Not simple BLDC controller It RUNS! :)

[oldswamm]

That last post confused me.
2 comparators, could be on the same shunt, bit more complicated with 2 shunts instead of 3.
<edit>3 shunts? One, optimized for speed, for HW shutdown, the other two optimized for accuracy, for phase current.<endedit>
One comparator would drive the enable pins on the drivers, to off.
The other would be connected to the fault input on the PIC, and would be set to a lower current than the first.
Normally the first wouldn't trip because the fault input would take care of it, and the program would probably temporarily reduce the pulse width so it didn't happen again?
If the first trips, it would probably indicate a 'fault' of some sort, and, if there's an appropriate input, the processor would act accordingly. The fault input keeps the PWM off for the rest of the cycle, so the program would have time without resorting to RC delays like I implied in the last post.
The big question is 'how fast is the enable pin based shutdown compared to the fault input on the processor'. I find no delay specs for the fault input on the dsPIC30F4012. OK, I did find it in the 3011 pdf, 50ns.
I think maybe I'm still confused. Is redundant fault protection needed, or is the driver delay the problem, or just in case of processor or firmware faults.

[Arlo1]

LOOK what came in today! Man im going to need a lot of chocolate to do all this thinking... luke??? Help?

[Lebowski]

cool stuff let us know how you get on with the included course, I'm very curious as to how good and elaborate it is.

[Njay]

It's a piece of cake Arlo, you'll see

[oldswamm]

It's a piece of cake Arlo, you'll see

Does this refer to the software, or getting it into the PIC?
<edit rather than make a new post>
Just wanted to mention to those who don't realize it, the ADCs on the PICs Arlo intends to use can do 2 simultaneous conversions at 500khz. Up to 4 @ 250khz. (Actually samples, the conversions are at 1mhz.)

[Lebowski]

It's a piece of cake Arlo, you'll see

Well, , in a 16F88 I run a control loop to control the current through a motor (so a current
based throttle control). This comprises a 1st order integrator implemented with 16 bit fixed point
math (including fixed point multiplications and in addition to this, for speed, an eeprom table),
this runs at 4 kHz sample freq. On top of this (via interrupts) it runs a noise shaper at 10 kHz sample frequency.

Definately not a piece of cake but lots of fun (and definately a case of: you can make stuff
as complicated as you want)

[Njay]

I was referring to learning to develop software

Lebowski, try an AVR, you'll like the 4x+ speed increase over the PIC

[Lebowski]

I was referring to learning to develop software

Lebowski, try an AVR, you'll like the 4x+ speed increase over the PIC

With the AVR I really dislike their datasheets, impossible to read while those of PIC
are very clear. Plus I didn't find any programmer on their site, same for good
assembler design environment...
Plus Microchip gives you free PIC's if you ask nicely, haven't found this option
on the AVR site...


By the way, with the 30F2011 and a standard motor stage (using only 2 out of 3 motor wires)
you can build a wickedly powerfull (kilowatt) audio amplifier

[bigmoose]

Njay, I would recommend a "system comparison." For a complex PWM controller we are looking for a combination of system features. First a PWM section optimized for 3 phase control. Deadtime inputs, period centering, etc.; then we need a good, fast AtoD, then some decent DSP features in single cycles for the PID loops, followed by user interfaces like free input pins RS232, USB, CAN features. Take the device I recommended and do a relevant feature comparison to your AVR's (which honestly I am not experienced with as one head can only hold so much manufacturer data... ) I am open to new an better devices!

[Njay]

BigM, my "AVR try" suggestion was only a comment on Lebowski's PIC16F88 project. I could rebate all of your points except the free samples Lebowski (which is the only advantage 8bit PICs have over 8bit AVRs, not considering some special peripheral a certain PIC model might have better over AVRs), but I suspect that would leads to kind of an avr vs pic argument

Nevertheless I guess this project already has a platform picked up, Arlo already invested in one. In the end I guess there won't be relevant BLDC control hw differences in comparable platforms.

[bigmoose]

Njay, know that I am open to new stuff!! The AVR/PIC/Cygnal/etc. debate are sort of like the Ford/Chevy arguments of my youth. You tend to stick with the devil you know, rather than the "other" one.

