Paralleling Gate drivers?
Kolas, Sun Sept 30 2007, 08:01AMHi guys.
It's been quite some time since I posted on the site... been busy.
I'm trying to drive a full bridge of some fairly large To-247 IGBT's and I roasted a few UCCx732x's before I realized what is happening.
The gate charge for one of my IGBT's is 405nC as listed in the data sheet.
Using the equation P=Qg*V*F I've found that I need about 5W to drive this full bridge at 200khz. Since most of this power has to be dissipated inside the gate driver, it's obvious a pair of UCC's cant provide this amount of power. (at least not the P package)
I'm trying to work with free parts here. I know there are a few TO-220 package drivers that are quite capable of delivering this much power...
So. finally to the question.
What would be the best way to parallel UCC's to get a rating that is suitable?
I've seen stacking, but this seems like it would further lower the heat dissipation the chip can handle.
So can I put the chips side by side and clamp a heat sink atop them to help apease the overheating that is going to be unavoidable?
Any input would be greatly appreciated.
Thanks for your time.
Kolas
Re: Paralleling Gate drivers?
Marko, Sun Sept 30 2007, 09:28AM
For CW drive, I would sample some of those TO-220 packaged UC3710's. They do a great job dissipating heat large amounts of heat.
30nF after all isn't so-much-big, and with appropriate series resistor a single UCC would probably work well too, although it would run a bit hot. You must note that most of this power anyway will be lost in a resistor, not the driver itself.
In DRSSTC's some people paralleled UCC's simply by stacking them atop each other, so you shouldn't have problems with that as long as you take care for cooling.
Marko, Sun Sept 30 2007, 09:28AM
For CW drive, I would sample some of those TO-220 packaged UC3710's. They do a great job dissipating heat large amounts of heat.
30nF after all isn't so-much-big, and with appropriate series resistor a single UCC would probably work well too, although it would run a bit hot. You must note that most of this power anyway will be lost in a resistor, not the driver itself.
In DRSSTC's some people paralleled UCC's simply by stacking them atop each other, so you shouldn't have problems with that as long as you take care for cooling.
Re: Paralleling Gate drivers?
Steve Conner, Sun Sept 30 2007, 12:28PM
In audiophile circles, there is a fad for paralleling TDA1541 digital-to-analog converters right now. They run hot. I once saw a guy make a tower of them soldered together, the same way as people here stack DIP-packaged UCCs, but with copper fins wedged in between with thermal grease.
Or, the UCC drivers come in a SMD package that has an exposed metal chip header (called a "Powerpad") that you can bodge onto a heatsink somehow. This allows it to dissipate a lot more than the DIP version.
Steve Conner, Sun Sept 30 2007, 12:28PM
In audiophile circles, there is a fad for paralleling TDA1541 digital-to-analog converters right now. They run hot. I once saw a guy make a tower of them soldered together, the same way as people here stack DIP-packaged UCCs, but with copper fins wedged in between with thermal grease.
Or, the UCC drivers come in a SMD package that has an exposed metal chip header (called a "Powerpad") that you can bodge onto a heatsink somehow. This allows it to dissipate a lot more than the DIP version.
Re: Paralleling Gate drivers?
Kolas, Sun Sept 30 2007, 12:40PM
Yes i've read the smallest package UCCX732X can disipate upwards of 3 watts per unit. however with pin spacing of .65mm I cannot work with them.
The T-package UC3710 slipped my mind; I forgot that I had a pack of them coming already. Since each of those can dissipate 25W I will use them. Thanks for you guy's input.
Kolas
Kolas, Sun Sept 30 2007, 12:40PM
Yes i've read the smallest package UCCX732X can disipate upwards of 3 watts per unit. however with pin spacing of .65mm I cannot work with them.
The T-package UC3710 slipped my mind; I forgot that I had a pack of them coming already. Since each of those can dissipate 25W I will use them. Thanks for you guy's input.
Kolas
Re: Paralleling Gate drivers?
thedatastream, Sun Sept 30 2007, 08:49PM
Well those audiophools will do anything!
Yes your calculations show a required power of 5W, but where is that power dissipated? The UCC drivers should be quite efficient at transferring power from input to output due to their low internal impedance. The series resistor in the gate line sounds like it will dissipate a lot of power, the high peak currents will cause some I^2.R losses and the rest will just be current returned to the power supply.
So really, I wouldn't have thought that your UCCs would be dissipating 5W. I have a feeling that you would know about it if they were!
thedatastream, Sun Sept 30 2007, 08:49PM
Well those audiophools will do anything!
Yes your calculations show a required power of 5W, but where is that power dissipated? The UCC drivers should be quite efficient at transferring power from input to output due to their low internal impedance. The series resistor in the gate line sounds like it will dissipate a lot of power, the high peak currents will cause some I^2.R losses and the rest will just be current returned to the power supply.
So really, I wouldn't have thought that your UCCs would be dissipating 5W. I have a feeling that you would know about it if they were!
Re: Paralleling Gate drivers?
Kolas, Mon Oct 01 2007, 03:59AM
well I do know, they constantly ran extremely hot, and the uccxxxx2's always burned out first. however the simpleton setup i used included NO gate resistors. Therefore most of the power was infact being dissipated by the gate drivers them selves.
Kolas, Mon Oct 01 2007, 03:59AM
well I do know, they constantly ran extremely hot, and the uccxxxx2's always burned out first. however the simpleton setup i used included NO gate resistors. Therefore most of the power was infact being dissipated by the gate drivers them selves.
Re: Paralleling Gate drivers?
ragnar, Mon Oct 01 2007, 12:13PM
The UCC37322s burn out first because the positive output is usually longer than the negative output, given a 50% duty cycle. Compare the outputs of UCC37321 and UCC37322 for the same input signal, at 2MHz or so, and you'll know what I mean.
I think I documented this pretty well in a thread about UCCs running at 17MHz. Note also that the P-channel MOSFET that pulls the load positive has a higher resistance than the N-channel MOSFET that pulls it to ground again, so there's more dissipation in that half.
ragnar, Mon Oct 01 2007, 12:13PM
The UCC37322s burn out first because the positive output is usually longer than the negative output, given a 50% duty cycle. Compare the outputs of UCC37321 and UCC37322 for the same input signal, at 2MHz or so, and you'll know what I mean.
I think I documented this pretty well in a thread about UCCs running at 17MHz. Note also that the P-channel MOSFET that pulls the load positive has a higher resistance than the N-channel MOSFET that pulls it to ground again, so there's more dissipation in that half.
Print this page