Why Does My Gut Tell Me I'm Going To Blowup A Capacitor?

[DDCE88]

Okay I'm Building A Capacitance Multiplier Circuit to filter out a 2Vpp ripple coming from a PWR supply I got off ebay. The Ripple Is centered around 34V
It was sold to me as a 30V 500W PWR supply. But apparently In china 30V means 34V with a 2Vpp ripple at 120Hz. 
So from there I was said okay I can figure out how to work with it.
I've designed a CAP Multiplier Circuit to filter it out and I know I'm never going to ask more than 100W from it. But I'm designing for 200W.
I've already tested the power supply with a load to make sure it will push at least 450W. And It does this  into a load resistor network no problem. (And it makes a great space heater too)
So I fired up good ol' LtSpice and got to work.

And here are the output wave forms.

Now you can see on the input cap C1 there is a >1A spike for ~100 micro sec causing a >50W spike across the cap for the same duration.
The cap I've selected for it will be rated for 100V with ESR of 45m.
I also found this in the Cap's datasheet.

Why do I still feel like that cap is going to blowup?
Thanks for any thoughts or suggestions.

[bob91343]

What usually destroys a capacitor is heat.  If you have a brief spike of current, the energy is small and can only cause local heating.  If that heating causes a rise in temperature above what the design expects, there will be a shortening of life.  You can integrate the pulse (make it last longer and be of reduced duration) with a series inductor.

But worrying about some potential fault because you have a feeling in the pit of your stomach seems to be an unscientific approach.

[T3sl4co1l]

Rather more worried about the lack of gate voltage protection, and current limiting.

Capacitors can handle overload for whole seconds, semiconductors for milliseconds at best.

Pedantic note: that's not a capacitance multiplier, because the MOSFET is not a current multiplier -- rather it's just a filtered source follower.  The same can be said for the BJT emitter follower case, but because h_fe is reasonably stable, it can be said to have a capacitive output characteristic, whereas here it's largely the source impedance r_s ~= 1 / y_fs.

Tim

[DDCE88]

I assume you mean a zener diode from Source to Gate, for gate voltage protection. As far as input protection I am going to a thermistor on the input to do some inrush limiting but I have never been able to successfully model one in spice.
And I apologize for calling it a capacitance multiplier its just what I've always called the circuit even my professors call it that.

[DDCE88]

Gave it a crack with protection circuitry at least what I think you were talking about @T3sl4co1l.
What do you think?

[T3sl4co1l]

In event of output short circuit, will D1 actually survive discharging C7?

What's D2 doing?

Limiting Vgs will actually have the effect of limiting output current, but not very well (transconductance is very high).  I'd go with a more traditional source-shunt resistor, and BJT.

It's possible for the MOSFET to have low enough resistance that, under fault conditions, it dumps all the current the input side can offer; but this is usually only reasonable at low voltages, where the peak power dissipation is lower, and you're likely to have a beefier FET anyway.  You'll have to check the SOA curves and other data to see if that's acceptable here.

It's also fine to go without, just let it blow up should that ever happen (shorting).  It might never happen in a finished project.  Or if this is more of a breadboarding thing, or something that'll be wired externally, etc., it may be somewhere between a good idea, and mandatory!

Tim

[DDCE88]

@T3sl4co1l: D2 was a suggestion from a professor he didn't explain why. Just told me to put it in there and see what happens. He just shrugged when I told him nothing happened.
I did add an over current circuit.
I wanted something that would reset when I unplugged the circuit, and I didn't want to use a relay. What do you think. I may have gone way too complicated.

 

[T3sl4co1l]

Yeah, a bit complicated...

Also, how often will you have a load that can be low-side switched?  Can be a big downside to that approach, unfortunately.  Upside, you have total control of thresholds and timing, and it's easily programmed (i.e. to add an external reset signal or whatever).

Foldback current limiting will do the same sort of thing -- you can fold back to "below" zero current when the output is heavily enough loaded.  Downside, it won't start up into a nominal resistive load.  (If the load is intermittent, or doesn't turn on until some startup voltage, it may be okay.)

This is all that's needed: https://www.eeeguide.com/wp-content/uploads/2016/09/Foldback-Current-Limiting-1.jpg
A resistor may also be needed from +V to the current sense transistor base, this allows full cutoff at low voltages.

Instead of the control circuit, you can use the RC filter, just wire in a series resistor between the capacitor and gate (and put a zener from source to gate), so that peak current into the gate circuit is limited.  Use a pulse-rated resistor.  Also, you should need little if any output filter capacitance; and if you do, it should probably have some ESR to maintain stability.  (A ferrite bead on the gate pin may also be helpful -- MOSFETs generally don't like being in the linear range with small capacitances strapped across them, in this case the zener diode of all things.)

Tim

[Bud]

This may come as a shock to you , but simulators are not always correct. When in doubt, verify using an other simulator. Try Tina TI for instance. I remember modeling a Gilbert cell mixer and tried several simulators, some of which like Orcad one were clearly off the mark. The one that worked properly was the old good HP one.

[DDCE88]

I think I came up with a better solution.
I'm using an lm317.
This is giving way better results on the output rather than a few mV of ripple I'm getting uV of ripple. Already ordered the parts for this one and a I am excited to try and check this out.
I only wish I could set this up to regulate automatically based off of the input. But that will be a point to work on eventually. Also this really only my second or third project with BJT's I've ever done I usually only ever use MOSFETS, and they have become sort of a safety blanket for me.
So I have a soft start set up running in to the regulator and then into a current fold back circuit. Still not totally thrilled about the fold back method because it doesn't just shut off the circuit when it encounters an over current event. but I'm also wondering if I'm not using a large enough sense resistor. It appears that once you go low enough in load resistance (IE Rload<=1) current seems to spike back up.
Another thing that is great about this circuit VS the old one is the thermals are much more manageable I was really having trouble finding a heatsink that would work with in the footprint I wanted for the final PCB.