DIY 6FET Controller help

SjwNz

100 W
Joined
Jul 19, 2011
Messages
146
Location
Hamilton, New Zealand
Hello every one, this is my first post there but have been following endless-sphere off and on for the last 3years. A little about myself, I work for a small electronics development business where I mostly do small Microcontroller jobs and only went to Polytech/Uni for about 2years which I never finished and learn most of my skills (or lack of) during the 15years I have been working. My theory is not that good which is my downfall but I have managed to get by, until I tried to make a Speed controller :D

I was hoping for some advice with my 6fet Sensored BLDC speed controller I have been working on.
This is my 2nd version as my 1st model only got as far as running a motor but decided the whole design was bad so I started again.

After finding a number of mistakes on my V2 board, I finally got it into a bike and sort of running . I am using a 2kw Outrunner motor and the battery pack was made from 120 A123 cells which I wired into a 40V 23A/h pack.

After riding the bike to my friends house which was only 5mins away, when I got home
I decided to increase the current from 40Amps(1500Watts) to 50Amps (2kw)
About 5sec into the ride the fuse went, so the next day I decided my V2 PCB design was also bad but wanted to see if I could make it work by beefing up the ground tracks to the Current shunt. That night the controller worked for about 10mins and then the fuse went again.

After fixing it 3 or 4 more times, the same fets died and I think it is still a ground loop issue with the Driver chip ( NCP5181 ) Or it could be something else. The fets that keep failing have the longest track length .

I am going to redesign this controller again as V2 is rubbish and I will use have a GND plane and will not have a Current shunt in the Neg rail and put it in the Pos rail before all the Caps and try again. But in the mean time I will fix up V2 again and back off the current to 35 to 40Amps and see how long it will last.

I also have questions about the best way for current limiting as the method I am using is to take the voltage across the current shunt, amplify it and feed it into a comparator so if I set the trigger current to 40amps and when the output of the comparator goes high my Microcontroller code will quickly back off the PWM duty until the current drops below the set point. I will then start to increase the PWM duty again at a slower rate until the PWM value matches the Throttle input. I hope this makes sense.

1.jpg

2.jpg

3.jpg

4.jpg

5.jpg

6b.jpg

7b.jpg
 
have you seen this thread about building an output stage:
http://endless-sphere.com/forums/viewtopic.php?f=30&t=55641&p=828490#p828490
the secodn iteration is shown in this thread:
http://endless-sphere.com/forums/viewtopic.php?f=30&t=57877&p=863162#p863162

about your current limiter, you're limiting the battery current, but the currents the FETs see is actually the phase current, which can be much higher than the battery current...
roughly speaking the phase current is the battery current divided by the PWM dutycycle.
 
Is it a highside or lowside MOSFET that's failing?

Your schematic is wrong, the bootstrap capacitor must be connected between pin 8 and pin 6 of the NCP5181.

Regarding the bootstrap capacitor, it's capacitance seems very high at 22µF, are you sure it is fully charged during the ON-time of the lowside MOSFET?
 

Attachments

  • 5.jpg
    5.jpg
    31 KB · Views: 8,106
I always use 47uF // 100nF combo for the boodstrap, not had any problems so far (getting on to 10000km)

I typically have a 15 Ohm or so resistor in series with the bootstrap diode. Also, I would add a diode / resistor combo
to the FET gate lines such that you have around 10 Ohm switch-off resistance and 43 Ohm switch-on resistance. Make
sure the gate drive lines of the gate and source of a FET are very very close next to each other, to minimize inductance.
 
I have used 2.2µF for the bootstrap capacitor with minimal voltage drop during gate charging, so 47µF seems like a lot. But I guess, if the capacitor is just fully charged before switching the highside MOSFET, the capacitance can't get too big. But it can get too small, it needs to be able to fully charge the gate without too much voltage drop so it depends on your MOSFET gate charge and leakage current in the MOSFET and gate driver and of course the switching speed (when the highside MOSFET is on, the bootstrap capacitor is slowly discharged by leakage current).
I guess the application notes I have read on the topic, was very cost-conscious and that was the reason they didn't go "overkill" with capacitance :)

Anyway, take a look at Lebowski's low inductance output stage from the links he provided, and note that the smaller capacitor on the Vbus should be a film type capacitor with high dV/dt rating.
 
