Hello every one i am trying to precisely control the position of a BLDC motor. Need an accurate way to do so.With what accuracy can the hall sensor output be used to control the position of the shaft?

Asim, my experience has been that people who want precise positioning DON'T use the motor hall sensors for that purpose. They would use a high resolution encoder, generally an optical type. Only those people who don't need precise positioning use the hall sensors. But to answer your question, Bodine Electric brushless motors have 3 hall sensors and a 4-pole magnet. If you combine the outputs from those 3 sensors, you can read 12 pulses per motor shaft revolution. However, if you are able to read the transitions from high to low and low to high, you can double the resolution and see 24 transitions per revolution.

If you still - after Terry comment - want use Halls for position feedback, the right way will be convert its to standard A,B signals by logic device (state machine). More details about such state machine you can find from description of HP quadrature decoder/counter interface ICs.

Leonid, you reminded me that I was considering speed feedback only and not position feedback. You are correct in that 2 channels would be needed to know direction.

You have to tell us what kind of accuracy you want. Repeatability should be considered too - accuracy isn't the only thing that counts.

Hall sensors can be used for position sensing with the correct Grey code decoding. If using to sense rotor position for commutation, it is not very accurate. Usually good enough for speed control, but not good for torque generation at zero or low speeds.

Thanks everyone for the advice!

My objective is to control the position for some defense applications and NOT for electronic commutation of BLDC. Hence require high precision on upto 0.3 degree.

Encoders such as ENX 16 from maxonmotor has 3 Number of channels and Max. counts per turn is 1024 . That i believe will give me an accuracy of about 0.35deg which I think is good for my present application.

As far as repeatability is concerned ,if i understand it correctly the motor application requires it to go into N number of 360deg turns.

More so what does number of channel denote in an encoder.?

Typical encoders can have 1, 2, or 3 channels. 1-channel can provide a pulse train, which can count in one direction only, or be used for speed sensing. 2-channel(quadrature) encoders allow you keep track of position by sensing direction and counts. 3-channel encoders are just 2-channel encoders with an index pulse that occurs once per revolution. That is mostly used for precise homing.

0.3 degrees requires 11 bit resolution (0.3/360 = 1200 round up to 2048). A 1024 line encoder provides resolution up to 12 bit (4096) after quad decode. You can dither +/- 1LSB and that is 0.175 degrees. for 0.3 degree accuracy that leaves 0.124 degrees for encoder inaccuracy - or 7.5 acr min. You might want to consider going to a 2048 line. (Hope the math is easy to follow).

But unfortunately your referred paper source is not open one...-(

Doesn't matter - you are not going to get 20 arc min accuracy without hardware.

Positioning with the the BLDC hall sensors will not be reliable. An external encoder is recommended if you want accuracy which is what your original statement/comment indicated. Bill K stated it best with common pulse trains that create round number control algorithms. Quadrature also improves resolution and is useful if your application is bidirectional. Other points to consider - do you need absolute position if power is lost? Do you need a marker point for a 'home' position?

Re: correction to last post - Q2: I meant do you need a marker 'pulse' to an external device. This as opposed to a mechanical home alignment position.

If your motor is a constant velocity application, you can get good accuracy by monitoring the back emf of the un-energized phase. With a phase locked loop, I could get one part in 10^4 of shaft rotation..