Indeed, I have experienced this with disastrous results. I turned off one of my Pete Millet amps w cap multiplier, then switched it back on -- it was like a charge pump and "there be smoke".
Thanks, Rod.
jackinnj; the use of switching Mosfets in the Pete Millett amps is what makes me think that there must be some fairly predictable middle route between blithely following the SOA graphs of switchers and predictions of explosive destruction if, for instance , expensive IXYS L-series devices are not used (although it may be a no brainer for transmitter valve amps).
jackinnj; the use of switching Mosfets in the Pete Millett amps is what makes me think that there must be some fairly predictable middle route between blithely following the SOA graphs of switchers and predictions of explosive destruction if, for instance , expensive IXYS L-series devices are not used (although it may be a no brainer for transmitter valve amps).
I use the switching types as grid driver mosfets all the time. I've never had one die. I've also used them in Maida-style regulators. I just try to stay very well away from the limits of the SOA.
I just used the linear fet in this project because it is a 750V transmitter tube supply and I really don't want any chance of this going up in smoke.
I just used the linear fet in this project because it is a 750V transmitter tube supply and I really don't want any chance of this going up in smoke.
Looking at the SOA (blithely), of the (now discontinued) FQPF8N60 is at the edge of the chart for 340V operation with the DCPP Amplifier. The cap multiplier is going to get excised.
In the DCPP amp there is only about 15V across the Mosfet when the amp is up and running. The gate will be at about 2/3 of the final voltage in about half a second, long before the amp draws current. The 47uF cap might worry me, but I think the time constant of the gate filter may be about 2 orders of magnitude greater than that of the output impedance at the source (someone please correct me if I've eyeballed that incorrectly) in combination with the 47uF cap. I would expect this would avoid any large current surges at start-up. I see this as conservative design vis-a-vis SOA. Switch off may be a bit more on the edge.
Ooops, my bad, I was forgetting the time to warm up the output tubes. My smoke experienced had to do with rapidly flicking the amplifier back on.
You mean an uncharged capacitor at startup? Yes.
Surviving a permenant dead short would depend on heatsinking and SOA
Hungry output capacitors aren’t the only failure mechanism here. Power cycling will cause problems if the Vgs isn’t properly clamped. A big (ish) capacitor directly on the gate will remain charged, and when the input voltage collapses, BANG. Power reapplied, dead mosfet. This does get worse with an output cap, which may also remain charged. The solution is a RESISTOR between the cap being multiplied and the gate. Zener on vgs, fast diode on Vgd. 100 ohm 2W MOX doesnt add any real amount of noise, and limits peak clamp currents well enough for a few events back to back (might not take cycling on and off repeatedly all day). I use exactly this with a regulated supply, where it’s not just a cap but a string of zeners - bypassed by a 100 uF cap to quiet them. Conceptually, it is similar. I test it by collapsing the input with and without load, and with momentary shorts. I also use single slope SOA protection when drain resistors can’t be employed, which isn’t that hard to add. If you want it to survive a long duration short that becomes a must-have.