I have spent the afternoon looking for code ideas for reading the 3 hall effect current sensors for each phase with not much luck,
I am only using a PIC18F252 and not sure if the ADC is fast enough for reading the 3 sensors.
So I am going to use this crude method for now using a bunch on Comparators, one is used for Current limiting
and the other for a over-current shutdown.
I guess if I write some ADC code later on, I just need to read the 3 sensors quickly looking for a set point to trigger a current limit or
if it gets to high a Over current shutdown.
I haven't even made V3 yet and I am all ready planing V4 controller which will use a 16bit PIC as I have a HiTech PIC-C dsPIC & PIC24 compiler at work
which we have not used for a long time so I better make use of it.

thanks for the info guys , I have been spending hours reading trying to soak up all the info.
 

Attachments

  • Schematic Prints.pdf
    7.7 KB · Views: 279
why are you monitoring all three phases? the only benefit of monitoring all three is if you blow a fet on one of the phases but then it already too late anyway.

have a look at this project:
http://endless-sphere.com/forums/viewtopic.php?f=30&t=36479

he has the current sensor between the source of the lower mosfets and the capacitors, so he is monitoring "phase" current
 
I don't want to put anything between the Source and the Neg supply bus, Just about all the controllers I have seen so far have current sensors in the phase outputs to the motor. But I was thinking about it as this is only going to be a sensored controller, I could get away with only monitoring the current of 2 phases as 1 of the sensors will show a neg current of the phase I do not have a sensor on.
I will just work on the ADC code to read the sensors and forget about the comparators and then its easy to read a Pos or Neg current flow.
 
Controller design on hold, busy making 700mm high 7segment displays for my friends Dirt Drag racing.
Have been slowly working my way through reading Arlos power Stage so far page 25 of 66 , and Lebowski controllers, so much info
to take in. Lebowski is that method of using a Cap & resistor between Drain and Gate to help reduce overshot still current, as I saw that in a post about 2years ago. As I am only running 40 to 50V and using 100v fets I have lots of headroom.
Still not sure how this method works I will have to fine the post again and read it a number of times.
I have made a start on the power section and have cut the length of it down by almost 1/2.
I still want to mount the fets to a PCB but like your very simple low inductance output stage Lebowski and may try this one day.

I am picking I also need to calculate the Avg battery current using the phase currents or maybe just put a current sensor in the +Battery line to look at
the Avg battery current and back off power to motor if phase current or Avg battery current gets to high.

Futterama I just found the 3 Allegro sensors I got 3years ago, they are only 50amp sensors so I may have to order higher current sensors.
 
I spent about 2hours playing around with a layout for the power stage for my 6fet controller using TO-247 packages.
The fets will be under the PCB laying flat and going to use twisted pair connections from the fets to the gate drivers
to get the gate signals over the DC bus.
I don't think I have enough DC link caps but have not calculated anything yet and
Not to sure if I should have DC caps between phases or have them at the end of the bus bars as in my layout.
The main power tracks are on the bottom layer which will be built up with wire and solder.
I am trying to use what I have learnt on this forum but would like to know if anyone has any
comments or suggestions.


View attachment Driver.pdf
 
SjwNz said:
I have spent the afternoon looking for code ideas for reading the 3 hall effect current sensors for each phase with not much luck, (...)

What voltage comparator are you using in that schematics?
 
I was going to use a number of LM358s, I know this is not the best OpAmp for this but we had lots of them at work and they have been ok.
I decided today I am going to change my design big time, use a dsPic and the ACPL-333J gate drive with isolated DC-DC supplies
so I can have Miller clamp and Desat protection.
This maybe over kill for a 2kw controller, but I really want to play with all these new features.
 
Why did you put the red and blue over each other in the supplies for the output stage ? This will not help for making it low inductance, as there is no current
flowing through the red ...
 
SjwNz said:
I decided today I am going to change my design big time, use a dsPic and the ACPL-333J gate drive with isolated DC-DC supplies
so I can have Miller clamp and Desat protection.
This maybe over kill for a 2kw controller, but I really want to play with all these new features.
I'm building a controller with the ACPL-333J gate driver too. There are a few things to be aware of with that driver, I haven't figured out a solution for all of them yet, but maybe we could join forces and help each other out?
 
