Did some testing with my encoder concept.
@Manh_Ti30122.5K followers
66,723 views
28 Comments
Comments from the original post on X (Twitter)

@EERandomness
The only difference between your design and "professional" encoders is probably in the firmware. There are dozens of control theory tricks you can use to improve the performance, like characterization, a "PLL", differential auto-gain, an estimator or even a Kalman filter. For magnetic sensors, estimating the axial and radial offset is very important, not just the rotation
8 likes

@Manh_Ti3012
yeah, my firmware is still primitive and simple compare to industrial one thanks for the suggestion, I'll try to implement them and see.
4 likes

@EERandomness
Prompt for Astra: "Build a real-time fixed point control loop for . It should implement a Kalman filter using the raw ADC values from each hall as input, and also use a pre-calibration LUT for phase and amplitude offsets. A second LUT should be used for stator magnetic field variation compensation. Calculate an auto-gain term based on the max/min hall value over 1 rotation closure. Finally, build an estimator that estimates the speed, angle and acceleration of the stator, compensating for offset, pitch, yaw and roll. At high speeds, use a PLL for accuracy."
3 likes

@boxcardavid
Engineers who’ve worked extensively with magnetic ring encoders
7 likes

@boxcardavid
Awesome. How sensitive is it to air gap changes?
5 likes

@digitelektro1
Amazing! I could use it on my satellite dish rotator for position feedback!
2 likes

@tonyaajffleury
This is awesome! What kind of magnet-sensor separation can this design tolerate, and what magnet size & grade are you using?
2 likes

@Manh_Ti3012
I'm using random 4mm neodymium magnet I bought online. From my testing, as long as you have a strong magnetic field that cover at least 2 sensors at once, it will work.
7 likes

@J_B_N_P
Cool. You could also use the concept used by digital calipers. Cheap and precise. Analog and SW might be a bit more challenging though
1 like

@Manh_Ti3012
The problem with this is that they are capacitive and not absolute encoder, maybe need precise mounting too.
2 likes

@VarnerRants
considering some 10 bit encoders go for $32 in bulk quantity, closer to $50 in singles I'd say you're doing great here. Not to mention the 'customization' capable and ability to integrate it into the design rather than bolt on. Nice.
1 like

@ck401uni
すごい!
1 like

@DomingoRivelino
Meus parabéns amigo, um projeto desse facilmente pode ter N utilidades... Com as possibilidades e a tecnologia de hj podermos criar muitas coisas fantásticas, eu tbm tenho muitas ideias mais por hora fica só na parte de projeto mesmo
1 like

@GhostOfFrederic
Cool! I still think variable capacitance rotary encoders are the GOAT. Like
1 like

@MaxClerkwell
This is absolute cool! Thank you for sharing!
1 like

@PcbBuy
“Promising results for a $20 prototype! Have you tested how sensitive the readings are to slight misalignment between the PCB and rotor?😀

@HenkPoley
Is that 11.8 bit = `log(365/0.05)/log(2) - 1` ?

@sadieOX360
The prototype look a bit too big , wht if tou you want to make one like this but of a total radius of 2-3cm ?

@deeperflows
Really neat, well done

@sahil_harriram
How are you calculating angle from the Hall array? Are you interpolating between the nearest sensors, fitting the magnetic field across the whole array or using some kind of calibrated lookup model?





