Designing a brushless motor controller

[Michael]

Howdy - I've been thinking alot about brushless motor control lately.
I've been using some COTS brushless motor controllers and have been
very unsatisfied with them - they've all been overly large, overly
complicated, and overly costly. So I'm thinking about trying to build
my own. I'm thinking about having it be something like this:

25-75VDC supply voltage
10A continuous output
hall, quadrature encoder, and analog position inputs
small (in physical size)
CAN connectivity
~25KHz switching speed or faster

The first thing that I'd tackle, of course, would be the hardware
design. I'm looking for very precise control - so I want a good deal
of built in smarts. My gut instinct says to go the FPGA route. Somehow
that just strikes me as a nice clean way of handling things. I'm more
experienced with MCUs - but I'm always happy to learn a new
technology.

Next up - and the topic that I'm having the most trouble with - is the
design of the half bridges. I'd want to use N-FETs throughout to
maximize efficiency. Driving the high side FET gates is troubling. The
only method of driving FETs like this is with a floating power supply
grounded on the source of the FET, and then optotransistors switching
the FET with that. There has to be a simpler solution. First of all -
is there a clever way to combine the floating power supplies such that
you only have one high side FET power supply?

This is the problem that I'm having the most trouble with. I mean - to
keep from blowing through the gate of the FETs (by going over the max
VGS) - you could put a zener across VGS on each high side FET (with a
series resistor of course) - but then you're going to be draining the
floating supply when you turn that FET on until the source voltage
goes high enough to turn off the zener. That is why it's very
attractive to have a separate floating supply on each FET's source -
but yeah... size is important here.

Lastly - driving the FETs - the standard solution I seem to see for
problems like this is to have some sort of a low side high voltage
(well, high enough to handle the bus voltage) N-FET with the source
grounded and a resistor between the gate supply and the FET's drain,
and the gate of the half bridge high side FET connected to the FET's
drain. This has the clear problem of having a huge trade off between
FET turn on speed and power drop across that resistor. I've also seen
optoisolators used for this, which takes care of that problem but also
adds large parts and would seem to me to place a limit on the
switching speed (as I feel like optos are typically substantially
slower than normal transistors).

So I guess - my big questions here are:
1. FPGA/MCU?
2. How to combine gate power supplies?
3. How to drive high side FET gates?

Thanks so much!

-Michael

 

[sycochkn]

Howdy - I've been thinking alot about brushless motor control lately.
I've been using some COTS brushless motor controllers and have been
very unsatisfied with them - they've all been overly large, overly
complicated, and overly costly. So I'm thinking about trying to build
my own. I'm thinking about having it be something like this:

25-75VDC supply voltage
10A continuous output
hall, quadrature encoder, and analog position inputs
small (in physical size)
CAN connectivity
~25KHz switching speed or faster

The first thing that I'd tackle, of course, would be the hardware
design. I'm looking for very precise control - so I want a good deal
of built in smarts. My gut instinct says to go the FPGA route. Somehow
that just strikes me as a nice clean way of handling things. I'm more
experienced with MCUs - but I'm always happy to learn a new
technology.

Next up - and the topic that I'm having the most trouble with - is the
design of the half bridges. I'd want to use N-FETs throughout to
maximize efficiency. Driving the high side FET gates is troubling. The
only method of driving FETs like this is with a floating power supply
grounded on the source of the FET, and then optotransistors switching
the FET with that. There has to be a simpler solution. First of all -
is there a clever way to combine the floating power supplies such that
you only have one high side FET power supply?

This is the problem that I'm having the most trouble with. I mean - to
keep from blowing through the gate of the FETs (by going over the max
VGS) - you could put a zener across VGS on each high side FET (with a
series resistor of course) - but then you're going to be draining the
floating supply when you turn that FET on until the source voltage
goes high enough to turn off the zener. That is why it's very
attractive to have a separate floating supply on each FET's source -
but yeah... size is important here.

Lastly - driving the FETs - the standard solution I seem to see for
problems like this is to have some sort of a low side high voltage
(well, high enough to handle the bus voltage) N-FET with the source
grounded and a resistor between the gate supply and the FET's drain,
and the gate of the half bridge high side FET connected to the FET's
drain. This has the clear problem of having a huge trade off between
FET turn on speed and power drop across that resistor. I've also seen
optoisolators used for this, which takes care of that problem but also
adds large parts and would seem to me to place a limit on the
switching speed (as I feel like optos are typically substantially
slower than normal transistors).

