Gate drive push pull(FDS8958) heating up

[prabhatkumar]

Hello all, I have been working in building a discrete gate drive system just like Loneoceans SSTC4, but then encountered certain observations which I could really not answer.
My setup is as follows. For testing purposes I have used both outputs of my function generator configured as 180 degree out of phase to mimic my tesla coil signals going into it. I am using a TC4427 MOSFET driver since that is the one which is available locally.
Now the output is as expected and is also driving the primary of the GDT without any issue( No load on secondary), but then there are few observations about the current drawn from the power supply.
I have used a input voltage of 15 volts and then using a 9 volt linear regulator to power the TCC4427. Now with the exact same circuit diagram and a frequency of 100kHz, the current drawn from the gate driver was 50-60mA (without the GDT connected). Now as soon as i put the 4.7uF capacitor near the FDS8958, the current drawn from the power supply rises to 120mA. This makes the FDS8958 really toasty very soon though the output remains as good as before. This is the problem which I am not able to explain. How can the circuit draw different current at different capacitances.?
 Also please tell me how much current should the circuit draw while idling( not driving GDT)
Thank you.

[davekni]

The schematic image ended up too low resolution to read.  Can you post a larger version?

[prabhatkumar]

I really apologise for the problem. I am reposting it here.

[davekni]

Are you adding C7 and/or C8 to cause the increased current?  Running without those two local bypass capacitors will cause behavior dependent on your layout details - parasitic inductance of power traces etc.  I suggest ignoring any measurements without C7 and C8 in place.

Unused logic inputs should be tied to ground or +5V, not left floating.  Floating inputs can cause extra power dissipation and make ESD damage more likely.

Since you are using TCC4427 without enable inputs, I presume you aren't using IC4 and IC6.  (There is a problem with that circuit if using a driver with enable.)

What voltage are you using for the supply shown as 12V in the schematic?

 TCC4427 has lower output current than UCC27425.  Perhaps the extra current is shoot-through, since both PFET and NFET will be conducting simultaneously for a short period during each transition, a bit longer due to lower drive current.  Reducing TCC4427 supply voltage down closer to 5V will help if shoot-through is the issue.

Are you driving the primary of a current-transformer with your function generator outputs?  One input of the schematic shown is ground, so you could drive that input with a single function generator output.

If the above suggestions don't help, sharing scope traces of signals will help figure out what is causing trouble.

[prabhatkumar]

Again thanks for the quick reply. This forum is probably the only place where a noob like me can find answers to variety of problems.
As for my setup, I am just using the push-pull part from Gao’s schematic. The input voltage to the TC 4427 is the 9V in my case. The input voltage to the FET amplification stage is 15 volt.
I am only feeding in the two 180 degree out of phase pulses from my function generator to the TC 4427 directly. Is this I a problem?
Also I tried building this circuit with irf540 and irf9540 and was experiencing this issue but with much more major heating in the FETs. I suspected that there was some short of some sort which I was not able to diagnose. But then on reading through the Kaizer drssstc pages I found another user reporting very similar problems.
SS Of the comments section
I will put the scope traces also if that’s required.
For a while suspected shoot through but that’s looks for a very brief moment only.
Another doubt which I have in the Gao’s schematic, why are the P CHANNEL FETS having diode from their gate to VCC. Is it to do something with faster turn and slower turn off thing ?
Another thing I noticed, I get output on the MOSFET driver even when the supply voltage is removed and signal is applied? Is it normal ?
Thank you
Scope traces attached herewith.
1) N channel and P channel gate of first FDS8958.
2) N channel and P channel gate of second FDS8958.
3) Two outputs of the TC4427.
4) Zoomed rising part of the output.
5) Zoomed falling part of the output.

Comments section screenshots in serial order discussing about heating of FETs(from https://kaizerpowerelectronics.dk/tesla-coils/kaizer-drsstc-ii/)

[davekni]

OK, your two 180-degree generator outputs make sense now that I noticed the TC4427 is non-inverting on both channels.  Of course, you will eventually need to invert one of the inputs for Tesla coil operation.

The diodes on the FDS8958 PFET gates is unrelated to the turn-off speed boost function.  It is to make the PFET gate waveform go from 15V down to 6V.  Since the PFET gate drive is AC-coupled (through a capacitor), it would be 15V+-4.5V (10.5V to 19.5V) without the diode.

Shoot-through current explains your measurements.  Thank you for the detailed scope traces.  There is ~30ns with ~4Vgs on the NFET and ~-5Vgs on the PFET during each edge.  At those gate voltages the FETs are conducting ~20A.  30ns/5us is 0.6% duty cycle of shoot-through.  20A * 0.6% = 120mA.  (Is your 120mA measurement for both FDS8958 parts together?  If so, lead inductance and ECB inductance and/or resistance must be reducing the shoot-through current.  That also makes sense for current being lower without the 4.7uF bypass capacitors.  Supply inductance is higher, so shoot-through current can't ramp up as high.)

Solution options include:
1) Reduce TC4427 supply voltage.
2) Use FETs with higher threshold voltages.
3) Use FETs with lower current capability.
4) Add intentional resistance and/or inductance to FET source or drain leads.  Removing the 4.7uF capacitors is doing that, but may cause other issues.  It depends on ECB layout and what other components may see the resulting glitches on 15V supply.  This solution also risks voltage spikes beyond FET maximum ratings - so needs careful simulation and/or measurement.

Many ICs, apparently including TC4427, have intrinsic diodes from logic inputs to supply rails.  The signal generator outputs are powering the TC4427s through these internal diodes.

[prabhatkumar]

A big thank you again for your response. I understand that your reasons make good sense. But I was wondering how could the same circuit work for loneoceans and some other people too and why only I am getting stuck. I tried powering the TC4427 with 5 volt and now the average current dropped to 10 mA, which I think is fine.
What I am worried about is that when I am powering my circuit with 5 volt and also giving pulses which are 0-5v then does gate driver even do anything there? I know it’s a very stupid question.
Also what is other way around this problem if I had to change the driver IC and also maybe the FETs, what would be your opinion on that given I already had problems with the IRF540 and IRF9540 mosfets and TC4427 drivers.
Thank you

[davekni]

At 5V the gate driver is not boosting voltage, but it is boosting current.  The logic can drive a few mA.  TC4427 drives 1.5A.  UCC27425 drives 4A.  (That 4A drive of UCC27425 reduces output switching time, so reduces the cross-conduction time, which reduces average current.)

IRF9540 and IRF540 have higher threshold voltages, which will reduce cross-conduction current some.  However, they have much more gate capacitance than FDS8958.  The high capacitance slows transition times, so increases the duration of cross-conduction.

I suggest staying with 5V and FDS8958.  Perhaps change the PFET gate diodes to schottky.  Schottky diodes have lower forward voltage drop, so will keep more of the 5V swing on the PFET gate below 15V.

[prabhatkumar]

A big thanks for your reply. I will be staying with 5V and FDS8958 as you said. I will be trying out 1N5819 later as you said. For now I think I am content with its working. Thank you for helping !!