But if you have a recommendation that is superior to the dsPIC30F3010-30I/SO I for one would be interested!

[Lebowski]

Njay, know that I am open to new stuff!! The AVR/PIC/Cygnal/etc. debate are sort of like the Ford/Chevy arguments of my youth. You tend to stick with the devil you know, rather than the "other" one.

But if you have a recommendation that is superior to the dsPIC30F3010-30I/SO I for one would be interested!

At a certain point you'll have to settle for 'good enough', else you keep looking for newer and better parts
but never build anything. And I agree with Njay, it's the algorithm that counts not the processor on which
it's implemented.
It's definately more fun just to dig in, build stuff and collect ideas /keep your ideas for the next version
during the current build.
Why did you choose the I/SO version (small outline, difficult to solder) ? I always go for the standard 0.1" pin
distance DIP versions... I cannot solder SMD's or really small IC's

[Njay]

I don't have any recommendation at all for the task at hand, I've never done nor studied BLDC motor control. I know only that all vendors have chips oriented to that task. If this were a "mainly software" project, then I would "blindly" recommend an AVR (if decided to go with 8bit platforms), it has a fast core and free/open source development tools (including simulator).

By the way, is there a big difference between this thread's controller and this one's here?...

[Arlo1]

By the way, is there a big difference between this thread's controller and this one's here?...

Yup Im not just going to talk about doing it

No I wanted my own thread so I can build it the way I want to build it, and not have to fumble through all that rambling to find the little details. I will also edit my first post to show the scematics and video and parts list ect.

[bigmoose]

Why did you choose the I/SO version (small outline, difficult to solder) ? I always go for the standard 0.1" pin distance DIP versions... I cannot solder SMD's or really small IC's

I just copied and pasted from my stock list. I have the capability (microscope, solder station, hot air rework tools) to do surface mount parts and in fact part availability and packaging often requires them. If you are breadboarding, the through hole versions may be the more appropriate choice for your build.

Also as a general note, you will find that you need the tools in the toolshed for the firmware to exercise. Any processor can be "programmed" to run an air conditioner fan motor, but if Arlo goes where he wants to go, he is at the cutting edge of development. Low inductance, low resistance 3 phase motors have severe problems. He is in the quest of a $10,000 controller, not a $100 one. This will be a long, hard journey of many steps. It will need many hands, so we need you all here!

[oldswamm]

Njay, that's the thread that drew me to ES in the first place.
Deteriorated into a debate about processors, drivers, etc. Read it from end to end. Lots of info (scattered amongst the chaf), but no building.
The DIFFERENCE is, Arlo's actually building something!
<edit> Thought this was going to be right below Njay's or I would have quoted.
I think putting 'BEST' in the title was the biggest problem with that thread, there is no such thing, unless maybe you specify exactly what it's best for, and then somone else will build one that's better, at least in their opinion. <endedit>

Big M, the reason they debated ford/chevy is they all knew Mopar was superior, right Arlo?
You're right, the reason I use PICs is that that is what I've always used, and will use unless there's a REASON for me to go through the learning curve to change. Were I part of a group effort to design a controller, I would go with the consensus, (if one were possible), but if I designed it, I would use what I'm familiar with.

[bigmoose]

Lots of good points being made!

That said, I would recommend that we not focus on all the bells and whistles and interfaces here at the start. No CAN bus, no logging, no fiber optic interface to a remote display. Start with the brute force core. Keep It Simple Stu... KISS!

First decide if it is PWM with over current limiting, where the throttle input sets the PWM period and you only abort on a blow up the FETs overcurrent; or true phase current control to terminate the PWM cycle, where the throttle sets a current threshold that will be met irregardless of the PWM period required to achieve it.

Then design the micro as bare bones as you can on one board, perhaps with the main power supply bus. Decide if this power supply bus is 24V, 15V or 12V.

Design the 1st round FET power stage, perhaps a 12 or 18 FET so it's relatively cheap when the smoke gets let out.

Rough design the ultimate FET power stage for the Collosus motor.

Design the driver stage to drive both, with the isolation power supplies. Now only component changes are needed for the lightweight FET power stage. The basic driver will drive both.