So I guess - my big questions here are:
1. FPGA/MCU?
2. How to combine gate power supplies?
3. How to drive high side FET gates?

Thanks so much!

-Michael

A small inverter supply for the floating supply. Torroids are not that big.

Bob

 

[GPG]

A small inverter supply for the floating supply. Torroids are not that big.

Bob

http://ww1.microchip.com/downloads/en/AppNotes/BLDC MC 00957a.pdf

 

[[email protected]]

Howdy - I've been thinking alot about brushless motor control lately.
I've been using some COTS brushless motor controllers and have been
very unsatisfied with them - they've all been overly large, overly
complicated, and overly costly. So I'm thinking about trying to build
my own.

Modern brushless drives are very complex, you should reconsider.

I'm thinking about having it be something like this:

25-75VDC supply voltage
10A continuous output
hall, quadrature encoder, and analog position inputs
small (in physical size)
CAN connectivity
~25KHz switching speed or faster

The first thing that I'd tackle, of course, would be the hardware
design. I'm looking for very precise control - so I want a good deal
of built in smarts. My gut instinct says to go the FPGA route. Somehow
that just strikes me as a nice clean way of handling things. I'm more
experienced with MCUs - but I'm always happy to learn a new
technology.

Most new ones have FPGA, older ones had microcontrolles. If you've
never done FPGA design before you have a very big mountian to climb.
Another good reason to forget the whole idea.

Next up - and the topic that I'm having the most trouble with - is the
design of the half bridges. I'd want to use N-FETs throughout to
maximize efficiency. Driving the high side FET gates is troubling. The
only method of driving FETs like this is with a floating power supply
grounded on the source of the FET, and then optotransistors switching
the FET with that. There has to be a simpler solution. First of all -
is there a clever way to combine the floating power supplies such that
you only have one high side FET power supply?

This is the easy part, not only have there been one chip soloutions
around for years, they are also well documented look on IR's website.
If you can t work this bit out on your own you have yet another good
reason to forget the whole idea.

This is the problem that I'm having the most trouble with. I mean - to
keep from blowing through the gate of the FETs (by going over the max
VGS) - you could put a zener across VGS on each high side FET (with a
series resistor of course) - but then you're going to be draining the
floating supply when you turn that FET on until the source voltage
goes high enough to turn off the zener. That is why it's very
attractive to have a separate floating supply on each FET's source -
but yeah... size is important here.

Lastly - driving the FETs - the standard solution I seem to see for
problems like this is to have some sort of a low side high voltage
(well, high enough to handle the bus voltage) N-FET with the source
grounded and a resistor between the gate supply and the FET's drain,
and the gate of the half bridge high side FET connected to the FET's
drain. This has the clear problem of having a huge trade off between
FET turn on speed and power drop across that resistor. I've also seen
optoisolators used for this, which takes care of that problem but also
adds large parts and would seem to me to place a limit on the
switching speed (as I feel like optos are typically substantially
slower than normal transistors).

So I guess - my big questions here are:
1. FPGA/MCU?
2. How to combine gate power supplies?
3. How to drive high side FET gates?

Well if your still determined get one of internationals "power train"
kits (I'm not sure if they still make them) for the power design side.
Fujitsu have a new brushless design kit available from Farnell for all
the control side, mcu, software theory ect. Of course you still have
time to FORGET the whole idea.

 

[Michael]

Modern brushless drives are very complex, you should reconsider.

I appreciate your concern, however, I am not worried.

 I'm thinking about having it be something like this:





Most new ones have FPGA, older ones had microcontrolles. If you've
never done FPGA design before you have a very big mountian to climb.
Another good reason to forget the whole idea.

I've done a decent bit of FPGA work - it's just been a while so I'm
rusty. That's good to hear that I'm on the right track with the FPGA
route. It just seems the more logical route for high performance
stuff.

This is the easy part, not only have there been one chip soloutions
around for years, they are also well documented look on IR's website.
If you can t work this bit out on your own you have yet another good
reason to forget the whole idea.

