magnetic flux variation is not the same as and optical encoder. It worked for me but even slight changes on the gear's position can change the output even if the sensor didn't get to the face of the tooth. i.e: you can sense a "magnetic edge" because you turned back slightly but not as far as getting to the "mechanical edge". This can send false signals to your code making it believe it moved more than it really did. this happens on direction changes only as far as I saw. Moving further in the same direction get's your signals correctly, it's just on the direction change.
Alex