Re: Did anyone used linear hall sensors on BLDCs for position sensing?

HI AndrewBCN,

Good points, let me elaborate on them.

Quote
AndrewBCN
1. The price of adding the many required linear hall effect sensors to any brushless DC motor assembly is not negligible. First, quite a few sensors are required, second, they must be precisely positioned, and third, each additional sensor requires wires which have to connect to the closed loop control circuit. And of course, each sensor/wire/component adds one more failure point to a mechanical device designed to operate thousands of hours. All that is not easy to do cheaply.

I have been using optical incremental encoder successfully but motor manufacturer suggested this idea that would replace the three Hall sensors used for commutation by three linear Hall-effect sensors. Depending on rotors number of poles the sensors will recreate more than one cycle per rotor revolution. If an encoder can be had for $3, I expect this change to be cheaper than that.

Quote
AndrewBCN
2. The closed loop control circuit for each brushless DC motor appears to require as much (real-time) processing power or even more than our very basic 8-bit AVR MCU - that presently controls an entire 3D printer - can provide. So basically for a five-stepper Prusa i3 we would require five 32-bit MCUs just for the control loops. And again, we are adding quite a few more failure points to our 3D printers.

It is true that field oriented control, using Clarke transform, can be too much for an 8-bit microcontroller. We can use a not so fine control or stick to inexpensive 32-bit MCUs. One advantage in this application is that high-speed is not needed, which cut us some slack.
[/quote]

Quote
AndrewBCN
3. One of the details that the paper brushes over (pun intended) is that each individual linear hall effect sensor needs to be calibrated separately and requires a properly designed analog amplifier circuitry for signal conditioning. Again, more costs and more failure points.

This may be the reason that could kill the idea at the start: If such a calibration cannot be done easily by either the manufacturer or by the motor controller, then we might have a problem. Given that PID controller will need to be adjusted for each motor anyway, I would assume that a self-tunning mode will need to be triggered on the first use of the motor or every time it is added to a different axis/printer. Hopefully such self-tunning should include a Hall sensors calibration process.

Quote
AndrewBCN
Of course I find the idea of closed loop control of rotary or linear motion actuators extremely attractive, but I am guessing the reason servo motors for e.g. CNC applications are so expensive is that there is no known way to make them any cheaper.

It's getting cheaper lately, on one hand Tecnik Clear Path line is getting more affordable for CNC applications, and it can get much cheaper if you use smaller motors and some ingenuity, like adding an encoder with a sensorless brushless motor [cache.freescale.com] . One implementation example for CNC may be this [www.cnczone.com]

Anyway, my point is to open the discussion and the exchange of ideas and experiences. If it were easy it alredy would be done (:P)

I already have a working solution for either brushed DC motors or for brushless motors with built-in encoder and drive electronics that I shared openly in the link above within the cost limit I mentioned, but I am interested on alternatives that could be cheaper and/or better.