Sensorless at standstill, development updates and status

[Lebowski]

As you may know from another thread, I'm busy developing sensorless start from standstill with full torque. Basically as if the (BLDC) motor had hall sensors or is using a resolver, but without either of these. This is where I am now:

[fechter]

Wow, that's pretty impressive. Does it still work when you get some phase current going?

[Lebowski]

I had it working with a little bit of phase current about a month ago (before some mayor rework, back then there was still a 50-50 chance of the direction being wrong), there's a video in another thread. I still have to add the capability to drive phase current to this version, and then I'll make a new video. Then from a cold start it should immediately rotate in the correct direction. After that I will go for higher and higher current...

[parabellum]

Cool. Will it be discrete chip managing position or all going into main brain unit?

[Lebowski]

It is all going in the main brain unit...

[district9prawn]

I'm definitely keen to try your new chip when it you begin offering them.

[Lebowski]

Now I added the capability to run with a (15A) phase current. I set a fixed phase current as there is no throttle connected yet.

In the video you can hear me flipping on the 15V/5V power supply (and thus cold-starting the processor). Note how the motor always starts and pulls in the same direction (and with quite a lot of force).

Just to place the 15A, this will generate a magnetic field from the coils about 1/3rd in strength of the permanent magnets.

[csc]

Brilliant !

[liveforphysics]

Wow!!
I've never seen instaspin show that level of control or precision rotor position calculation!

Congratulations my friend on raising the bar for controllers!

[spinningmagnets]

I second what Luke said. I very much like sensorless controllers for big DD hubmotors, and I have one now, in spite of the current state of how crudely they operate. If you can improve them even just slightly, that is a very good thing. Thank you for posting here.

[kiwifiat]

Genius as usual, waiting excitedly for the initial release.

[Lebowski]

cleaned up my table to show the (simple) wiring more clearly. Also strapped it in to try to prevent any motor movement and to measure the pulling power (this is with 15A amplitude phase current again).



At the end of the video I mention I went all the way to the limit of my power supply, must have been between 25 and 30 A phase current....

Its really weird to see this operate. We all know that to have good torque at standstill you need a sensored controller, and then to see a sensorless one do the same thing is just freaky. Even when you (think you ) know in the minutest detail how it all works....

[spinningmagnets]

The type of ebike that is still sold the most today is a geared hubmotor from 250W to about 1400W. When builders try to hot rod them (instead of just getting a bigger motor) geared hubs have a poor heat-shedding path, so they die from heat.

One of the most-often seen failure points is a generic Hall sensor. Although running sensorless can be useful, it does not change the fact that geared hubs are not a good base for high power. Plus...mass production makes existing controllers for these very cheap.

at 3000W and above, direct drive hubmotors are slowly fading out of the off-road market, and being replaced with mid drive, and mid-mount motors.

The one place where large DD hot rod hubmotors are growing in popularity is street ebikes...These builders will pay a premium for a product that makes their life simpler and more reliable.

I am happy enough with my current 12FET sensorless controller, but...I don't know anyone who makes an 18-FET that is sensorless. Kelly has a compact 18-FET in a slender 9X9 configuration but it is a buzzy trapezoidal wave, instead of a quiet sine.

Adaptto uses six large FETs that output as much current as a common-sized 12-FET. The FET model is high efficiency, so it generates less waste heat. If someone made a slender 6X6 using these FETs, it would not get hot at 52V and 60A.

There are currently batteries available that use 4S of 18650 format cells, which is small enough to actually fit in a wide variety of frames. 4S of a 30Q cell can provide peaks of 60A. I would suggest that as a goal for your first controller with this feature.

I would love to see this development spread to every available controller as an option. However, I must be realistic when I state that any new development that is useful will be reverse-engineered and copied. So...your first controller model must be the most profitable to you. 48V/52V providing 60A in a small form factor IMHO is the sweet spot for Version-1

[thorlancaster328]

It's almost like magic, being able to do full torque with no hall sensors!
How are you doing it? My guess is that you feed a electrical pulse into a coil and measure the di/dt and thus the inductance of each phase. Once you know the inductance, you can calculate where the magnets are in relation to the phases and get the rotor angle.
Very ingenious. Too bad I don't have room for 2 more current sensor ICs on my controller board that I'm making.

As for reliability, how immune is it to noise? On a lab bench at low currents is one thing, racing with 50-100+ phase amps is another.

Good luck with your design. I'd love to see a R/C car ESC with such a feature.

[amberwolf]

I still haven't finished building the old version of the controller...and now you have started the ideal version of what I'd like to have on the trike.

Well, I'll finish the ones I have and see how they're doing, then when you finish this design I'll see if there's any way to "patch in" the new brains into the old controllers....

(unless the MCU for the new brain could possibly be "plugged in" (with an adapter?) to the old brain board design. (the one from Whereswally) Somehow I doubt it would be that easy.)

[fechter]

That's pretty amazing. It really proves it is possible to make sensorless control work as well as sensored. I'm totally impressed.

[Lebowski]

I still haven't finished building the old version of the controller...and now you have started the ideal version of what I'd like to have on the trike.

Well, I'll finish the ones I have and see how they're doing, then when you finish this design I'll see if there's any way to "patch in" the new brains into the old controllers....

