Offline halfbridge smps gate drive

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hello,

I am currently designing an offline halfbridge powersupply. The supply uses UC3825 and IR2110 to drive IRG4PC50U IGBTs. These are connected to the 2110 by 10Ohm resistors and 1N4148 antiparallel diodes and a 10k from gate to emitter. The supply is running at 115kHz.

Now, when i look at the gate emitter waveform of one IGBT on my scope, i see a spike of about 2.5V at the moment the other IGBT turnes on.

I don't seem to be able to get rid of this. I allready disconnected the trafo and replaced it with an omic load, and it does not make any difference. When I remove the power to the supply and leave only the 3825 and the 2110 running, the spike disappears.

I assume this is a problem because the IGBT can turn on at Uge=3V. Can anyone help?
 
This is a usual problem with half bridge SMPS.
I'm now developing a 2000W, 4300V output SMPS and I have got the same problem.
The cause are parasitic capacitance between IGBT gate and collector, giving you parasitic current throught the gate and, thus, spikes.
There's a lot of info on the net, please check HERE and HERE.
In my projects I usually do prefer to drive the gate of the IGBT negative fduring its turn-off period, in manner to be absolutely sure it never turn-on.
I'm currently using two IXDH30N120D1 igbt, DRIVEN by two IXDD414 drivers; each driver has its floating, separate power supply.
Each driver is NOT powered by a sdingle power supply, but by a dual supply (+15, -15V). the IGBT gate is connected to the driver output, the emitter is instead connected to the common point of the dual supply.
In this way I can use the driver to drive each IGBT gate up to +15V and down to -15V.
You can also develop a full bridge driver; in such a way you can use a single 15V supply for each driver.
Of course, if you have a floating driver you must drive it through an optocoupler, like the 6N137.
The "Active Miller clamp" does only work if carefully designed, the gate capacitor will make the driver job harder; if you don't mind please follow the negative supply path.

Ciao,
Giovanni
 
I'm not sure you can do it with IR2110, never tried it.
Probably -probably- it can be done.
First of all, remember the IR2110 maximum supply voltage of 20V: you cannot use 15+15V=30V or you'll burn the driver.
You can instead use two different supply, +15V and -5V, for a total of 20V.
I reccomend you to use a separate floating dual power supply for each driver, because the charge/discharge currents are so high they can interfere with the rest of the circuit.

For the low side:
Vcc = +15V
COM = -5V (actually you must move the COM pin to -5V)
The central point of the dual supply must be connected to the IGBT emitter.
Remember you must also change the input circuitry, because now the Low state is reffered to -5V, not to ground.

For the High side:
Vb = +15V
Vs = -5V
The central point of the dual supply must be connected to the IGBT emitter.

I'm quite sure you must leave the Vss and COM pins at the same potential, just to avoid problems.
I repeat, I never done it, you must try it and find the right circuit.

Ciao,
Giovanni
 
Schematic:
 

Attachments

  • gate drive.jpg
    gate drive.jpg
    78.5 KB · Views: 925
I don't like it, I'm sorry. I'm afraid it won't work.
if you have a parasitic current flowing from the collector to the gate through Cgc, the internal impedance (ESL) of the capacitor you placed on the gate will further increase the spike.
You really need something that will keep firmly the gate to ground or, better again, to a negative potential. Without any impedance in between.
I told you before that I don't like the "Active Miller Clamp" solution explained HERE, because the IGBT internal impedance, summed to the BJT internal impedance (some nH) make the solution useless with the frequency you are using. Imagine what the internal impedance of a capacitor can do.

Look at THIS: as you can see, the Cgc is loaded with the full bridge supply voltage, and thus we have a current flowing through it trying to load the capaciutance to, let's say, the supply voltage.
Any impedance placed between the gate and the Cgc loading path will create a voltage potential, and this is what your capacitor will do.
I've seen a 5V spike induced by a 4 cm. cable, measured with my scope at the ends of the cable (connecting the driver and the gate).

There's another way to reduce the spike: increase the turn-on time for both the IGBT, reducing the dV/dT and, thus, the current through the Cgc. But you can't inrease too much the Ton time, the IGBTs don't really like it.

Ciao,
Giovanni
 
The outputs of the IR2110 switch between 0V and +20V. The capacitor and the zener clamp DC-offset this signal to -5V to +15V. So the gate is at -5V during its off period isn't it? And isn't the capacitor's internal resistance negligable compared to the 10 ohm resistor in series with it? I can put an 1N4148 antiparallel to it.

I have just made the circuit on my breadboard and it seems to be working. The spike is still 2.5V, but because it is superimposed on -5V, Uge is still negative.

Or am I overlooking something?
 
It's depend.
With High voltage and/or fast turn-on the local turn-off PNP transistor could not work: the IGBT and BJT leads, themselves, could introduce up to 50 nH of inductance, which means that on fast spikes the solution is useless (if you have a 20nS turn-on time, the spike could reach frequencies up to 200Mhz, and thus the impedance of the BJT is too high, and the BJT itself is too slow).
I tried the PNP BJT solution with my application (half bridge driving a resonant transformer at 300V, about 2KW, 30nS of turn-on/off): nothing to do, the spike was always there.
With low voltage and not too fast timing the solution is perfect.
Please let us know your circuit parameter (main voltage,turn-on time, current, etc) and I will check it.

Ciao,
Giovanni
 
The IGBT's I use are IRG4PC50U.

The supply runs on 230V and must provide power for a 2000Wpk audio amp. I have an ETD59 transformer with 3C90 cores and 14 turns on the primary. (Is this ok?)

I have tried the PNP shut down scheme, but it did not work. I have even tried it with a very low Uce(sat) transistor (FMMT718 from Zetex).
 
Well,
with the parameters you listed (230Vac -> 310Vdc, 2KW) simply forget any other solution than negative turn-off voltage.
Increasing gate resistor increase turn-on time, but also increase the spikes when the other IGBT turn on...
Right way to proceed (In my opinion) is to calculate the gate resistor in manner to have the desidered turn-on time (let's say, 100ns), using the formulae you can find in the document switchmodepower has linked in his previous post, then design the driver for negative voltage.

Are you sure the IGBT you are using are the correct one? They are rated for 600V, you are working with 310V and more (you can go up to 350V with line fluctuation), it seems to me their voltage is underrated.
More important, are you using a resonant circuit? if you are using the IGBT in "hard switching" mode, they will not be happy with 115Khz.

Ciao,
Giovanni
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.