How to combat high gate capacitance?

[tostino]

Well I have a couple 12 fet controllers build up and ready for 133v operation. They are filled with 4115 fets.

I have heard that these have higher gate capacitance than 4110's, or most other fets for that matter, and also heard that the Infinion boards are all the same for the fet driver circuitry. I was wondering what possible mods should be done to increase the power for driving the fets, considering that is a big portion of switching losses, correct?

Personally, I am a noob just trying to learn as much as I can, and I have no idea where to even start. But if we can get some more knowledgeable people in on this, maybe we can find a solution!

 

[mwkeefer]

Toast,

In theory the little SMT resistors which sit between the LM317 rail and the gate power section could be reduced (i can't tell they may be pull up or down resistors, traces are too small right now) since I believe they function to limit current - decrease their resistance a bit and (again in theory) you would flow more current faster into the gate which would handle the additional capacitance inherint in the high voltage FETs.

In the real world, I'm willing to bet you will see not too much difference in performance over 4110s - just higher voltage handling.

Others (Luke, Jeremy, Methods) could give you a better answer with more precise cause / effect and suggestions to combat it.. to the best of my knowledge though, several people have just removed stock fets and added these for 120+v useage with good results - though none of them are running > 100v really (24S max right now) but I would assume gate capacitance would have an identical effect at 100v as 120 and higher - again... others are better suited to the real "engineering" answers (and much more elequant than I).

** EDIT:

I spent some time comparing the datasheets for the 4110 and 4115... in the end, other than the possibility that the gate threshold for the 4115 could be 5v vs 4v maximum, I don't think there is a big enough difference to warrant doing anything to the gate drive of an infineon. They have different rise and fall times, different delay times but in the end... ON will require approx 92ns no matter which you run (4110 or 4115) but the performance (provided you stick to 70A max continuous per FET in parallel) should be quite on par with the 4110's - perhaps a tad more heat, so be sure to use appropriate thermal compound and fasten all the sinks down completely.

-Mike

 

[John in CR]

Methods once told me that 14V or 19V supply was better than a flat 12V on that rail...somehow easier on the fets. Could this be related and this as a partial solution?

 

[mwkeefer]

John,

I'm fairly sure the gates on 4110/4115 can withstand 20v maximum... higher voltage = lower current required and thus I would "assume" faster gate charge or full ON time.

So it makes sense that increasing that voltage (the output of the LM317 regulator) above the 13.5v stock (I think) would enable faster ON switching atleast... won't really effect the OFF time, that's dependent on the gate capacitance and the value (if any exists) of a resistor to ground which bleeds out the gate to make the power off curve much steeper (sharper).

Like I said - I'm not the best at explaining these things, I get it but rephrasing whats in my head more oft than not turns out gibberish.

-Mike

 

[John in CR]

Cool,

I know I can be dangerous when I know only a little bit, but sometimes I might understand just enough to throw a little gold nugget (or only silver) out there as a contribution.

 

[rhitee05]

I have heard that these have higher gate capacitance than 4110's, or most other fets for that matter

That's not true, nor is that the useful spec for comparison in this case. The spec you care about is the gate charge, which is labeled Qg. The 4115 is better than the 4110 in this respect, 77 nC versus 150 nC. So the 4115 will switch more quickly with the same gate drive.

As far as tinkering with the gate drive circuits, there are probably some things you could do to make small gains, but don't expect huge improvements. Anything you change also risks make the drive less stable, or more likely to blow a FET. It's complicated to work with gate drive circuits. To make any substantial improvement, you'd have to use an off-board gate driver.

 

[John in CR]

Thx Eric,

That sounds great for my motors that spend the vast majority of their life at partial duty, and much of that time in current limiting. A cool controller would be something totally new.

 

[rhitee05]

Don't forget that the 4115 still has about 2.5x higher Rdson than the 4110. Lower gate charge probably isn't enough to make up for that unless you're over 100V.

 

[mwkeefer]

In a 12 FET won't that put 2 FETs on each high and low side per phase, in effect halfing the rdsOn to approx that of single 4110 (ie: 6FET with 4110) but with faster switching times?

-Mike

 

[John in CR]

Regarding the rdsOn, is that the resistance while on, or is it something that's a problem when switching more frequently? Every controller failure I've had (all with bigger brushless motors) occurred while at partial duty and partial throttle position. I understand that lower resistance is good, but maybe this other spec plays a big part too. Those of us who have 4115 controllers want to understand how to take best advantage of them, since they're what we have to get past the 100V limit.