(unless the MCU for the new brain could possibly be "plugged in" (with an adapter?) to the old brain board design. (the one from Whereswally) Somehow I doubt it would be that easy.)

I think a retro fit should be possible as the through hole PCB contains necessary systems (I mean no matter what you need power supplies, output stage with drivers, current sensors etc). It would be something like this though:

stamp.jpg (98.14 KiB) Viewed 3863 times

to make a translation of a 64 pin TQFP plus the additional analog processing down to the 40 connections of the DIL package (the 40 pin DIL has connections to all necessary signals, also for what I'm building now... well so far at least).

But to go from where I am now to a complete controller IC is still a long way of...

[thepronghorn]

Wow. This is amazing.

[amberwolf]

I think a retro fit should be possible as the through hole PCB contains necessary systems (I mean no matter what you need power supplies, output stage with drivers, current sensors etc). It would be something like this though:
stamp.jpg
to make a translation of a 64 pin TQFP plus the additional analog processing down to the 40 connections of the DIL package

I'm ok with that. I don't have any particular space or shape limitations atm, since these are my SB Cruiser trike controllers.

My main goal is just to have reliable smooth quiet performance that can push enough power thru the hubmotors to quickly accelerate me up to the 20MPH max, even if I'm carrying a dog or pulling a trailer full of heavy stuff (dog, food, piano, metal, wood, etc).

Right now with generic controllers I have most of that, but not quiet or smooth, especially with the one sensorless on there, and reliable presently depends entirely on the hall sensors/wires/etc in the one sensored motor/controller to stay working (otherwise it doesn't startup from a stop very well unless I pedal it up to get moving, which is really hard and painful and thus the whole reason for the motors to do it for me).

So...will be awaiting the eventual release of the new chip.

But to go from where I am now to a complete controller IC is still a long way of...

That's ok, too--I'm a long way off from a complete controller build.

[emmgee]

I wish I knew so much about physics, electronics etc. to push things further than global companies like Texas Instruments are capable.
Just fyi: I stumbled across a video explaining the TI Instaspin technology, I think I haven't seen it on the forum yet: https://youtu.be/szgVUfyX8JM

[Lebowski]

Trying it out with my small (I think 4025) RC motor.

With phase current and the computer showing the real time (rotor position) phase info.

Note how the phase info is continuous and not just 6-stepping.



I added some AGC's inside the code to make it easier to try different motors. I tried all 3 iron core motors I have with all of them having the sensorless at standstill algorithm working as good as with the MAC. Turns out the MAC I started with was actually the most difficult one giving the weakest signals to work with. The RC motor was the best with the big 10 kW rated Enertrac 602 in the middle.

[Lebowski]

I would love to see this development spread to every available controller as an option. However, I must be realistic when I state that any new development that is useful will be reverse-engineered and copied. So...your first controller model must be the most profitable to you. 48V/52V providing 60A in a small form factor IMHO is the sweet spot for Version-1

I think once its out there and indeed does what people want, you're correct, everyone will try to reverse engineer it and some will succeed. And then you'll start seeing more controllers with this feature...

I've thought about patenting but don't believe in it. It will be costly especially when the patent lawyer will want to split the system into multiple patents. And patents are only as good as the lawyer you can afford to defend them... Plus China laughs at patents, and will be happy as a patent will be a good description of how it all works. Patenting makes sense for a big company as they are used in negotiations with other big companies, not so much for actually protecting ideas.

I'm still not really into starting building and selling complete controllers, though I do think this is the type of tech to start a company on. Looking at Adappto, you have to deal with customer returns, questions, warrantee, legal battling you competitors etc etc. I am a Research guy who likes to tinker building stuff, and then throw it over the wall when stumbling across something useful. All the other stuff associated with a company would keep me awake at night, and give me nightmares if I do manage to sleep...

[Lebowski]

I wish I knew so much about physics, electronics etc. to push things further than global companies like Texas Instruments are capable.
Just fyi: I stumbled across a video explaining the TI Instaspin technology, I think I haven't seen it on the forum yet: https://youtu.be/szgVUfyX8JM

I looked at this video and it is indeed the integrate voltage to get flux method. I think a good method for when you are 6 stepping, nothing wrong with it.... well as an analog engineer integrators are kind of scary in a sense that you always have input referred offset and the thing always runs away. If you build custom chips like TI does then there are many tricks to keep the offset of integrators extremely low and to get it to work. But it will not work at standstill. TI also mentions in the comments that this is not their standstill FOC method...

[madin88]

Trying it out with my small (I think 4025) RC motor.

With phase current and the computer showing the real time (rotor position) phase info.

Note how the phase info is continuous and not just 6-stepping.

Impressive how accurate it tracks the position. Awesome!!
Do you think that it will work at very high phase currents as well?

[Lebowski]

For what I know it should work at higher phase current as well, but it will put some demands on the electronics (especially the current sensors).

I think it'll be a case of just trying and seeing when and why it fails, and then fix that. At the moment though it will go to the max of my power supply, so the MAC motor I tested with 26A amplitude phase current which was the max my power supply would allow.

At these higher current levels I want to start using a throttle and maybe some error detection, so stuff to add just to keep testing safe...