The less chips there are the happier I am. I prefer doing things
discretely instead of connecting together black boxes. However, the
IR2101/2 looks like it might take care of a lot of problems for me.
From the datasheet, it uses the way I was talking about with a high
voltage FET + resistor driving the half bridge FET gate.

Well if your still determined get one of internationals "power train"
kits (I'm not sure if they still make them) for the power design side.
Fujitsu have a new brushless design kit available from Farnell for all
the control side, mcu, software theory ect. Of course you still have
time to FORGET the whole idea.

Software I'm not worrying about at this point - only the hardware.
Somebody else is signed on for that side of things. I don't plan on
forgetting the idea I don't seem to see anything about "power
train" kits on IRF's website, unfortunatley. Thanks.

-Michael

 

[[email protected]]

I appreciate your concern, however, I am not worried.







I've done a decent bit of FPGA work - it's just been a while so I'm
rusty. That's good to hear that I'm on the right track with the FPGA
route. It just seems the more logical route for high performance
stuff.



The less chips there are the happier I am. I prefer doing things
discretely instead of connecting together black boxes. However, the
IR2101/2 looks like it might take care of a lot of problems for me.
From the datasheet, it uses the way I was talking about with a high
voltage FET + resistor driving the half bridge FET gate.







Software I'm not worrying about at this point - only the hardware.
Somebody else is signed on for that side of things. I don't plan on
forgetting the idea I don't seem to see anything about "power
train" kits on IRF's website, unfortunatley. Thanks.

That was some time ago but search for this instead on their website,
it's just what you need. IRMD22381Q

 

[Michael]

Sounds like you may be about to find out why.









You'll need a separate supply for each since each phase is moving
relative to the others.  Check the IR 2100 and friends for one approach
to high side driving.  Note that there is a limit on how long their
scheme can hold the high side on.

Well - if you implemented my suggestion in using a resistor in series
with a zener then you could use a combined supply. I did exactly this
fairly recently for some low speed high side switching. Worked fine -
but you do waste some power in the zener and reisistor, while also
slowing switching speeds. The IR2101 seems to be exactly what I was
describing in my first post - except all wrapped up in a single chip.

For low threshold voltage FETs it can be usefule to have a negative gate
drive to help turn off the FET and keep it off.

I've never seen this done before - are there any half bridge driver
chips that implement this? Also - why would this be necesarry? I mean
once your VGS is below VTH there should be no conduction channel...

One item missing from your spec list is the number of phases.

Three!

-Michael

 

[Robert Adsett]

Howdy - I've been thinking alot about brushless motor control lately.
I've been using some COTS brushless motor controllers and have been
very unsatisfied with them - they've all been overly large, overly
complicated, and overly costly.

Sounds like you may be about to find out why.

So I'm thinking about trying to build
my own. I'm thinking about having it be something like this:

25-75VDC supply voltage
10A continuous output
hall, quadrature encoder, and analog position inputs
small (in physical size)
CAN connectivity
~25KHz switching speed or faster

The first thing that I'd tackle, of course, would be the hardware
design. I'm looking for very precise control - so I want a good deal
of built in smarts. My gut instinct says to go the FPGA route. Somehow
that just strikes me as a nice clean way of handling things. I'm more
experienced with MCUs - but I'm always happy to learn a new
technology.

Next up - and the topic that I'm having the most trouble with - is the
design of the half bridges. I'd want to use N-FETs throughout to
maximize efficiency. Driving the high side FET gates is troubling. The
only method of driving FETs like this is with a floating power supply
grounded on the source of the FET, and then optotransistors switching
the FET with that. There has to be a simpler solution. First of all -
is there a clever way to combine the floating power supplies such that
you only have one high side FET power supply?

You'll need a separate supply for each since each phase is moving
relative to the others. Check the IR 2100 and friends for one approach
to high side driving. Note that there is a limit on how long their
scheme can hold the high side on.

For low threshold voltage FETs it can be usefule to have a negative gate
drive to help turn off the FET and keep it off.

One item missing from your spec list is the number of phases.

