# Failure of MOSFET-Based Capacitor Dump Circuit After 168 Joules…Why? Parasitics? Thermal-Transient-Land? Jupiter?

In both simulation and reality, I'm simply unable to get this circuit to oscillate in a way that creates extra power dissipation. The only direction I can get it to go, is less power dissipation due to I think two things: 1). The good high-frequency gain of the 2N7000 and 2). The severe restriction on available current to charge Q1's gate (~340uA @ Vbuss = 340VDC). My biggest concern are transients that could be coupled onto Q1's gate via its C_gd (Miller) capacitance so in LTspice, I've inserted PCB and external-wiring inductances around Q1 and Q2, and modified these inductances in many different ways while pushing transients into the system. My hope was that something goes awry, but the current, under the shamefully few conditions that I know of to consider, seems to me to be truly clamped to around 700mA maximum, thus placing a hard limit on Q1's power dissipation at about 250W or less.

As soon as my microcontroller hits the "DUMP" button (turning off the PC817 optocoupler, hence allowing Q1's gate charge up) Q1 fails catastrophically. The three-ohm resistor makes an alright fuse, keeping the 0.6 kilojoules mostly in the capacitors unless a flying piece of this resistor-fuse gets across the buss terminals, in which case the threshold-of-hearing in my left ear went up a little last night.

280VDC remained on the buss after failure, so it took (starting_energy_578J - end_energy_392J) 186 joules to kill it.

Why is this too much to ask from a Minibloc? Either transient thermal parameters are coming into play here that I've missed, or there's sustained oscillation that I've missed.

Tj starting temp = heatsink starting temp = 30 degrees C. Heatsink: 1/4" thick anodized Aluminium, 4" x 5", 1.5" fins.

Looking forward to your response; thanks.

• Did you measure current in ixfn48n50 in the simulation? – StainlessSteelRat Dec 14 '20 at 19:09

From ixfn48n50 datasheet:

So $$R_{DS_{ON}} = 0.10\Omega$$, which means KVL (Kirchhoff's Voltage Law) when your Power MOSFET turns on is:

$$V_{DD} - V_{DS} - V_{R_G} = 0$$ $$I_{D} = \frac {340V} {0.10\Omega + 3\Omega} = 110A$$

From data sheet, the maximum current is 48A constant and 192A pulsed. And pulse width is limited by $$T_{JM}$$ <= 150°C, based on an Ambient Temperature of 25°C.

$$P = I_D^2 R_{DS} = (110A)^2 \times 0.10\Omega = 1.21kW$$

$$P = I_D^2 R_S = (110A)^2 \times 3\Omega = 36.3kW$$

This explains why your 3Ω resistor works as a fuse. This is a symptom, not the problem. The 2N7000 has no impact on this, aside from triggering the destruction.

The resistor gets more current than it can handle, so it acts as a fuse.

Now this is just theoretical numbers. The true question is as I asked in the comments Justin/Flame: Did you measure current in ixfn48n50 in the simulation?

IXYS Power MOSFET Datasheet Parameters Definition

• Hi I've made an account - I'm the OP. StainlessSteelRat, My perspective on your answer is that everything's there minus the critical role that the 2N7000 plays; which is to clamp Q1's gate voltage if its current exceeds more than about 700mA. The circuit is designed to be a constant-current sink, so I'm unsure why you've focused so closely on Rds(on), I_max(continuous) and I_max(pulsed). The magnitudes of those parameters should never come into play, but something's obviously wrong, so if you could explain how the ON-resistance ever gets that low, or the drain current could ever exceed 7 – Flame Dec 16 '20 at 21:53
• Did you read the datasheet? – StainlessSteelRat Dec 17 '20 at 0:26
• The problem was that it was simply oscillating. I'm so glad I read the datasheet. – Flame Dec 18 '20 at 20:28