This is cumbersome and not optimum; but will give you "building blocks" for experimental development. When you have it working, you can add the cute user interfaces and roll it into two boards.

[oldswamm]

First decide if it is PWM with over current limiting, where the throttle input sets the PWM period and you only abort on a blow up the FETs overcurrent; or true phase current control to terminate the PWM cycle, where the throttle sets a current threshold that will be met irregardless of the PWM period required to achieve it.

Thank you, I think this is the first I've understood exactly what was meant by a current control throttle. Obvious once you describe it concisely. (Knew what was wanted, but was vague on how to achieve it.) At it's simplest (with no processor intervention), the 'over' current comparator could have its reference connected to the throttle, and it's output to the fault input to the processor? All the processor would do then would be to provide commutation and the pulse period.
I would assume some analogue controllers work like that? Haven't honestly thought about analogue BLDC.
And I still don't think he/we should do anything because it's stupid, lets 'keep it as simple as possible'. (sorry, couldn't resist)
Bob

[Arlo1]

Lots of good points being made!

That said, I would recommend that we not focus on all the bells and whistles and interfaces here at the start. No CAN bus, no logging, no fiber optic interface to a remote display. Start with the brute force core. Keep It Simple Stu... KISS!

First decide if it is PWM with over current limiting, where the throttle input sets the PWM period and you only abort on a blow up the FETs overcurrent; or true phase current control to terminate the PWM cycle, where the throttle sets a current threshold that will be met irregardless of the PWM period required to achieve it.

Then design the micro as bare bones as you can on one board, perhaps with the main power supply bus. Decide if this power supply bus is 24V, 15V or 12V.

Design the 1st round FET power stage, perhaps a 12 or 18 FET so it's relatively cheap when the smoke gets let out.

Rough design the ultimate FET power stage for the Collosus motor.

Design the driver stage to drive both, with the isolation power supplies. Now only component changes are needed for the lightweight FET power stage. The basic driver will drive both.

This is cumbersome and not optimum; but will give you "building blocks" for experimental development. When you have it working, you can add the cute user interfaces and roll it into two boards.

Yup thanks that's it I did not want the simple controller design with out programing capabilities but I didn't think the best controller design in the world was needed either. I just want something I can control and program and understand. This is the second ever circuit board I have ever built from scratch! And will be my first time writing code from scratch. I fumbled with megasquirt and got some rough coding figured out but burning gas is so lame now that I have my true EV taste for a couple of years. I made a knock sensor curcit from scratch and it works well I am almost tempted to use it to the Stop pin on the fet driver but a flip flop looks better and simpler. Now I just need to figure out how to implement it!

[Lebowski]

I think you guys should start with something simple. Seeing as there's limited or no experience with the
PIC microcontroller... My first try was to make a LED go on / off based on a push button switch. Second
project was to connect the motor with 3 NMOS transistors (via drivers) to the PIC, force commutate with
speed control via the PIC's clock frequency (clock was coming from a variable frequency signal source).
Then I tried simple sensorless commutation by detecting ground supply crossing. Only then I added PMOS
(or high side NMOS) and PWM.

How much understanding do you guys have of how a motor works ? Are you at the level where you understand
that a (3-phase) motor is like a triangular wheel with 3 spokes ? A bit like a Wankel motor ? Rotating vectors ?
Do you know why 3 low-side NMOS transistors would be enough to run a motor in WYE config ? All this is
stuff you need to know...

I found taking baby steps a lot of fun as every evening I had something cool to try which had me hopping
on my seat with excitement (yey it works ). If you try to implement all your wishes in one go you'll
first spend 3 months coding followed by a big disappointment when you finally throw the switch and find
out it doesn't work.

[Arlo1]

I think you guys should start with something simple. Seeing as there's limited or no experience with the
PIC microcontroller... My first try was to make a LED go on / off based on a push button switch. Second
project was to connect the motor with 3 NMOS transistors (via drivers) to the PIC, force commutate with
speed control via the PIC's clock frequency (clock was coming from a variable frequency signal source).
Then I tried simple sensorless commutation by detecting ground supply crossing. Only then I added PMOS
(or high side NMOS) and PWM.

