MOSFET detection of conduction - shoot through protection

[TrickyNekro]

Hallo ladies and gents,

I´m here yet with another question. I want to build an H-bridge. A heavy duty one. I want to drive a motor with a DC stall current of almost 100A.
For the reason I want to implement any fail safe mechanism possible in my design, one of them being shoot though prevention.

I want to detect when a MOSFET on one side is conducting, so that I prevent the other from turning on. I though that the best way to do that,
is to have an ultrafast comparator which detects when Drain and Source are shorted. I also uploaded a schematic of what I have in mind, so that
I can make things easier.

IMPORTANT NOTE: Please bare in mind that the schematic is only a proof of concept and is rather crude in nature. It works in the sim and right now
I only want that it works in the sim. I am only testing the proof of concept here.

Is the concept correct or should I use another method? What do the experts say?

Thank you in advance for your time and effort,
Best Regards,
Lefteris, Greece

[Jon86]

If you use a proper H-Bridge driver you shouldn't have shoot-through anyway?

[TrickyNekro]

Hmmm, kind of yeah but the driver doesn´t know if a mosfet is dead though and is shorting anyway.

Have you ever imagined what kind of fireworks you can have with a system designed for 100A failing?
Funny ones, I assure you :-P

I just need to know if a solution like this would be viable. Most specifically if I am going to have problems with the
comparator due to the inductive nature of the motor. Most probably no, but if there are some extra precautions,
I should take, I really need to know them :-P

I am more afraid of false triggering of the comparator. I need to make the system stable ;-)
I can use Schmitt trigger for this, I know. Any other precautions, I should take though?
Has anyone else tried something like that? What I should be really careful of?

That kind of things :-)

[eneuro]

I want to detect when a MOSFET on one side is conducting, so that I prevent the other from turning on.

I have some idea, using fast optocoupler, and maybe even it could be possible to integrate it into my optoisolated gate driver, so mosfet conducting on one side of the H bridge could simply disable another optoisolated driver on other side without microprocessor intervention in mosfets drivers circuit   

It is possible to disable this prototype driver and output low (force to switch off driven mosfets) by pulling down VE in 6N137 if needed.
It is interesting if it could work, but I need redesign my optoisolated gate driver and test it, while this was also only a concept (revision v0000) of using fast opto-isolator 6N137 with TC426 to drive a few power mosfets on one side in H bridge or standalone heavy load.

BTW: In your concept circuit above you never get 100A @ 15V (<15A) with this current limiting resistor 1 ohm in series with IRFZ44N (RDSon @ 25*C 0.017 ohm) 

[T3sl4co1l]

What is a short?

A short is a low voltage even under high current.

So,

You just need a resistance detector.  Which works for any voltage and current, including zero.

Umm, okay.  Good luck with that...

Suppose we apply some bias.  A lot of bias.  A hundred amperes.  That should make a pretty strong signal, eh?

But isn't that dangerous?  Heck no -- the magic figure is around 10 microseconds.  An overload condition (full gate voltage with the drain at full rated voltage) dissipates a lot of power, which heats up the silicon right quick, but heat is a slow thing, and there's relatively a lot of time available to detect something going wrong.  So don't worry about actual shoot-through too much.  If it's happening continuously, your driver is shit.  If it happens once in a while, well... maybe you've got RFI problems or something, but it's not going to explode.  If it happens for more than 10us continuous, yeah, you've probably got a short circuit somewhere, which should be dealt with.

So how should this be detected?  Use a desat detector.  This works better with IGBTs, but is applicable to MOSFETs all the same.  I've designed and built MOSFET inverters in the 5kW, 500kHz range, that detect a "desat" condition within two microseconds.  The difference between IGBTs and MOSFETs is mainly that you set the threshold higher for MOSFETs, at least for the high voltage devices.

IGBTs exhibit a constant current characteristic at much lower voltages than MOSFETs do, so they normally saturate at a low voltage (even the 4kV devices drop hardly 5V), whereas MOSFETs might drop 50V in normal operation (but >100V in fault, so it's not hard to tell or anything).  At low voltages, you're looking at, like, under a volt for normal operation, but a substantial fraction of VCC under fault, so it's still easily done.

Tim

[diyaudio]

Hallo ladies and gents,

I´m here yet with another question. I want to build an H-bridge. A heavy duty one. I want to drive a motor with a DC stall current of almost 100A.
For the reason I want to implement any fail safe mechanism possible in my design, one of them being shoot though prevention.

