Odd# of mosfets vs even#

[swbluto]

Thanks rhitee05, I think I can now accurately correct the heat formulas.

 

[John in CR]

Great discussion on the odd verses even fets. I will throw in some more thoughts which have already been touched on in other peoples postings. I have been reviewing some ebike controller designs that use an even configuration and they tend to be under designed for the starting torque conditions. When you are battery current regulated, and you don't have the matched motor that the controller was set up for, you run the risk of cooking the lowside fets during hard accelerations at low speeds. But as your motor speeds up, the limiting factor is the battery, which is a much lower current then the stall or low rpm motor current. So from a reliability point of view, buying a 9 fet design over a six fet design or a 15 fet over a 12 fet design makes total sense. The controller has been optimized in cost since you are not populating extra fets that are not required. By the way, I would say based on consultants and my own observation that the vast majority of controllers are design for geared hub motors, so regen is not a big issue(except for this crowd maybe), but for those that are using it in direct drive motors, limiting motor regen current to no more then battery current would deal with that problem. I have had two BMC 500 watt controllers driving a cutler-mac 500watt geared hub motor fail very quickly when they have been driving the bike up long hills. Had those controllers been 9fet designs, they would not likely have failed except maybe because of quality issues :lol:

Thanks WR, finally something I can fully sink my teeth into and not get lost in the EE language barrier. The idea that a bit of cost savings at no real sacrifice makes perfect sense to me. Is it possible that we can get a general consensus on this?

Regarding the regen comment, except going down a hill steeper than can be climbed with the motor/controller combo at reasonable speed, is there any possibility that regen current can exceed max battery side current limits in the controller?

John

 

[swbluto]

Can someone other than me give me an example of controller switching heat?

I was using the values that I have, but I'm getting something ridiculous like 300 watts with 40 phase amps, 1 uS turn_on and turn_off time, 80 battery volts and 10,000 hz PWM frequency and I'm wondering if my calculations are wrong. The freewheeling and resistive heat losses are only like 15 watts at that point.

Edit: It looks like an arbitrary '5' found itself in the formula, deleting it gave a more reasonable 60 watts.

 

[TylerDurden]

Can someone other than me give me an example of controller switching heat?

Might be some nuggets buried here:
http://endless-sphere.com/forums/viewtopic.php?f=2&t=2134&start=45 (bit of a wrestling match, but the fet info is worth the read).

 

[swbluto]

Can someone other than me give me an example of controller switching heat?

Might be some nuggets buried here:
http://endless-sphere.com/forums/viewtopic.php?f=2&t=2134&start=45 (bit of a wrestling match, but the fet info is worth the read).

Cool, looking at the numbers he was using, it seems like the formula is accurate. He assumes 50 khz PWM (What controller is that?), and a rise and fall time of ~60-80 ns. However, "the experts" online advise that the switching time is more like 1000 nS with the infineon and better designs by wrobinson can get down to 600 ns. The longer rise and fall times beyond predicted RC times are due to inherent trace inductance.

Does it seem right that switching losses constitute 75% of the total losses and is somewhere in the range of 100 watts with 80 volts and 40 phase amps? That's what my calcs are showing, roughly. And, this seems to suggest the high side fet is enduring the vast majority of heat during switching (At least in this case). At least, this is the case when using a high voltage (Like 80 volts) and having around 40+ phase amps.

Given methods experience with a hot 18-fet controller, I guess that might seem reasonable and it doesn't seem like 12 fets or 18 fets would alleviate that problem since the switching losses would be the same (Per mosfet it wouldn't, but I don't think it's per mosfet that's the problem... is it?)

 

[rhitee05]

Given methods experience with a hot 18-fet controller, I guess that might seem reasonable and it doesn't seem like 12 fets or 18 fets would alleviate that problem since the switching losses would be the same (Per mosfet it wouldn't, but I don't think it's per mosfet that's the problem... is it?)

This is splitting hairs a little bit, but in theory switching losses in an 18-FET could be higher than in a 12-FET. For a given drive circuit, the switching time will naturally be longer when driving 3 FETs than when driving 2. It's possible they do something to beef up the driver, or maybe not. I don't have any data to prove it either way. But if the drive circuits are identical, an 18-FET would switch slower than a 12-FET and thus incur higher switching losses. Resistive losses would be lower, though, so at higher currents that would make up the difference.

Assuming that somehow switching times remain constant, then switching losses would also be the same.