Mosfet driver IC's

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
They can simplify things but I believe they can also be prohibitive, mainly they are not a plug in solution and should not be viewed as such, they need some pampering.

There's really no reason to hate them either, they can be made to work. I believe manufacturers typically dislike them because they can build their own at a small fraction of the cost.
 
Chris: I would sure like to see how you would propose a discrete circuit that works as fast as a driver IC. Especially with low cost standard components. Delay and rise times are critical.

Look at this scope shot, showing a 120V jump (1:10 probe) in under 10 nS, using a commercial driver IC.

10ns.jpg


I have been working on beating this performance for years without finding a good discrete solution. Let alone cheaper. So your ideas would be very interesting :D
 
Kind of puts words in my mouth. I'm not saying such a simple driver can do the job at high voltage/high speed, but for typical class d at lower voltage I'm positive it's cheaper to use a few transistors than a driver IC, especially when you start burning them out, so I don't think "always cheaper" can really apply.

When I come up a better discrete version for highpower/speed though, be happy to sell it to you :D
 
OK, i wish you good luck! :D

Anyway i dont think the price of the driver solution, be it IC or discrete is of any significance, because it's only a small fraction of the whole circuit. And they roughly balance.

But if higher performance could be achieved with a discrete solution, i don't think anybody would mind an extra cost involved ;)
 
Very good, and i guess, cheap discrete driver is does exist! And it's UcD again. So UcD400 switching for 120v/20nS. For true-virtual comparison several drivers, we would be try to fit THD figures by the loop gain multiplier to the zero feedback state. E.g. UcD400@10w 30db THD=.007%, then THDopenloop=.007*31.6=.22%, for 100w .02*31.6=.63% etc. Anybody ready?
 
While we're on the subject, here's a discrete gate driver I came up with a couple weeks ago. It's only been built in spice, not real life. I was playing around with a UcD schematic classd4sure made and was unhappy with the amount of component dependency so I decided to take the comparator and feedback loop and design a new gate driver from scratch with no preconceived notions as to how it "should" be done.

Everything right up to the EF stage before the MOSFETs runs in current mode with little voltage swing so it should be pretty fast. Even though it's shown as UcD that may not be optimal, it requires more comparator drive than UcD feedback supplies at high outputs.

The .ZIP contains the .asc LTSPICE schematic file, and a .GIF version for those of you without LTSPICE. It was created on a clean four month old windows XP Pro x64 system running Symantec Antivirus and is completely safe.

So what do you experienced designers think, is this a workable design?
 

Attachments

  • mirrored-drive.zip
    63.2 KB · Views: 452
Hmmm, it's interesting.

Some of the interactions may be slightly reduced but it looks like they're all still there, even with 20V rails.

Watch how the thresholds and shoot through vary with output level.

I think the worst part would be trying to build in some adjustability over the gate waveforms with this one to achieve good performance, dead time, slew rates..
 
Tim, i think you have overlooked the problem of saturation delay in your voltage amp transistors. That would give uncontrolled turnoff delays in each transistor. Try looking at LS logic, how they have solved this problem. We are dealing with the same kind of speed here.

Further the devices you have used are not the fastest on earth..

Basicly you have only 3 ways of speeding up a drive circuit,

1..Use transistors with extremely low Ccb (can only go so low).
2..Use high collector current and high base drive to fight Ccb capacitance.
3..Make an ingenious circuit where the collector-base voltage is constant.

All the best from

Lars
 
"Try looking at LS logic, how they have solved this problem. We are dealing with the same kind of speed here. "

Ah, clamping the base drive, good idea, though it does significantly increase component count.

"Further the devices you have used are not the fastest on earth.. "

Yeah I know, I tend to SPICE with positively average transistors, I couldn't tell you why :cannotbe:

"3..Make an ingenious circuit where the collector-base voltage is constant."

Impossible to do directly before the emitter follower, and as far as I can see impossible to do with a bootstrapped high side gate drive. One could cascode the appropriate transistor and move the voltage swing to a different C-B junction, maybe even some sort of variable cascode putting half the total voltage swing across each C-B capacitance, I'll attach what I'm thinking.

P.S. The four Rs on the bottom do allow some optimization, I haven't fiddled with them too much, but it seems slightly higher values for R6 and R7 (15-33ohms) might be better.
 

Attachments

  • cascode.gif
    cascode.gif
    7 KB · Views: 799
Watch how the thresholds and shoot through vary with output level.

Yeah I noticed that, like I said " it requires more comparator drive", maybe giving the LTP more current gain/lower impedance as seen looking into the emitters would fix it, sziklai pair perhaps?

Edit: Ok, here it is with a cascode and a sziklai pair LTP.
 

Attachments

  • mirrored-drive2.zip
    1.9 KB · Views: 141
Interesting stuff. I was doing some work a while back on variable chopped AC supplies and we replaced the entire mosfet drive section with a few descrete parts and a pulse transformer. We were getting mosfet gate transition times of about 20ns as I remember. I'm just laying out the board to try it on the output of an amp, much more interesting than controlling the speed of motors. :)
 
Hi all,

Just thought i'd stick my oar in with my own experience on this matter. I've tried alsorts to get a clean switch in under 50n/s using discrete components. Every time i could only get under 50n/s in one direction. Either low to high or high to low. The gate charges i were driving into were not massive either, about half a nan or so. Thats why i gave up trying to use discrete although to this day i wish i could. Instead i have settled for a maxim part to do the job. They are admitidly a bit delicate and need a little pampering, but for the price, convenience, speed(40n/s each way) and space saving they are unbeatable. I still occasionally try to find a good discrete alternative whenever i have some spare time and empty breadboard but so far have failed. I need high switching speed as i always drive my amps between 500Khz and 1Mhz. P.S for those who are wondering i've found a new job:D

Have fun all,
Mad.P
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.