Robert

 

[Tim Wescott]

Howdy - I've been thinking alot about brushless motor control lately.
I've been using some COTS brushless motor controllers and have been
very unsatisfied with them - they've all been overly large, overly
complicated, and overly costly. So I'm thinking about trying to build
my own. I'm thinking about having it be something like this:

25-75VDC supply voltage
10A continuous output
hall, quadrature encoder, and analog position inputs
small (in physical size)
CAN connectivity
~25KHz switching speed or faster

The first thing that I'd tackle, of course, would be the hardware
design. I'm looking for very precise control - so I want a good deal
of built in smarts. My gut instinct says to go the FPGA route. Somehow
that just strikes me as a nice clean way of handling things. I'm more
experienced with MCUs - but I'm always happy to learn a new
technology.

Next up - and the topic that I'm having the most trouble with - is the
design of the half bridges. I'd want to use N-FETs throughout to
maximize efficiency. Driving the high side FET gates is troubling. The
only method of driving FETs like this is with a floating power supply
grounded on the source of the FET, and then optotransistors switching
the FET with that. There has to be a simpler solution. First of all -
is there a clever way to combine the floating power supplies such that
you only have one high side FET power supply?

This is the problem that I'm having the most trouble with. I mean - to
keep from blowing through the gate of the FETs (by going over the max
VGS) - you could put a zener across VGS on each high side FET (with a
series resistor of course) - but then you're going to be draining the
floating supply when you turn that FET on until the source voltage
goes high enough to turn off the zener. That is why it's very
attractive to have a separate floating supply on each FET's source -
but yeah... size is important here.

Lastly - driving the FETs - the standard solution I seem to see for
problems like this is to have some sort of a low side high voltage
(well, high enough to handle the bus voltage) N-FET with the source
grounded and a resistor between the gate supply and the FET's drain,
and the gate of the half bridge high side FET connected to the FET's
drain. This has the clear problem of having a huge trade off between
FET turn on speed and power drop across that resistor. I've also seen
optoisolators used for this, which takes care of that problem but also
adds large parts and would seem to me to place a limit on the
switching speed (as I feel like optos are typically substantially
slower than normal transistors).

So I guess - my big questions here are:
1. FPGA/MCU?
2. How to combine gate power supplies?
3. How to drive high side FET gates?

Thanks so much!

-Michael

To answer your questions, in order:

0. No you didn't ask it, but you should have: consider your engineering
time when you make your decisions. Unless your production volumes are
huge, or the controllers that you buy don't meet your specifications for
some reason, or you have to stand the rest of the system on it's head to
accommodate the drives you're using, you're probably going to end up
spending more money in the end by rolling your own, even if you drove
the BOM cost to zero.

1. I'd consider a DSP. TI, Analog Devices, and Microchip all make DSP
chips that are designed for brushless service. I _think_ that all of
them are available with CAN controllers built in, to boot.

2 & 3. Have you looked at high-side driver chips? These address many
of the issues that you have, and IIRC there are ones that are good for
up to 300V supply rails. I've seen these used with great success for
class D synchronous amplifiers (by a company with low production
volumes, but very stringent requirements, so they always ended up
rolling their own and paying the $$ for engineering).

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" gives you just what it says.
See details at http://www.wescottdesign.com/actfes/actfes.html

 

[Robert Adsett]

I've never seen this done before - are there any half bridge driver
chips that implement this? Also - why would this be necesarry? I mean
once your VGS is below VTH there should be no conduction channel...

There is a parasitic capacitance from the drain to the gate. If the
Drain-Source voltage changes quickly (say due to a low side turning on)
you get voltage coupling from the drain to the gate, in the worst case
turning or keeping the FET on. Even if it does not turn fully on you
can get siginificant shoot through currents. Driving the gate slightly
negative increases the margin. It's not a problem with high threshold
devices but a lot of the modern high performance FETs have low
thresholds and are more sensitive to this. For some the only difference
between a logic level FET and the normal FET appears to be the label.

I think IR has an appnote on developing a negative gate drive using a
2100. It does use a number of extra components.

IGBTs often need a negative turn-off drive as well I understand.

Three!

I've seen three and four. I recall there being a 3 phase (3 1/2 H's in
a package) chip. Maybe from IXYS? Capable of more current than you
need as I recall. It might be worth a quick look, Ah here's one of
them http://ixyspower.com/store/PartDetails.aspx?r=0&pid=3763

I don't know the cost of these but I'm curious.

I expect the needed capacitors are going to take more room than your
power devices anyway.

Robert