I want to detect when a MOSFET on one side is conducting, so that I prevent the other from turning on. I though that the best way to do that,
is to have an ultrafast comparator which detects when Drain and Source are shorted. I also uploaded a schematic of what I have in mind, so that
I can make things easier.

IMPORTANT NOTE: Please bare in mind that the schematic is only a proof of concept and is rather crude in nature. It works in the sim and right now
I only want that it works in the sim. I am only testing the proof of concept here.

Is the concept correct or should I use another method? What do the experts say?

Thank you in advance for your time and effort,
Best Regards,
Lefteris, Greece

In simulation with your proof of concept you are not considering a realistic impedance model of the motor, switches, control circuit at those current levels 100 amps+ .

H-bridge drivers already prevents pulse overlap and compensate with some dead-time either fixed or programmable,  the "error" amplifier design you have, will need many attempts at testing to get the tuning of the poles/zeros compensation to make it work reliably.

You can take the whole design and simulate it as a array of 4 ideal switches, complex RLC load and a modelled RLC parasitics around switches at those currnet levels, then the simulation will start to resemble real-world conditions.


I have learned some serious motor control tricks here all for free.

Teaching Old Motors New Tricks - Part 1

Teaching Old Motors New Tricks - Part 2

Teaching Old Motors New Tricks - Part 3


   
 

[eneuro]

Use a desat detector.

Do you mean something like this using HCPL316J?
However, this is rather low conductivity detection while IGBT is ON (should be but it is not saturated so something is wrong)?

Just thinking about it in my optoisolated gate driver, but I'd like to add fast optocoupler to this resistor and Ddesat diode, so it could be Fault signal to other H bridge side or microcontroler, but when it conducts.

[T3sl4co1l]

Yes, that looks like a good chip / example.

[diyaudio]

Yes, that looks like a good chip / example.

+1 second that.

[TrickyNekro]

Yes, that looks like a good chip / example.

+1 but... I´m entering TI European Analog Design contest.... So, I have to stay with TI parts as much as I can.
I know, I should have told that earlier, but I kinda want to keep my mouth shut as much as possible, until the contest is finished 
I will post though my results, but only after the dead lines, for obvious reasons...

So, I have to make the same thing but with whatever TI has available :-P
I will come back to you when I have a better example and have something working. I´m getting into it.

Probably, I will use a comparator as a schmitt trigger and then an isolator for sending on/off signals between the two sides.
But I really have a reason, to stay in this kind of design.

Thank you for all your replies, I will come back to you, when I have something more substantial.
But please, if you have something else to suggest, I am more than open to listen.

Best Regards, Lefteris
Greece

[T3sl4co1l]

Ah... well then, TI makes plenty of things, I'd be surprised if you had to do it discrete (well... TI doesn't make 2N3904s either, do they?  Hmmm...), though I don't think they really have anything in the way of gate drivers or isolators, not like Micrel/Microchip/Fairchild/etc. do.

A quick look at their gate driver product line, I see high/low side drivers up to 100V (maybe more, but not shown in the short description?), nothing claiming isolation.  The usual low-side drivers at reasonable ratings (2-7A peak output, single/dual/etc...).

So... yeah.  Oh well.  A good learning subject though, in any case.

Tim

[TrickyNekro]

They do have isolators, google iso7220 for example.

Actually, I gave a good hard look at the schematics above, the one eneuro posted. I searched around and I really think it fits to what I have in mind,
using, of course, other ICs and stuff like that, but I got the idea of how this works. But I ran into a small problem.

I´m trying to find a proper diode for this. Best bet is schottky by far (I guess...). So far I found bat30, although I am not really sure if it´s fast enough.
I need a diode that recovers in under 1ns, has a reverse voltage of 30V (to be safe) and can handle at least 30mA.
I´m searching around right now, but if you have something in mind, well yeah, that would be cool 

Best Regards, Lefteris
Greece

[T3sl4co1l]

You mean the desat diode?  1N4148.  Not worth using a schottky (too leaky, breakdown voltage isn't enough for transients).

Tim

[TrickyNekro]

Looking at mouser, for example, I saw BAS140W with 0.1ns recovery time. Is that any good of an example, for that kind of work?

