Protection for MOSFETs ?
Heater.
Posts: 21,230
I have been playing with a 12v to 200v boost/flyback converter circuit. As below:
I built this up on a breadboard and in my fool hardy way first tested it without the feed back circuit R4, R5, R6, T1. Amazingly it worked first time!
Then the neon load fell off and poof! The output runs up to who knows what volts destroying the MOSFET and the diode and the 555 chip
OK, to be expected, lesson learned. Must fit the feedback circuit to limit the output voltage.
Whilst I'm rebuilding this I have some questions:
1) The article from which that circuit comes suggests it will not operate without a load. Any ideas why that might be?
2) It further states that a short or overload on the output will destroy the transistor as there is no protection against that. Any ideas how to provide such overload protection.
3) All the world seems to use gate resistors on the MOSFET which I do not have. This all seems to be due to reverse voltages on the gate being created by inductive ringing in the switching which can pop the FET. Could anyone elaborate on that?
4) Would things be more robust using an IGBT instead?
All in all, anyone got suggestions for making this circuit bullet/idiot proof?
As an aside, I'd like to be able to drive the FET from a micro-controller, say a Propeller. So that I can play with the frequency and pulse width programmatically. But that would require a gate drive circuit and given the ease with which smoke can be let out in such a circuit it would need galvanic isolation.
http://www.dos4ever.com/flyback/flyback.html#flyback2
I built this up on a breadboard and in my fool hardy way first tested it without the feed back circuit R4, R5, R6, T1. Amazingly it worked first time!
Then the neon load fell off and poof! The output runs up to who knows what volts destroying the MOSFET and the diode and the 555 chip
OK, to be expected, lesson learned. Must fit the feedback circuit to limit the output voltage.
Whilst I'm rebuilding this I have some questions:
1) The article from which that circuit comes suggests it will not operate without a load. Any ideas why that might be?
2) It further states that a short or overload on the output will destroy the transistor as there is no protection against that. Any ideas how to provide such overload protection.
3) All the world seems to use gate resistors on the MOSFET which I do not have. This all seems to be due to reverse voltages on the gate being created by inductive ringing in the switching which can pop the FET. Could anyone elaborate on that?
4) Would things be more robust using an IGBT instead?
All in all, anyone got suggestions for making this circuit bullet/idiot proof?
As an aside, I'd like to be able to drive the FET from a micro-controller, say a Propeller. So that I can play with the frequency and pulse width programmatically. But that would require a gate drive circuit and given the ease with which smoke can be let out in such a circuit it would need galvanic isolation.
http://www.dos4ever.com/flyback/flyback.html#flyback2
Comments
fault detection logic.
Gate resistors aren't needed for a single device - driving two or more MOSFETs in parallel does
require some damping on the gates to prevent a differential oscillation mode.
The really nasty problem with switch-mode circuits using ferrite cored inductors is over-current.
This causes the core to saturate magnetically, then the inductance drops like a stone and the
rate of rise of current rockets - if this happens you have to cut the gate drive in a few microseconds
or your MOSFET is vaporizing as its basically short-circuited.
The boost circuit you have there doesn't have a fixed ratio since you use a diode + MOSFET
rather than two active switches - hence the output voltage can keep rising till something blows.
Using two MOSFETs and synchronous rectification can give a much more predictable output
voltage characteristic, set by the PWM ratio (and lower losses as a MOSFET on voltage can be
a lot lower than a schottky diode forward drop).
Thanks, interesting.
I was keen on sticking with the 555 for this. For the fun and learning and because I like the idea that the 555 will be around for years to come. Besides I want to use this to drive some NIXIE tubes and perhaps a 12AX7 circuit. Shamefully this is the first time I have ever used a 555 chip despite having started using chips in the mid 1970's. That was with a TTL and NIXIE clock that used back to back mains transformers to get the NIXIE drive voltage.
OK, I'll forget about the gate resistor. Whatever it was I was reading about them did not really make it clear when they might be needed.
I was weary of the over-current. Just now what saves me from that is that the 12v is supplied by a wall-wart that simply won't deliver enough current. I was having fun simulating this set up with LTSpice. Without the feedback control. Sure enough the output can go soaring up over 1000V. "Oh yeah", I though , "that's what I want!". Despite having done that when it came to assemble the real circuit I could not convince myself it would work that well, hence the lack of feedback control and consequent magic smoke letting. I can't visualize this. Do you have a diagram of such a topology to link to?
layer between the collector and the sensitive gate. The difference between MOSFETs and IGBTs sort
of parallels the difference between triodes and tetrodes I think.
IGBTs though have a fairly large saturation voltage (1.5 to 3V typically), so for low voltage circuits they are
inefficient like darlingtons. Think 100V+ before IGBTs come into their own. They are available in 600V
and 1200V parts mainly, reflecting this property.
