Make Life Accessible - Motor Suite Test - blog 8

How do you think will Hall effect sensors help you at lower speeds? The state of the Hall effect sensors changes every 60 degrees of electrical revolution. This is of little help to FOC algorithm. They may help you if you are using trapezoidal control as back-EMF zero crossing detection is not reliable at low speeds, but FOC algorithm will probably use some other means of sensorless rotor position estimation.

The way I understand it, BLDC motors require sensing to know when to commutate the phases. When the speed is high enough, there is enough back-emf in the un-powered phases to deduce rotor position and figure out commutation timing. When rotation is not fast enough for significant back-emf or perhaps when no computer is available, Hall sensors can be used instead. NXP has an application note that might help explain it better:

http://www.nxp.com/files/microcontrollers/doc/app_note/AN4058.pdf

I don't know if it will work better, since I have never done anything tricky with a BLDC motor, but I seem to recall reading somewhere that hooking up the Hall effect sensors would provide better low speed control.

Doug

Hi Doug,

The hall sensors should always allow you to know which phase in the commutation cycle you're at (i.e. even with a stall), whereas with the sensorless method you won't know at very slow speeds/standstill.

Good luck with your project, and super-interested to hear more about the control algorithms. I used a BLDC chip for driving a DC motor a while back, and used the sensor feedback for speed/position determination. However it was an old-school method of control (using analog integration of pulses to determine speed). Modern methods using microcontrollers/DSPs are far more advanced. The FOC method kulky64 mentions is quite new I think, it may well use things other than back EMF for sensorless control, but it is beyond what little I know about BLDCs. While the hall sensors may not help with that method (I have no idea), if back EMF is being used then at slow speeds there could well be issues.

I tried to write up some short motor notes a while back, here and here - in case any of that is useful or gives ideas.

All true, but applies to BLDC motor with six-step commutation (sometimes called block commutation). But you are dealing with PMSM motor and Field Oriented Control (FOC) with all three phases driven at the same time with sinusoidal current. From quick scrolling through Kinetis Motor Suite user's guide it seems that KMS does not support six-step trapezoidal control.

I'm guessing it must still be performing some measurement (if not back-emf then some other characteristic) otherwise there is no closed-loop. It is still likely that that measurement (whatever it is) is impacted at slow speeds/standstill, but it is just a guess. I don't know anything about FOC : (

Or is it treating it like a stepper motor, i.e. not closed loop at all?

Of course it does rotor position estimation. Accurate position and velocity are essential for FOC algorithm to work. At medium and high speed, a BEMF observer in d/q reference frame is usually used. At low speed, some other techniques like high frequency injection or open-loop start-up are needed to spin the motor until BEMF is sufficiently high for the BEMF observer. Usually 5% of the nominal speed is enough for proper operation in sensorless mode. I didn't check how things are implemented in Kinetis Motor Suite though. Of course things are easy if you have proper sensor (either encoder or resolver), but i'm afraid hall effect sensors won't improve situation much. 60 electrical degrees of resolution they provide is simply not enough for FOC to work. Don't know if low speed operation is needed for Doug's project.

Of course it does rotor position estimation. Accurate position and velocity are essential for FOC algorithm to work. At medium and high speed, a BEMF observer in d/q reference frame is usually used. At low speed, some other techniques like high frequency injection or open-loop start-up are needed to spin the motor until BEMF is sufficiently high for the BEMF observer. Usually 5% of the nominal speed is enough for proper operation in sensorless mode. I didn't check how things are implemented in Kinetis Motor Suite though. Of course things are easy if you have proper sensor (either encoder or resolver), but i'm afraid hall effect sensors won't improve situation much. 60 electrical degrees of resolution they provide is simply not enough for FOC to work. Don't know if low speed operation is needed for Doug's project.

There is an Atmel doc here http://www.atmel.com/Images/doc32126.pdf

page 15 (figure 4.9) looks like they can use a Hall sensor, but just a single one presumably to compute difference between the estimated position and the actual position.

I'd originally thought Doug's system used the typical commutation method but sounds like FOC is a controlled variant of sinusoidal drive like stepper motors.

Hall sensors don't provide 60 degrees of resolution in a BLDC as far as I understand (or maybe there are variations). They will latch approx every 180 degrees, i.e. at repeatable positions per revolution, so you know as it rotates when it passes a certain physical position - that would give you the accurate position and velocity you mention) just that there are three of them so they all operate approx 60 deg out of phase. I could be wrong - I'm no expert on BLDC nor on how flexible FOC is - e.g. if it can (or needs to) resolve the error more frequently than every revolution, nor on how slow a speed is needed by Doug.

Right. Hall effect sensor changes its state every 180 degrees of electrical revolution. Place three hall effect sensors 120 degrees (or 60 degress) apart and bam, you know absolute position with 60 degrees resolution. Check Figure 67 in this document:

http://www.st.com/content/ccc/resource/technical/document/user_manual/5e/5e/d2/cb/07/35/45/a6/CD00298474.pdf/files/CD002…

(This document also describes in Section 8.2 High Frequency Injection technique i mentioned earlier.)

I am tracking the sun - which is obviously pretty slow at one revolution per day, but I can use short, higher speed, motions periodically to emulate continuous tracking. I still will need to gear the motor down significantly. Its minimum speed right now is 288000 revs per day and I need 1 rev per day.

I think we're talking about the same thing. Section 4.3 of that document mentions a hall sensor can be used. Presumably it waits for a transition from high to low (or vice-versa) to get a repeatable precise position fix, that repeatable position would occur every  180 degrees or every 360 degrees (depending on what that algorithm needs).

Oh that's quite a bit of reduction. I know of some UK suppliers of gears (e.g. worm gear sets or just spur gears) or timing belt, that will get you perhaps 50:1 or 100:1 max reduction, so several stages of it would be needed (i.e. 3 or 4 stages). Would need sub-millimeter accurate 3D printing or some adjustable capability so you can put in the right place and tighten screws.

If you need the details let me know, I'll dig out the names I've ordered from them once or twice. But maybe you know of a supplier closer to you.

But the other method you suggest has a lot of value, i.e. no need to run at the slowest speed, and only power it on and off intermittently, e.g. power it on every minute and just count the revolutions using the hall sensor if the FOC algorithm implementation doesn't have a hook to program how many revolutions you need. I did that with a brushed DC motor, and the error was tiny (and the error won't accumulate since it is possible to compensate in each subsequent minute).

Edit media

Dimensions x Small Medium Large Custom

Subject (required) Brief Description Tags (separated by comma) Video visibility in search results Visible Hidden

Parent content

Getting Started with the Infineon DC Motor Control Board Comment on Make Life Accessible - Motor Suite Test - blog 8

Poster

Upload Preview

Ok

Cancel