How much understanding do you guys have of how a motor works ? Are you at the level where you understand
that a (3-phase) motor is like a triangular wheel with 3 spokes ? A bit like a Wankel motor ? Rotating vectors ?
Do you know why 3 low-side NMOS transistors would be enough to run a motor in WYE config ? All this is
stuff you need to know...

I found taking baby steps a lot of fun as every evening I had something cool to try which had me hopping
on my seat with excitement (yey it works ). If you try to implement all your wishes in one go you'll
first spend 3 months coding followed by a big disappointment when you finally throw the switch and find
out it doesn't work.

The pickit3 I got has 12 lessons the first one is to turn a LED on and off. I Am pretty sure I understand brushless motors enough. Its not like me to take small steps its more like me to bite off more then I can chew then chew the shit out of it Just ask methods my first email to him after ordering a 18fet kit a couple years ago just after joining ES. He said well how did you set it up on your other e-bike. I said other E-bike? If I wanted a Ferrari I would not buy a fiat 500 to get a feel for it first! He laughed then told Luke this stupid canuck will never figure out this controller.... A year later Luke told me he all this and said he thinks I built the best 18 fet controller. At that time I was fumbling though it. Now I feel like I am training for this and I have a great feeling its going to work. I will start will all this once my brothers bmx is finished most my parts are in for the controller other then the bread board for prototyping.

As for how long it takes well learning to write code is something I always wanted and there is also lost of other publicly listed BDLC code to use bits of to see what I think will work best. I have had a KICK ASS motor for over a year that nothing in the world can through real power at. (I don't thnk Kelly is going to work well) Its time to build somthing than can do what I need.

[Lebowski]

I looked at the lessons coming with the PICKIT... they're a good start but all written in C. Have you thought about which
programming language to use for this project ? C will be easier to code I assume (does it do fixed/floating point math
in C or does the C-code need to be close to assembler). I like assembler better though as you have more control
and it's closer to the hardware (and more challenging which makes it more fun). I don't know about portability (when
you want to use a different chip) but C should be more portable even though in the lessons I saw you really need to
know register names and things like that.
All in all, even though I'm an advocate for assembler I think sticking to C will make it easier in the beginning. When you
run into speed problems then you can still switch to assembler... my chf 0.02 (which translates to at least 10 US$ )

[nieles]

just found this fet driver
http://focus.ti.com/docs/prod/folders/p ... v8301.html

Features
Operating Supply Voltage 8V–60V
2.3A Sink and 1.7A Source Gate Drive Current Capability
Integrated Dual Shunt Current Amplifiers With Adjustable Gain and Offset
Integrated Buck Converter to Support up to 1.5A External Load
Independent Control of 3 or 6 PWM Inputs
Bootstrap Gate Driver With 100% Duty Cycle Support
Programmable Dead Time to Protect External FETs from Shoot Through
Slew Rate Control for EMI Reduction
Programmable Overcurrent Protection of External MOSFETs
Support Both 3.3V and 5V Digital Interface
SPI Interface
Thermally Enhanced 56-Pin TSSOP Pad Down DCA Package

maybe you can use this for your controller.

[oldswamm]

Interesting IC. Not sure if it's right for this, but it isn't up to me anyway.
He could put a discrete transistor buffer right at the fet.
For current controlled throttle, you would have to send throttle info over the SPI, don't see anything wrong with that.
They may have found an elegant solution to bootstrapping. They use bootstrapping to supply the switching current, then use a miniscule dc-dc that is just capable of maintaining the charge indefinitely (it probably wouldn't be able to supply external buffers indefinitely, though, it's intended for gates, not bipolar). I actually discussed using both dc-dc and bootstrap with someone, 'somewhere' on ES a while back and discounted it as impractical, but on an IC, like this.... Even on this project both might work if you thought the dc-dc might be a little marginal, all it would take is the addition of 1 diode per phase. (just typing as I think.)
<edit> Duh, I see why not, it's a 60v part. You weren't planing on going much over 45V or so, were you.
Would be great for a washing machine. A simple processor to handle commutation, then program the SPI from the MPU.

[bigmoose]

This might make things easier. Here is my personal compilation of FET 1/2 Bridge Drivers from various manufacturer's with key parameters listed.

FETdriving-V1.pdf

(35.38 KiB) Downloaded 81 times