Is antiparallel diode crucial?
dexter, Mon Sept 21 2015, 12:07PMI'm in the process of building a small DRSSTC using a half bridge of To247 IGBT's
after receiving all the parts i noticed that i ordered IGBT's without antiparallel diode...
How crucial the antiparallel diode is?
Can the coil work without it?
Or i need to order the version with antiparallel diode?
Re: Is antiparallel diode crucial?
Dr. Slack, Mon Sept 21 2015, 01:15PM
In theory, if the drive circuit commutates the IGBTs perfectly, with no under-lap, then the diode has nothing to do.
In practice, as you work to avoid overlap, to avoid shoot-thru, you will end up with some finite amount of under-lap.
What happens with under-lap is the bridge goes high impedance, so there is nowhere for any current flowing in the primary to go. The current will flow into the low (relatively low) capacitance of the node between the two off devices, changing its voltage rapidly.
There may be several reasons a bridge can still function like this without going bang.
a) You've got zero current switching really sorted, and there's no current flowing
b) the device capacitance is high enough to limit the voltage rise
c) one of the switching devices avalanches the energy safely
... but mess up on one or more of those and you are replacing fried devices.
Many people do not trust the performance of the anti-parallel diode, and fit an additional well-specified fast recovery device in parallel anyway. Don't throw the IGBTs away, just buy some suitable diodes and add them.
Dr. Slack, Mon Sept 21 2015, 01:15PM
In theory, if the drive circuit commutates the IGBTs perfectly, with no under-lap, then the diode has nothing to do.
In practice, as you work to avoid overlap, to avoid shoot-thru, you will end up with some finite amount of under-lap.
What happens with under-lap is the bridge goes high impedance, so there is nowhere for any current flowing in the primary to go. The current will flow into the low (relatively low) capacitance of the node between the two off devices, changing its voltage rapidly.
There may be several reasons a bridge can still function like this without going bang.
a) You've got zero current switching really sorted, and there's no current flowing
b) the device capacitance is high enough to limit the voltage rise
c) one of the switching devices avalanches the energy safely
... but mess up on one or more of those and you are replacing fried devices.
Many people do not trust the performance of the anti-parallel diode, and fit an additional well-specified fast recovery device in parallel anyway. Don't throw the IGBTs away, just buy some suitable diodes and add them.
Re: Is antiparallel diode crucial?
dexter, Mon Sept 21 2015, 05:08PM
right now i only have lying around some MUR1560 60ns recovery but with only 15A continuous 30A peak and 150A single pulse... which i don't think is enough
i can order some fast (130-200 ns) diodes with 75 and 100A continuous but they cost half the price of a single IGBT so i think i'll just order new IGBT's
also adding extra diodes complicate the layout, they'll require some heatsink and their speed is to low compared with 55ns of the IGBT with antiparallel diode...
dexter, Mon Sept 21 2015, 05:08PM
right now i only have lying around some MUR1560 60ns recovery but with only 15A continuous 30A peak and 150A single pulse... which i don't think is enough
i can order some fast (130-200 ns) diodes with 75 and 100A continuous but they cost half the price of a single IGBT so i think i'll just order new IGBT's
also adding extra diodes complicate the layout, they'll require some heatsink and their speed is to low compared with 55ns of the IGBT with antiparallel diode...
Re: Is antiparallel diode crucial?
Wolfram, Mon Sept 21 2015, 08:58PM
This analysis works fine as long as you're driving the primary. The problem appears when the burst is over and all IGBTs turn off. With the diodes across the IGBTs, most of the energy left in the tank at this point is rectified by the reverse diodes and returned to the DC bus capacitor. Without the diodes, the voltage across the IGBTs will rise until you get reverse avalanche, which usually means instant IGBT death. You can combat this by leaving the bottom IGBTs in the bridge turned on after the burst, to trap the energy in the primary, but this is not possible when you drive the IGBTs from GDTs. Getting new IGBTs with built-in diodes sounds like the best approach.
Unlike MOSFET reverse diodes, IGBT reverse diodes are not parasitic, so they can be optimized without sacrificing IGBT performance. This means that they usually don't need to be bypassed. Note that the internal diodes often have lower current handling capability than the IGBTs themselves, so external diodes can be an advantage with certain topologies where the diodes see significant average current, but in DRSSTCs, this is not usually a problem.
Wolfram, Mon Sept 21 2015, 08:58PM
Dr. Slack wrote ...
In theory, if the drive circuit commutates the IGBTs perfectly, with no under-lap, then the diode has nothing to do.
In practice, as you work to avoid overlap, to avoid shoot-thru, you will end up with some finite amount of under-lap.
What happens with under-lap is the bridge goes high impedance, so there is nowhere for any current flowing in the primary to go. The current will flow into the low (relatively low) capacitance of the node between the two off devices, changing its voltage rapidly.
There may be several reasons a bridge can still function like this without going bang.
a) You've got zero current switching really sorted, and there's no current flowing
b) the device capacitance is high enough to limit the voltage rise
c) one of the switching devices avalanches the energy safely
... but mess up on one or more of those and you are replacing fried devices.
Many people do not trust the performance of the anti-parallel diode, and fit an additional well-specified fast recovery device in parallel anyway. Don't throw the IGBTs away, just buy some suitable diodes and add them.
In theory, if the drive circuit commutates the IGBTs perfectly, with no under-lap, then the diode has nothing to do.
In practice, as you work to avoid overlap, to avoid shoot-thru, you will end up with some finite amount of under-lap.
What happens with under-lap is the bridge goes high impedance, so there is nowhere for any current flowing in the primary to go. The current will flow into the low (relatively low) capacitance of the node between the two off devices, changing its voltage rapidly.
There may be several reasons a bridge can still function like this without going bang.
a) You've got zero current switching really sorted, and there's no current flowing
b) the device capacitance is high enough to limit the voltage rise
c) one of the switching devices avalanches the energy safely
... but mess up on one or more of those and you are replacing fried devices.
Many people do not trust the performance of the anti-parallel diode, and fit an additional well-specified fast recovery device in parallel anyway. Don't throw the IGBTs away, just buy some suitable diodes and add them.
This analysis works fine as long as you're driving the primary. The problem appears when the burst is over and all IGBTs turn off. With the diodes across the IGBTs, most of the energy left in the tank at this point is rectified by the reverse diodes and returned to the DC bus capacitor. Without the diodes, the voltage across the IGBTs will rise until you get reverse avalanche, which usually means instant IGBT death. You can combat this by leaving the bottom IGBTs in the bridge turned on after the burst, to trap the energy in the primary, but this is not possible when you drive the IGBTs from GDTs. Getting new IGBTs with built-in diodes sounds like the best approach.
Unlike MOSFET reverse diodes, IGBT reverse diodes are not parasitic, so they can be optimized without sacrificing IGBT performance. This means that they usually don't need to be bypassed. Note that the internal diodes often have lower current handling capability than the IGBTs themselves, so external diodes can be an advantage with certain topologies where the diodes see significant average current, but in DRSSTCs, this is not usually a problem.
Print this page