Hi panyuming,
This was problem I saw. So make air gap larger at the stator tooth than air gap at radial director. Length of air gap is radial distance between iron face and magnet. The longer gap will have less force. Having lower force at the outer gap will allow rotor to move with the force at the inner director air gap.

The other variable influencing force at the air gap is the area or coverage of magnet face/tooth face. Per arc distance (or degree of rotation), this area decreases more at the stator compared to the inner director. Since this case uses same magnet (rectangular), at opposite ends, displacement at the ends are not equal, and affect each other due to common magnetic circuit.

I like to think the limit case. If outer gap is very large, there is no resistance to movement at inner gap. As outer gap decreases, attractive forces between magnet/stator tooth start to oppose movement of rotor to follow the radial director rotation. But the radial displacement of inner radial director/magnet will affect flux & force of outer magnet/stator tooth. At some point, attraction of director to magnet is just a bit greater than attractive force of magnet/stator tooth. Then rotation will start and continue, with cogging.

It is a complex puzzle to me.
bi

MOTORFLUX "kick-back"

I'm sure most are aware of the MotorFlux kick-back but I'll document it anyway.
Probably the reasoning behind the use of "one-way bearings."

Attached is a pdf of a series of snapshots from "The Motoflux Principle: How it Works"
video by Mike Corbin - https://www.motofluxpower.com/

SL

Here's a nice graphic of cogging torque per radial position.
From:
Torque Ripple of Permanent Magnet Synchronous Torque Motor
Autor: Ing. Jan Moravec
Vedoucí: Doc. Ing. Pavel Souček, DrSc.

bi

Motorflux Torque Curves

It appears the " Torque Curves" provided in Corbin's "Proof-of-Concept" video
follow the theoretical curves given above.

In particular the "Control Motor Input Torque" - attached.

The variance in RPM and Load is also somewhat apparent.

Attached a pdf for easy comparative study as well.

SL

Quote from: SolarLab on March 17, 2023, 08:40:28 PM
Motorflux Torque Curves

It appears the " Torque Curves" provided in Corbin's "Proof-of-Concept" video
follow the theoretical curves given above.

In particular the "Control Motor Input Torque" - attached.

The variance in RPM and Load is also somewhat apparent.

Attached a pdf for easy comparative study as well.

SL

Thanks SL,
I had not previously noticed that first chart in your pdf labelled "Start-up". I find it very interesting. One assumes the torque sensors are recording as the control motor is started. You can clearly see on the lower trace (torque input lb.ft.) a linear ramp up to ~0.5 lb.ft. and then a sudden break downward. I can't read the sample point on the x-axis, but you can see the vertical grid up through the upper trace. That upper trace remains steady right at 6.0 lb.ft. from the start to that grid line at the point (time) which input torque starts to decrease. As the input torque decreases (lower trace), the output torque also decreases (upper trace). While input torque decreases linearly about -0.65 lb.ft., the output torque decreases a bit less, about -0.5 lb.ft. slightly curved. From there on, the traces oscillate as expected of cogging torque, the lower trace around ~0 average and the upper trace around ~5.7 lb.ft. The magnitudes of the torque ripple for both curves appear to be about the same.

Very strange, in my opinion. It leads me to believe that output torque sensor was biased to 6.0 lb.ft. in the static condition prior to and throughout the test. Then that the measured and recorded input and output torques are approximately equal in magnitude.

Just offering my observations. I'll post a copy of your pdf chart for convenience. Also a copy of the cogging torque calculation from the pdf in my prior post.
bi