 

[mwkeefer]

RdsON is the Resistance once the FET is in the fully on state...

The ramp up and ramp down times are the nano seconds it takes to fully charge the gate and fully discharge it - this is going to effect partial throttle use more than RDSon of 2x.

That's actually the idea behind multiple FETs per phase and side (HIGH + or LOW -) to allow the current being passed into the given BLDC phase to be distributed across multiple FETs and thus reduce the total current across any given FET.

According to Jermey Harris the waveform coming out the infineon even at full throttle and > 100% programming never leaves PWM mode - that would in effect mean we are all running around in modulation mode where those FETs are just turning on and off fast as can be.

Well that's a crude explain - Eric can give you somthing more elegant

-Mike

 

[texaspyro]

If you want to have fun driving FETs, try driving 18 IRFP2907's in parallel switching 20,000 amps... been there, done that. Rise and fall times exceeding 2 billion amps/second...

If you want to parallel FETs you need to match them for gate threshold voltage. Also for Rds on (but if they are from the same batch this tends to be a non-issue). You also want to drive the gates from a common source. No separate gate drive resistors.

See: http://www.irf.com/technical-info/appnotes/para.pdf (warning: heavy techno-reading)

 

[John in CR]

So can I take it that while a 4115 controller may not be great for pushing very high currents, they may be better otherwise since they switch faster?

A question regarding the switching...When something switches slower does the electricity get bottlenecked in the other components creating voltage spikes? I always revert to looking at electricity like a flow of water in a hose, so a fast switcher would be like the noisy sprinkler heads that go around in a circle pulsing the spray off and on rapidly, but the water is more of a continuous flow in the hose. The slow switcher is like turning the water off and on by hand in a sink, where sometimes you can hear the hammer effect inside the pipes. Is there kind of a similar effect with electricity, or does the voltage (pressure) not build up like that behind a slow closing switch?

 

[mwkeefer]

I think the crux of this issue is simple... to best utilize the inherint abilities of the IRFB41XX series FETs we would need to build a seperate driver section and once we get there... may as well build a whole new controller.

Lucky for us - the IRFB tend to plug and play with the Infineon and no matter what, anything beats the stock FETs (regardless of the voltage you want to run).

I've used bldc gate ASICs and I've built gate drives from transistors - designing a gate drive can be uber simple or far too complicated - the drives in the infineons are simple transistor based circuits and could be much bettter... then again, so could the firmware, the PCBs, etc.

Personally I'm still at a loss to the advantages of running > 100v (really about 84)... I've yet to find a single hub motor (of any type, geared or DD) which doesn't burn up and melt down running even low currents (15A) at those voltages - oddly enough, the majority of heat build up seems to come from friction or perhaps a better way to put it - lousy bearings run 2-3x their rated RPM.

-Mike

 

[John in CR]

If you want to have fun driving FETs, try driving 18 IRFP2907's in parallel switching 20,000 amps... been there, done that. Rise and fall times exceeding 2 billion amps/second...

I can't wait till good batts are cheap enough that I can have a need for 20kA on an EV even if it is only at 75V! 2khp peak will be sweet. 8)

 

[John in CR]

I already changed the bearings, so hopefully 1500rpms isn't an issue.

My thought regarding the usefulness of >100V is to take a current limited motor to much higher power, and hopefully greater efficiency will come as a result. That installation would be out of the wheel and geared down, so it's not high speed I'm after. Instead it would be a higher power but silent and durable non-hub drive that only requires a 2:1 or less step down in gearing. Also, the higher rpms will make my ventilated hub motor approach work exceptionally well for cooling.

With my big hubbies, as direct drive in the wheel, 100V+ will make it truly highway capable, especially once I work out some real aero.

Thanks for the heads up about the bearings. I'll replace those on my 9C too while I have it open.

 

[mwkeefer]

John,
You planning a trip to Mars on your ebike-I'm thinking you could reach escape velocity with that much power!
-Mike

 

[John in CR]

John,
You planning a trip to Mars on your ebike-I'm thinking you could reach escape velocity with that much power!
-Mike

LOL! Well I don't have my Jetsons style atomic car that I was essentially promised to have by now when I was a kid back in the 60's. This DIY bike stuff is about as close as I can come on my own, so I might as well play around with the limits. If I can get a useful 8-10kw out of something like a 9C then a 2:1 gear down gives me real mountain climber that can still do 35-40mph without going to multiple speeds and still in a lightweight package.

Also, don't forget that I'm still street legal as a bike here, so I've gotta take full advantage of that lack of regulation while I still can.

