View Single Post
Old 06-24-2010, 03:12 PM   #3477 (permalink)
Join Date: Feb 2009
Location: phx
Posts: 260
Thanks: 0
Thanked 41 Times in 34 Posts
You bring up some good points - I've actually thought about a couple of these and thought I'd share my conclusions.

Originally Posted by Darxus View Post

1) Asynchronous mosfet switching. Instead of switching all mosfets on and off at once for one massive pulse, hook the batteries up to the controller in 6 (number of pwm outputs on the ATMega168) isolated packs, and stagger their mosfet switching. Shouldn't that significantly reduce load on things like capacitors and diodes, reducing heat / loss?
I believe this is called interleaving. You're right in that interleaving will allow you to run at a higher switching frequency and reduce the stress on the capacitors. Or, you can run at the same frequency, but have your fets alternate switching if you have 2 legs (or switch every 6th pulse if you have 6 legs) which will mean your mosfets are on for half (or 1/6th) the time. This will reduce the stress on the mosfets.

There are two drawbacks. First, whichever leg is on has to carry the full amount of current. So, if this controller were to have 2 banks of mosfets that alternate switching, they'd both have to be sized for 500A. That means twice the number of mosfets (roughly speaking - you might be able to get away with one or two less, but at some point the instantaneous current will blow them).

The second is that the heat losses will still be roughly the same. The same amount of current will be switched at the same frequency. The same amount of heat is produced, just spread around between two (or more) legs. Or, if you up the frequency, there will be even more losses than before.

2) Switch battery input between parallel and serial connections. Efficiency graphs from the Enertrac motor thread on endless sphere show that for every RPM, there is a most efficient voltage (related to Paul's great tests on the relationship between volts, amps, speed, and torque). So it would be nice if you could adjust the input voltage based on the motor RPMs, but doing it manually should still be useful. Battery packs could be hooked up to switch between, for example, 36v 40A, 72v 20A, and 144v 10a, by switching them from full parallel, to parallel / serial, to full serial. Implemented as a pair of banks of manual contactors? (Combining #1 and #2 should make for an interesting diagram.)
Yes, switching high voltage increases the switching losses, so a lower input voltage would increase efficiency. However, as of now, the most inefficient operation of the controller is the moment the car accelerates from a stand still. Calculations show that it is operating near 90% efficiency at this point and rises as the vehicle accelerates. I guess a lower input voltage would buy you a few percent for those several seconds of acceleration. You'd have to decide if it is worth the cost and complexity of added contactors. When cruising at 50% duty cycle and 200 motor amps, it's about 97.5% efficient.

3) Why wouldn't running some capacitors across the motor terminals be useful for smoothing things out?
They may, but I don't think there would be any significant benefit if the motor saw less ripple. There aren't any controllers out there with caps on the output, afaik.

(Freaking long thread.)
I'm glad I got in when it was only 40 pages long :P
ReVolt AZ testing thread:
  Reply With Quote