Active pre-charge/inrush control

The gate voltage sounds correct. It will jump up to the gate threshold and stay there until the controller caps charge, then jump up to 12v. Version with the switch should be fine as long as the switch is off when the battery is connected.

When FETs blow, it's quite typical that all the leads short together. That's normal.

Dang, about the only other thing I can think of is you got some bad FETs. That's very rare, but not impossible.
 
I did add one thing to the circuit but it doesn't seem to be to be what would cause the issue:

I used a DPDT switch and to the second pole i connected a high voltage buck converter to power my receiver. The converter is connected straight to the negative input and the positive is connected by the switch. Would this be causing the issue? I do realize i could have just ran it in parallel with the drain of the mosfet and just use a SPDT switch.
 
Problem solved!!!!!

Fechter you where right! fets where a garbage! lesson learned don't buy from ebay or amazon!

I tried this fets with no load and they never failed!!! FDP18N50 http://www.mouser.com/ds/2/149/FDPF18N50T-1008232.pdf

im not going to use these for my switch but they where a good test!

BTW Fechter thanks so much for helping! I really appreciate it! (was going nutz!!)
 
There are lots of counterfeit parts in the low cost channels. Beware.

Driving PWM into FETs requires more drive, and FETs in parallel may not share current very well. It helps to match them, but in any case you cannot assume the rating doubles with the second FET. There are guidelines for this, as I recall assume about 50% additional for the second FET. Again, counterfeit FETs will fare far worse as they are likely even less matched.
 
Alan B said:
There are lots of counterfeit parts in the low cost channels. Beware.

Driving PWM into FETs requires more drive, and FETs in parallel may not share current very well. It helps to match them, but in any case you cannot assume the rating doubles with the second FET. There are guidelines for this, as I recall assume about 50% additional for the second FET. Again, counterfeit FETs will fare far worse as they are likely even less matched.

The PWM part wasn't towards me was it?
 
Alan B said:
Driving PWM into FETs requires more drive, and FETs in parallel may not share current very well. It helps to match them, but in any case you cannot assume the rating doubles with the second FET. There are guidelines for this, as I recall assume about 50% additional for the second FET. Again, counterfeit FETs will fare far worse as they are likely even less matched.
I definitely agree on the counterfeit parts.

However, well matched FETs (same part #/mfr, same run) share pretty well. MOSFETs have a positive temperature coefficient*; as they get hotter, their resistance goes up. Therefore the FET with the highest load tends to shed load as it heats up, and the cooler FET takes more of the load. This is a mechanism that allows paralelling MOSFETs with relatively good sharing, which is why you often see paralelled FETs in controllers.

This is in contrast to bipolars, which have a negative temperature coefficient - the hottest device takes more and more current as it heats up. This means you need ballast resistors to "force" a certain amount of sharing, but this (of course) increases losses.

(* - MOSFETs have requirements to maintain this positive coefficient, like sufficient gate drive.)
 
Since this switch is operating at zero Hz, the gate drive requirements are pretty simple. Well, things do happen during the ramp up and turn off, but in general, the FETs will share pretty evenly as long as they have similar heat sinks.

Glad the problem was solved!

I got some counterfeit parts a while back and blew up all of them testing. For critical applications like this or a motor controller, I buy from the big supply houses like Mouser or DigiKey. You still have to watch them too, I've gotten the wrong parts a few times.
 
FETs do share a lot better than bipolars, but in fast switching (PWM) service they don't do all that well. Manufacturers app notes talk about only designing for about 50% improvement (or some number well below the full improvement) in capacity with the second FET in parallel due to imbalance. In the DC case it is likely better as the voltage threshold is not important, only the on resistance needs to match, if they are the same temperature, but if they are not matched the current will not balance well and I squared R magnifies the difference since R affects I. The same batch helps, but they still vary for some reason, or so that is the experience of folks who match them.
 
A 50% improvement for 2x more FETs sounds about right. I'm sure you can do better than that with careful design, but typical builds will be less than optimal.
 
I think i spoke too soon

I just tried putting the "bad" mosfets on a breadboard and testing them like that (instead of soldering them in circuit) and they worked fine!?

I put a nice size heat-sink on them as they where getting pretty hot.

IMG_20171122_093803-1036x1842.jpg

the orange wires were connecting all the gates (had 4 in parallel) and the black wire was for the source. the Drain was connected to the Heatsink. i made sure everything was insulted from each other before hand.

This there something wrong with that? this is the only difference from the breadboard testing.
 
It's a reverse Murphy's Law thing. If you solder them into a board nice and neat, it will blow for sure.

Just kidding... It might be random chance the one you picked actually meets the specifications.
 
The madness doesn't stop!!!!

So i took the last 2 remaining mosfets that worked on the breadboard and soldered them to the circuit.