John

 

[tostino]

Oh gosh, I have a couple days at work that I'm too busy to check back on the thread, and it has quite a few responses!

I have heard that these have higher gate capacitance than 4110's, or most other fets for that matter

That's not true, nor is that the useful spec for comparison in this case. The spec you care about is the gate charge, which is labeled Qg. The 4115 is better than the 4110 in this respect, 77 nC versus 150 nC. So the 4115 will switch more quickly with the same gate drive.

As far as tinkering with the gate drive circuits, there are probably some things you could do to make small gains, but don't expect huge improvements. Anything you change also risks make the drive less stable, or more likely to blow a FET. It's complicated to work with gate drive circuits. To make any substantial improvement, you'd have to use an off-board gate driver.

This is why I claim I know nothing , because obviously I don't!

So in your opinion, would increasing the driving voltage from the stated 13.5v help at all with at least the turn on speed? If it would, that in your opinion, what is the highest you suggest it should go?

 

[rhitee05]

In a 12 FET won't that put 2 FETs on each high and low side per phase, in effect halfing the rdsOn to approx that of single 4110 (ie: 6FET with 4110) but with faster switching times?

The switching time of 2 parallel 4115s will be on the order of a single 4110, but you'll still have about 1.25x higher Rdson. The only advantage left is you're sharing the loss across two parts instead of one. The total loss would still be higher than a 6-FET 4110 controller.

The ramp up and ramp down times are the nano seconds it takes to fully charge the gate and fully discharge it - this is going to effect partial throttle use more than RDSon of 2x.

Your point is valid, but the switching times are more like 1-2 us (as others have reported) rather than nanoseconds for the Infineon controllers. Theoretical switching times on the order of 10s of ns are possible, but no real controller would ever approach that. A very good design would be on the order of 100s of ns.

 

[rhitee05]

If you want to parallel FETs you need to match them for gate threshold voltage. Also for Rds on (but if they are from the same batch this tends to be a non-issue). You also want to drive the gates from a common source. No separate gate drive resistors.

The reference you quoted is excellent, but I disagree with your recommendation of no separate gate resistors. That same document recommends both a common resistor and parallel resistors on each gate, with the parallel resistors equal to about 10% of the total resistance. With no resistance between gates you risk oscillation.

 

[rhitee05]

So in your opinion, would increasing the driving voltage from the stated 13.5v help at all with at least the turn on speed? If it would, that in your opinion, what is the highest you suggest it should go?

It would decrease the turn-on time somewhat, but it would also increase the turn-off time. Switching loss is caused by the total of the two so I think it would be very close to zero-sum or at most a small benefit. It would also increase power dissipation in the drive circuit (probably not enough to matter). You'd also have to consider other circuits which draw power from that supply. I'm not sure if any other circuits are powered directly from this supply, but at the very least the power loss in the regulator for the 5V supply would be higher (again, may not be enough to matter). Finally, by placing a higher voltage on the FET gate there's a (slightly) higher chance that something will cause a spike and blow the FET.

So, I think you might get slightly better performance (very slightly), but you end up losing in other areas and I doubt it's worth the effort. The usual way of making a "better" gate drive is to decrease the resistance to allow a higher peak current for the same voltage. You could attempt this mod on an Infineon controller, too, and would probably have better results. Caution is still advised, though, since faster switching can lead to shoot-thru and another realm of problems...

Perhaps I should suffice to say that it's complicated. :-D There's a lot of good reading on the 'net about gate drive circuits if you want to educate yourself.
I like this document: http://focus.ti.com/lit/ml/slup169/slup169.pdf

 

[texaspyro]

The reference you quoted is excellent, but I disagree with your recommendation of no separate gate resistors. That same document recommends both a common resistor and parallel resistors on each gate, with the parallel resistors equal to about 10% of the total resistance. With no resistance between gates you risk oscillation.

Yeah, I've done it both ways and have never seen any real difference... YMMV

In my CD welder (which can switch 20,000 amps), I went without the separate resistors. I wanted the best drive match that I could get. I bought 100 fets and macthed them into 5 groups for Vthreshold. There was no need to match Rds on. They all came from the same lot and were within +/- 30 micro ohms or so.

My gate drive resistance is pretty high to be driving so much capacitance (20 ohms into .3 uF with a real crappy DS0026 gate driver). At 10%, that would be 2 ohms for those parallel resistors. I wanted to slow the edge rates down quite a bit to help tame the di/dt (which was over 10 billion amps/sec).