View Single Post
Old 02-03-2011, 05:12 PM   #4274 (permalink)
DJBecker
EcoModding Apprentice
 
Join Date: Nov 2010
Location: Annapolis
Posts: 159
Thanks: 0
Thanked 29 Times in 24 Posts
Quote:
Originally Posted by MPaulHolmes View Post
DJBecker! You are doing some SR experiments?! ya! I still really want to do that too.
Our initial plan was to do SR entirely under software control. To that end I wrote our firmware with OCR1B configured as the SR gate output.

For those that haven't looked at the details, the SR gate drive is not as simple as inverting the primary gate drive. There needs to be a "deadband' period between the gate drives to prevent both from being turned on at the same time. And you have to handle various edge cases correctly e.g. turn on/off and 1%/99% throttle. One illustrative case is when you drop to zero throttle. Initially there is positive freewheel current, which then reverses. If you leave the SR MOSFETs blindly turned on, you change to hard regenerative braking at the transition. So you have to monitor the motor current, waiting until the freewheel current drops to a low level before turning off the SR gate drive.

Our 'version 0' bench tests used a diode instead of a SR MOSFET to minimize risk. It also had no gate isolation. We might have had a semi-working configuration, but we couldn't really tell. Our Syscomp computer o-scope crashed just being in the same room and we had frequent USB disconnects to the controller itself. Between the 18V driver supply and no gate resistor, we must have been generating some serious broad-spectrum noise. (Along with some pretty hefty voltage spikes back into the controller board. It still runs, but now draws about 3x the current of a new one. Doh!)

The next step was to rework our gate driver board with an optoisolator and gate drive resistors. The first round used a standard 2501 optoisolator. This immediately raised a design issue for SR gate drive: optoisolators are slow and asymmetrical. This one took well over a microsecond to turn on. Worse, the turn-on and turn-off times were different, and varied with temperature, voltage and pulse width.

That's not a problem with single gate drive and feedback. If the gate pulses are narrowed or widened by the isolation circuit, the feedback loop adjusts. But it's a major problem when you need to avoid overlap or a long gap with both sides turned off.

Hoping to improve the speed, we ordered the same 'high speed' optoisolator that Cougar uses. The change created its own problems as we belatedly discovered gate drive signal ringing that we didn't have with the 2501. The fix was to increase both the drive and sense current e.g. change from a 470ohm to 220ohm drive resistor. (No, it wasn't that easy to find and fix. "Belatedly" means "after we added to the exploded parts bin". Only after did I read that the Cougar 2d board made a similar change, from 330 to 220 ohms.)

Even with the supposedly faster optoisolator, I still wasn't comfortable with the signal delay and delay symmetry. So we explored other options. The two current contenders are using a Si8233 isolating gate driver, or using a IR11672AS SmartRectifier driver.

Both are surface mount parts, and thus pose a challenge for putting in a kit for other people. Other than that, they are both great parts.

The Si8233 has both dual isolation and dual gate drivers in one package. It's far faster than an optoisolator, requires fewer external parts, is easier to design with, and shouldn't degrade in performance over time. It also enforces a minimum dead time between the two gate drivers -- a useful guard against software bugs. It's down-side is that it's rated at only 4 amps per driver, which might be a little weak for the IXFX230N20T power stage we are planning to build later this year.

The IR11672 takes a entirely different approach, changing from software to hardware SR. It's a stand-alone chip that generates the SR gate drive by monitoring the voltage on the MOSFET pins. It doesn't need software support, a controller output pin or an isolation stage. Nor does it need to be tuned for different isolator delays, turn-on speeds or motor current L/R time constants. Best of all, it might be usable as a drop-in efficiency improvement for existing Cougar controllers.

There are a bunch of other similar SR controller chips, but the IR chip looks the best to me. Unless most others, it directly handles the full voltage. Alternative chips require a bunch of external components to clamp the sensed voltage without degrading the sensitivity or speed too much.

Last edited by DJBecker; 02-03-2011 at 06:14 PM..
  Reply With Quote