I turned on the circuit and everything turned on but the output voltage was only at 30v. I checked the gate voltage and it was only 2.5v. Weird.
So I de-soldered the circuit and tested it with jumper wires. And magically it worked! The output voltage was 46v and the gate was 12v! I soldered all the connections besides the gate and connected the gate with just the jumpers and it worked as well. But once I soldered the gate it didn't work?! I connected the gate to the circuit by just clamping them together (they where definitely connected) and it also didn't work?!
I soldered another wire connecting the gate and the circuit and now it works!!?!
What in the world is going on?!!
Here is a small video:
https://photos.app.goo.gl/KbeM7M8bJaN5JRhd2
 
One possible idea came up:

The gates are super sensitive and fed by a very high resistance. Residual soldering flux between the gate lines and any other part can have enough resistance to throw things off. I've seen this cause problems before. Carefully clean things after soldering. Alcohol sort of works if you use a stiff brush (toothbrush). Otherwise they make flux remover solvent that really dissolves it fast, but it's kind of nasty. Use a small screwdriver or pointy object to scrape any thick deposits between traces.
 
i thought of that but i soldered another wire on to both ends and it still worked. sorry the last image didnt load properly. its the white loop wire that im talking about
 
Here's my design for controlling inrush when using the Zero power pack with an inverter that has 2500uF to charge. I read this entire thread and learned that a switch is required. I didn't like having to remember to switch off before connecting and possibly damaging the contactor if I didn't. Many design iterations later I came up with this using a momentary toggle switch. I think it is fool proof, time will tell, haven't built it yet but LTspice says it works. Pin 1 on the pack is always hot so M2 keeps the output off when first connected. A momentary enable signal on pin 3 closes the contactor. That starts the output voltage ramp after C1 delays M1 turn on giving C2 time to charge. C2 then discharges through R8 for a nice constant current output. The contactor latches the enable signal on until a momentary off turns off the output, opens the contactor and resets C1.

pack adaptor.png

This is my first post here. I'm a retired EE with an electric motorcycle and a soldering iron. Comments appreciated.

pack adaptor sim.png
 
I'd recommend 12v zener diodes from source to gate on M2 and on the bank of M3-M6. In theory they would not ever do anything but in practice you can get some capacitive coupling between parts and the gates are pretty easy to blow up. They're cheap insurance.

The FETs will be throwing off a bit of heat, so you will need some heat sink. You could probably find some 150v rated parts with lower Rds to reduce the heat, but those aren't too bad.

I ran a 600w inverter off my 52v bike battery, which works great, but it does make a nasty spark when first connected.
 
Yes, I tried 10pF from the B+ to M2 gate and spice shows a 45V spike at contactor close so zeners are a good idea. I could just use some small capacitors since fast switching is not required. Even 0.1uF doesn't affect anything.

I like the 200V fet's for the sake of more SOA, that can also cause surprise failures.

I'm not sure about heat yet. Spice predicts less than 0.5W per device at full load current. Of course the turn on pulse will make some heat but that varies with load. The inverter has a slow turn on (and should be off anyway) so it probably will only be the capacitor charge, about 4 joules per device. Vishay provides spice models for thermal capacitance, it might be interesting to know the transient junction temperature.

That pop sound on connection is what alerted me to the inrush problem. I was surprised that the inverter capacitance was that high. It is actually eight 220uF 250V capacitors. Inverters for the 90-120 voltage range are scarce. Mine is the Reliable Electric 2500, so far working well, under $300.
 
I wouldn't worry about the heat generated during the precharge since it happens so fast. Just worry about the steady state current.
0.5w per device is close to not needing a heat sink, but I would suggest at least some kind of sink, especially if it's inside some kind of enclosure. In real life, you probably won't be running the thing at 100% capacity for long periods of time, so should work fine.

Yes, that nasty little spark can erode your connectors pretty fast if you do it often enough.
 
I couldn't resist doing a thermal analysis since I have the model. With no heatsink the ramp charging the inverter capacitor creates a rise to about 35C that quickly falls and then slowly rises to around 44C with the 0.5W constant power.
thermal cap charge.PNG
If the inverter is on and driving a 1kW load, turn on power causes the junction temperature to reach about 115C with no heatsink.
thermal 1kw.PNG
 
For sure you wouldn't want to turn it on with a load. During the ramp up, the transistors would be dissipating a huge amount of power. Most inverters wouldn't let you do this anyway, so it should be pretty bulletproof.

44C isn't bad, but I'm sure this still assumes free air, not inside some enclosure.
 
I'm going to build it in a small aluminum enclosure with the TO-220's mounted insulated to it. Can't do free air because of the high voltage and I want it compact. I estimate the junction temperature will be a little higher, maybe 55C, but that should be fine. I'll post a picture when it's done. Here's what the parts look like, did a board layout just to see.
image.png
 
Back
Top