IGBTs don't have a body-diode built in, so circuits with IGBTs typically need a high current high voltage diode
as well (often they are packaged together as a unit though).
A preliminary look at the datasheet show they are good for 600v and have a fast switching time. That large saturation voltage may be a downer in this 12v circuit.
I have a bunch of big fat high voltage diodes here, where would I put them?
I may have to use those IGBTs when I have blown the last of the MOSFETS
Another difference between FETs and IGBTs is switching speed. FETs are just a lot faster. Looks like the ultimate voltage limit is similar though. (I.e. digikey has 4Kv IGBTs and 4.5Kv FETs)
To protect from saturation, add a current sense resistor between the FET and ground, then use a little BJT to pull down the reset pin. Like this inductance tester (source page)
Marty
I do like the sense resistor idea.
As as far as I can tell the circuit I'm using will change frequency with load anyway.
[ http://forums.parallax.com/showthread.php/156366-Wire-Wrapping?p=1277498&viewfull=1#post1277498 ]
Sadly not. Too expensive for me. Besides most of the components here are not wire wrap friendly.
Currently this is a bunch of components and spaghetti poked into a bread board.
Strangely in decades of forcing the magic smoke out of electronic components this is the first time I have ever used a bread board. Normally I solder things up on vero board or make a PCB. Tube circuits of course get point to point wiring on pieces of tag strip bolted to aluminium chassis.
Now, I know you like to be a good troll. A "wind up" as we used to say back in the day. I can appreciate that.
Whilst you are at it, do you have any helpful suggestions about this topic? Or are you just full of wind and giggles?
Anything specific?
Oh my.
I was fishing for someone with lots of experience of using MOSFETs to pop up with and offer some sagely advice.
I'm attracted to the idea of driving the PWM and monitoring the feedback programatically from a Propeller just so that I can experiment easily with different frequencies, pulse widths, inductors, MOSFETs etc. But that is going to require some opto-isolation to protect the poor Propeller and some further protection against software error causing runaway current consumption or over voltage.
Perhaps though, that is too much of a diversion from what I really want to do which is spark up some tubes !
If this thing ever works my next boost converter has to go up to 1000 volts. I have a big old CRT from a 1940's radar that needs lighting up.
Isn't the feedback circuit itself enough?
2) It further states that a short or overload on the output will destroy the transistor as there is no protection against that. Any ideas how to provide such overload protection.
If the NPN switching that FET never turns on then the FET will be in a constant state of conduction where the only resistance in the drain is the inductor's DC resistance. When the FET is getting pulsed then the current is limited by the inductor reactance (vs. time, an ideal inductor has infinite ohms at the start, going down to 0Ω.)
The FET should be mostly off. For this (primitive) circuit, stay away from supplies that can deliver lots of current that way such that if the FET was in "DC mode", the main supply will load down (no harm, no foul.)
Use a 7555 (or other CMOS version).
Hope your CRT from the 40's hasn't gone down to air (DTA) in all this time.
the gate drive due to abnormal conditions - a current-shunt is normally required (unless using
an internal current-sensing FET). Similar techniques allow cycle-by-cycle current limiting and
high efficiency low-duty-cycle mode switching - read a few datasheets is probably the best way
to see what's done.
Re: 1)
Yes, I would have though the feedback circuit was enough to cater for the no load situation. So I still don't get what the author was driving at.
Re: 2)
I still don't get it. If I had a low impedance 12v supply and a dead short on the output surely I end up with only a diode drop voltage at the top of the MOSFET.
Ah...thinks...in that situation the diode blows first due to over current. If it fails "open" then the top of the MOSFET now flies up to thousands of volts due to inductive kick back, that blows the MOSFET and probably everything else !
With a high impedance 12v supply voltages everywhere drop to almost zero and all is OK.
My supply is a carefully chosen (yeah right) wall wart that simply won't deliver that much current. So I might be safe with regard a shorted output.
I have no idea of the state of that old CRT. Similarly aged tubes have been know to work so I'm still hoping for some luck there.
Cheers.
It looks to me like you're trying to accomplish something similar.
Yes, thanks. That's heading in the right direction. Changing the single inductor to a transformer has the great advantage that the MOSFET does not have to switch huge voltages. I have been scrounging suitable ferrites to experiment with at some point.
which a fault condition is assumed and the drive disabled.
Protection is not the same as the normal feedback loop normally, the protection may have to react
fast, whereas the control loop is typically adjusting the input into a relaxation oscillator as here. If the output is short circuit the inductor will saturate then the MOSFET may blow next time is
on as the inductor is now just a piece of wire to the supply. The diode is likely to blow too
of course. A inductor with a core will catastrophically saturate when sufficiently overloaded -
as it starts to saturate the inductance drops dramatically, so the rate of increase of current
accelerates, so it saturates faster and faster until its just an air-cored inductor in effect, with
< 1% of the former inductance if that. Basically a piece of copper wire!
