Deadtime in an IGBT Full Bridge
thedatastream, Fri Aug 03 2007, 12:55PMHi all,
I'm just running up my repaired bridge on Inazuma and was playing around with phase, frequency and deadtime settings on the controller. Setting the phase is fine, as is the frequency (tuned to the upper pole at the moment) but the deadtime control is giving me some food for thought.
As I understand it, the purpose of deadtime is to insert some space between the switching intervals so that shoot through cannot occur doe to slow turn off of the IGBTs. When I set the deadtime to maximum, I get a reduced power output obviously because the on time is less.
However, I get much more noise on my bridge (when operating with this amount of deadtime) from when the IGBT turns off and the diodes are supposed to take over whch I'm guessing is caused by reverse recovery time of diode and the stray inductance of the circuit.
When I operate with no deadtime, the switching waveforms are _much_ cleaner with far less ringing and voltage spikes. However I'm now concerned about some limited amounts of cross conduction due to slow turn off. I could put a current probe in series with the collector of one of the IGBTs to see what's going on and if I'm getting shoot through.
Does anyone have any opinions on the issue of deadtime? How much (if any) do people typically operate with? I'm inclined to operate with none simply because of the cleaner switching.
Thanks in advance,
James
Re: Deadtime in an IGBT Full Bridge
Steve Conner, Fri Aug 03 2007, 01:42PM
I saw exactly the same things as you did when using small, fast IGBTs, so I ended up operating with the deadtime control turned to minimum. That still gives a few hundred nanoseconds of deadtime, but nothing much.
The reason why it happens is that the current is only zero at one instant. So, with deadtime, one or other of the IGBTs must miss the zero current switching point and switch with a substantial current flowing through it, which is what causes the noise. The longer the deadtime, the worse it will get.
The deadtime control on thet driver was intended for later when I was going to hook it up to big slow IGBTs the size of pizza boxes that have fall times of over 1us. Still haven't got round to it, though.
Steve Conner, Fri Aug 03 2007, 01:42PM
I saw exactly the same things as you did when using small, fast IGBTs, so I ended up operating with the deadtime control turned to minimum. That still gives a few hundred nanoseconds of deadtime, but nothing much.
The reason why it happens is that the current is only zero at one instant. So, with deadtime, one or other of the IGBTs must miss the zero current switching point and switch with a substantial current flowing through it, which is what causes the noise. The longer the deadtime, the worse it will get.
The deadtime control on thet driver was intended for later when I was going to hook it up to big slow IGBTs the size of pizza boxes that have fall times of over 1us. Still haven't got round to it, though.
Re: Deadtime in an IGBT Full Bridge
thedatastream, Fri Aug 03 2007, 01:47PM
OK that's great, thanks Steve.
mmm... Pizza!
thedatastream, Fri Aug 03 2007, 01:47PM
Steve Conner wrote ...
I saw exactly the same things as you did when using small, fast IGBTs, so I ended up operating with the deadtime control turned to minimum. That still gives a few hundred nanoseconds of deadtime, but nothing much.
I saw exactly the same things as you did when using small, fast IGBTs, so I ended up operating with the deadtime control turned to minimum. That still gives a few hundred nanoseconds of deadtime, but nothing much.
OK that's great, thanks Steve.
Steve Conner wrote ...
...big slow IGBTs the size of pizza boxes...
...big slow IGBTs the size of pizza boxes...
mmm... Pizza!
Re: Deadtime in an IGBT Full Bridge
Steve Ward, Fri Aug 03 2007, 08:58PM
James, if you can probe your DC rails, check that out. If there is cross conduction you should see a pretty serious and short dip in the DC supply. The supply will normally have some noise from the Cce charging up during switching, but real cross conduction will be much worse. Ive never used programmable logic deadtime... but i often used the diode/resistor trick to slow down the turn ON and not the turn OFF time.
Steve Ward, Fri Aug 03 2007, 08:58PM
James, if you can probe your DC rails, check that out. If there is cross conduction you should see a pretty serious and short dip in the DC supply. The supply will normally have some noise from the Cce charging up during switching, but real cross conduction will be much worse. Ive never used programmable logic deadtime... but i often used the diode/resistor trick to slow down the turn ON and not the turn OFF time.
Re: Deadtime in an IGBT Full Bridge
Steve Conner, Sat Aug 04 2007, 09:54AM
Steve Conner, Sat Aug 04 2007, 09:54AM
Steve Ward wrote ...
If there is cross conduction you should see a pretty serious and short dip in the DC supply.
Hey, can you really see cross conduction like that? Have you actually seen it? If it's visible just by looking at the DC bus voltage, that makes life a lot easier. Well in relation to SSTCs at least If there is cross conduction you should see a pretty serious and short dip in the DC supply.
Re: Deadtime in an IGBT Full Bridge
Steve Ward, Sat Aug 04 2007, 05:42PM
This is how ive checked for shoot-through, but i cant actually remember if ive seen anything bad as far as dipping goes. Ive measured shoot-through like this on my P/N bridges and there was a bad dip until i fixed the problem with gate zeners (for proper biasing).
Steve Ward, Sat Aug 04 2007, 05:42PM
This is how ive checked for shoot-through, but i cant actually remember if ive seen anything bad as far as dipping goes. Ive measured shoot-through like this on my P/N bridges and there was a bad dip until i fixed the problem with gate zeners (for proper biasing).
Print this page