Interesting topics about electronics, product development and more.

Common cause of EMI in power electronics - the diode

by Timo Stehmann7. May 2013 16:47

If you were involved in some power electronic project you probably got involved in a discussion about power MOSFET drivers. More specific the one statement: "Listen, don't use that driver. It always latches up and acts weird. Rather use this fancy opto-isolated driver." Yes, that stupid driver! Bad driver! Do what all professionals do and use that opto-isolated driver!

I started becoming skeptical about this statement when one person mentioned an IR2102 as being one of those bad drivers. For those who don't know the driver: This is an International Rectifier high- and low-side driver rated for 600V. I never had problems with them. Why was it a problem now? The simple answer is that the latching drivers pointed out an inherent problem in the power supply. It was acting as a giant EMI emitter! Using fancy drivers just masked the problem. Sure, there are probably drivers out there that are too susceptible to EMI, but in general they are relatively robust. So, which component in a switch mode power supply can emit the most noise? The inductor? Maybe. The transformer? Mmmh, don't know. No, in many cases it is that reverse/free-wheeling diode! Yes, that one! More specifically the reverse recovery mechanism of the diode.

This famous picture of the reverse recovery of a diode tells a simple story. The turn-off of the diode can be very rapid. Especially, the second half of the reverse recovery. Here the dI/dt can exceed the dI/dt of the MOSFETs in the power supply. Combined with the inductance of the conductors (i.e. PCB tracks) this can cause significant voltage spikes both on the signal ground and over the MOSFETs. These spikes can confuse the hell out of any MOSFET driver. In some cases it can turn the driver on again with detrimental results.

The first strategy to eliminate the problem is to use soft recovery diodes where the second half of the recovery has lower dI/dt:


However, this is the one reason why this problem often goes "undetected". Most MOSFETs already have build-in reverse diodes and people don't think much about the diode. The focus is on buying a kick-ass switch. An additional external diode can always be added, but I never liked this, because there is always the question of current sharing. However, for lower voltage applications (<200V) it is possible to use additional parallel Schottky diodes if the MOSFETs use ultra-fast diodes. The Schottky has much better turn-off characteristics and a lower forward conducting voltage compared to the ultra-fast diodes used in MOSFETs.

The second strategy is the poor man's solution, but very effective. With modern MOSFETs the switching speeds increase every year. The severity of the turn-off snap recovery is a function of the MOSFET switching speed. A MOSFET turn-on is what caused the diode to turn off in the first place. So, a simple solution is to slow the MOSFET down. Yes, why use fast MOSFETs. Well, we just want to slow down the MOSFET turn-on:

Yes, as simple as that. If you have an existing design that gives this kind of problem, just rip out that small gate resistor and replace it with 10 to 20 times bigger resistance and solder a diode (e.g. the trusted 1N4148) over it. The power supply will run a bit hotter due to higher switching losses, but at least it will work. Another variation on the theme is this:

Use a low-ohmic resistor for Roff and again something 10 to 20 times bigger for Ron.

The third strategy is a complete or partial re-design of the power supply topology and get rid of hard-switching. ZVS and/or ZCS remain hot topics. Here there are many different topologies. The more difficult ones to understand and control are the resonant inverters and in general there are much less PWM controllers available. Most controllers are aimed for the traditional tried and trusted Buck, Boost and flyback power supplies.However, for power supplies with output power above 500W (even 100W) there is no other option if you are serious about your product.

On the other end of the spectrum are the passive snubbers with energy recovery. These are also very promising and enable you to change your high-power Buck or Boost converter into a ZVS/ZCS power supply. In essence the passive snubbers add small resonant inverter components to your power supply and only play a role during the switching transients.

That is all I have time for today, but now you know ... that stupid diode! Bad diode!

Happy designing!

Tags: ,

Power electronics

Permalink | Comments (0)

Add comment

  Country flag

  • Comment
  • Preview