I do take your point about protection vs control loop. It's the difference between a servo motor controller and the end of travel cut out switches.
Seems to me that in that circuit powered from a puny supply if the output is shorted the 12v gets dragged down and the MOSFET cannot be driven again. Not
that I would want to bet on it.
The saturation thing is a bit of a mystery to me. I have read about it a bit. Even remember it from uni days. So from a simple conceptual point of view I understand it. However I have never seen it. It's time to borrow a scope from work and do some experiments. The article I linked to has a nice circuit for observing saturation on a scope.
I'd feel happier using my old tube based Tektronix scope for these kind of experiments. Unlikely to blow the inputs on that. Sadly I don't have it anymore.
Just now I'm short of a transistor to complete the feed back loop.
Is there any merit in the good old crowbar technique here? Stick an SCR across the input, have it triggered from current overload, thus immediately shutting everything down and blowing a fuse on the input supply.
Have you finished your design?
Would you please share the final schematic?
What is the maximum current available at 200V?
Having blown up all the semiconductors on my original 200v converter, MOSFET, diode, 555, etc, I took some time off from that idea and then started to take a totally different tack.
I got fascinated by the idea of the "SEPIC Multiplied Boost Converter". http://www.analog.com/static/imported-files/application_notes/AN-1126.pdf
Fascinated because I love a circuit like this that I have never seen before.
As a practical matter it offers the promise of requiring lower voltage MOSFETS and diodes. Which in turn means being able to use cheaper and faster devices. Fast is good for efficiency. Cheap is, well, just good.
The down side is the need for two inductors.
At the same time I decided to play with the MC34063 switching regulator chip instead of the humble 555. The 34063 is cheap and readily available and comes in an DIL8 package.
So, where did I get to with this? The circuit I came up with simulates in LTSpice nicely. 12 volts in and 200 volts out at 100ma. More than enough for the Nixie tubes I want to drive.
As it happens I built that circuit up on a breadboard at the weekend but have yet to power it up.
I have no idea how well that simulation will match reality when driven by a 34063. And the whole "build it on a breadboard" thing is very dodgy.
Perhaps next weekend I will work up the courage to apply power to it.
Welcome to the forum by the way.
This is not exactly Parallax related stuff. Apart from that fact that at the end of the day the logic driving my Nixies will be a Propeller. Perhaps even the Prop could take the place of the switcher chip.
Thank you.
Would you please post the ***.asc file of the above schematic in LTspice (as ****.txt)
It would just save me a few minutes. Thanks
OK, here we go...
Interesting. Here is another switcher app note by OnSemi (sorry no link but you can search for SMPSRM. Its Rev D)
AppNote Switch Mode PS Ref Man OnSemi SMPSRM-D.pdf
BTW Your link to the app note has an extra http// in the link. Nice app note.
In 1970 I built an inverter for my ham rig (old taxi valve rig modified) for 6V to 400V - I had a 6V VW!
BTW this packs a punch if you get caught across the secondary - at least I lived to tell the tale (400V DC 100mA). Remember, its the mils that kill
Anyway, used a pair of 2N4033 for the switcher forming an oscillator. Later I would have used 2N4036/2n4037 ??? as they fail open whereas the 2N4033 usually fail short circuit.
But now, I would just use a switcher IC. I have a need to run a 30W LED spotlight (32-36V constant current) from 12V which is why I am looking at switchers. Then I have to build a switcher (buck) to convert my solar panel from 30V to 12-14V using MPPT to charge my boat batteries.
The 34063 is a good cheap choice, better than a 555.
34063 is nothing stellar in efficiency, but it has a FreqSet CAP, missing on many modern parts.
It is well suited to transformer operation and means you could also experiment with Mains transformers run at ~ 400Hz to get high voltage in a quick and easy way, if size is not critical.
Gate resistors can also be used to lower RFI, by slowing the FET edges and that also reduces general ringing too.
The typical values can sort-of act like a fuse, but with 200+ V on a Drain, if the FET really wants to fail D-G short, a gate resistor is unlikely to 'save' a connected part.
Some look down on the 34063, what with it being ancient. But, hey, I'm just catching up with this switcher technology idea.
I was hoping too stay away from transformers. An experiment with such a set up for a tube supply a few years back did not work very well. Besides I was using a transformer to drive Nixie tubes back in the 1970's!
I guess I'll have to move to a transformer for my next iteration, a 1000v converter.
Size is not such an issue, that SEPIC converter I put together starts to get pretty big.
Cheers for the comments re: gate resistors. RFI? I don't care about no RFI