Lebowski, I think on page 65 of Arlo's power stage thread there is something about this method on the laminated bus bars, but I could be wrong.

Futterama
It maybe a month or so before I start testing the 333J, but I have ordered 10 of them and should be here next week.
And yes, I would be keen to help each other out if needed.
 
I kinda lost track of what Arlo1 is doind in his thread, but the general idea for a busbar... well:
- in a sinewave controller during part of the cycle all motor currents go through the + or - supply rail.
- To minimize the inductance form this, in both cases the currents must go (as close as possible) through the same 3 dimentional space.

Just a groundplane doesn't do that.
 
SjwNz, I haven't even ordered mine yet, but I have some similar devices from Fairchild Semiconductor that are pin compatible and has the exact same features, these are for testing purposes. It might be more than a month for me before I get the real parts so no worries.
I'm in the process of designing the isolated supplies for the ACPL-333J, maybe you can use some of that work when it's done.

The things to look out for with the ACPL-333J is the max UVLO voltage of 12.5V (which means the recommended supply voltage is 15V), and the DESAT threshold min/max values of 6-7.5V, which for MOSFET use, means you will have a very big DESAT trip variation which almost makes DESAT useless, especially with low RDSon MOSFETs. I'm working on a workaround for this, using a high speed comparator for more accurate detection of the voltage drop across the MOSFETs.
 
Put a zener diode in series with the Desat diode to lower the Desat trigger current. Its not in the Datasheet for the 333J but
I is in datasheets for the ACPL family of gate drivers. This method is used on the TD350 driver chip that zombiess used in his Controller.

If putting a zener in series does not make sense, I saw a post by HighHopes and then it clicked and now I understand it.

I am not to worried about the UVLO, I was using 13V on the gate of my last controller (Ok it would only last 10mins)
so I will just go to 15V. I may put a zener on the G/S but I have seen posts that say it may cause ringing so I will deal
with that somehow if it happens.

Are you going to make your own isolated supplies? I got a quote from Element14 for a 1W 15V DC/DC supply for around $5 each so I was
just going to use them.
 
The zener in series with the DESAT diode is exactly to lower the trigger current. But if the DESAT threshold can vary between 6V and 7.5V, you will have a large variation in the trigger current threshold.

Example: Let's say you use the IRFB4468 which has a typical RDSon of 2mOhm. And perhaps you would like to trigger the DESAT at 100A. The voltage drop across the MOSFET at 100A is 0.2V. So you want to trigger the DESAT, which has a threshold anywhere between 6V and 7.5V, with a voltage drop across the MOSFET of 0.2V?
Let's say you choose to set the DESAT voltage to 5.8V at 0A so the DESAT voltage will be 6V at 100A. If your ACPL-333J does trigger at 6V, no worries. But if your specific part triggers at 6.5V, due to temperature or part-to-part variation, the trigger current is not 100A but 350A. What if your specific part triggers at 7.5V? The trigger current would then be 850A.
If you then choose to set the DESAT voltage to 6.5V which is the typical threshold for the part, and hence have a DESAT voltage of 6.3V at 0A, but this specific part triggers at 6.2V?

zombiess DESAT works because he uses MOSFETs with higher RDSon so the voltage drop across the MOSFET is larger.
 
SjwNz said:
Are you going to make your own isolated supplies? I got a quote from Element14 for a 1W 15V DC/DC supply for around $5 each so I was
just going to use them.
I forgot to answer this. Yes, I have an idea for a smaller and perhaps cheaper way to get isolated power for the 6 gate drivers. I'm getting help to build this, and when it is finished, I will post it all here on ES.
 
My ACPL-333 chips turned up today and then I found this SOIC socket which I can use to test the difference between the Desat threshold on the 10 chips I have.


I'm working on a workaround for this, using a high speed comparator for more accurate detection of the voltage drop across the MOSFETs.
Futterama
I came up with this today, is this similar to what you were planing ?
I have ordered some parts to test this little circuit.
View attachment 1
 
Back
Top