I already chose a 4.5ns comparator for the job, and with taking into account the input capacitances, that 4.5 is about 5 - 5.5 with
the resistors at the inputs, worst case scenario, I guess. And the MOSFET is expected to start full contacting in 70 - 80ns.
The time is comparable and really good enough for the moment. This diode is only the last part before PCB building and testing.
(Ooops, forgot that I have to order the parts... that wouldn´t take long though :-P, especially when Uni is paying  )

Anyhow, thanks guys (and maybe, who knows, ladies) so far.
My best regards from Greece,
Lefteris

[TrickyNekro]

You mean the desat diode?  1N4148.  Not worth using a schottky (too leaky, breakdown voltage isn't enough for transients).

Tim

They are simply not fast enough. Sure I am getting a little bit paranoiac here, but the MOSFET would be conducting long before
it signals the other side of this happening.
It´s for a contest after all, I have to make it pretty 

[T3sl4co1l]

Nanoseconds?  Where did nanoseconds come from?

I think you'll find you'll have a hard time making that work at all.  If nothing else, gate voltage goes up before drain voltage falls, so it always starts in desat.  For maybe 20ns even for rather fast drivers on fast transistors.  More if it's slower or hard switching.

The natural thing to do is add an RC delay, so it doesn't respond to fast events.  Usually a microsecond or several.  In which case diode recovery and comparator speed matter none at all.

Tim

[eneuro]

Actually, I gave a good hard look at the schematics above, the one eneuro posted.

But if you go into details then it looks like this 2 channel TC4426 has different delay times for td1 (20ns) & td2 (40ns), but datasheet says typical delay time is 40ns.
I've also designed this concept for isolated gates powered by supply using toroidal SMPS in full bridge topology, so now I want to try simply use this gate power source 15V to power resistor, optoisolator diode and 1N4148 in series to mosfets drain, while optoisolator diode has typical only 5V reverse voltage and might be too slow, so we need fast diode of course, but typical optoisolator diode current needed is 10mA in case of 6N137, so maybe 1N4148 might work with 100V reverse voltage.
I will test it soon, too 

[tszaboo]

Look at all this thread. Lets say, your mosfet is shorted, damaged. You still dont want to die, that is a great though. You know what is the proper way to prevent this? You put a FUSE there. Not a crappy AC glass, one, a proper DC rated UL recognized fuse. No additional protection is required, no current sensing, comparator or whatever. If your mosfet is dead, you need a soldering iron, replacing a fuse should not be an issue. And if it takes 2 seconds to blow, so what? Your circuit is dead anyway.
You will need the fuse ANYWAY if you want a safe circuit. Why do people want to invent something which is not needed?

[T3sl4co1l]

"Safe" and "destroyed" are two different things.

Thanks in part to such circuits as are under discussion, I have blown very few transistors in the course of my experiments.  It's also a labor saving feature, having to replace just one overloaded transistor instead of the while pile, all of them top bottom side to side in parallel, because one decided to make things difficult for the rest.

Fuse, absolutely, whether it's battery or line operated: that should go without saying.  But remember, the fuse only serves one purpose: to protect the wiring from a fire hazard.  A fuse will not protect against dead transistors, not against arc flash if applicable.  (Even in industrial equipment with superfast "semiconductor" fuses, there is plenty of current and time (1-10ms) available to generate arc flash where the former transistors used to be.)

Tim

[TrickyNekro]

Nanoseconds?  Where did nanoseconds come from?

An IRFZ44N with a 22? drive from a 0 - 10V source needs about 86ns until full conduction at 15A drain current. There are MOSFETs much faster than
the IRFZ44N. Then again, the slower you drive the transistor, the better this circuit will work, but what I want to say is that these times can be comparable.

I am not really building the Desat circuit. I modified it, a lit bit, to my needs. Instead of biasing the diode from the gate driver, I am just applying a
constant voltage. Much more practical, cause you can both detect fault conditions and whether or not the transistor is conducting. The fault detection will
be implemented by a microcontroller, so I don´t have to worry about any other circuitry.

All in all, I have a solution that is probable to work, the only thing remaining is to test it in reality.

Best Regards,
Lefteris, Greece

[TrickyNekro]

Look at all this thread. Lets say, your mosfet is shorted, damaged. You still dont want to die, that is a great though. You know what is the proper way to prevent this? You put a FUSE there. Not a crappy AC glass, one, a proper DC rated UL recognized fuse. No additional protection is required, no current sensing, comparator or whatever. If your mosfet is dead, you need a soldering iron, replacing a fuse should not be an issue. And if it takes 2 seconds to blow, so what? Your circuit is dead anyway.
You will need the fuse ANYWAY if you want a safe circuit. Why do people want to invent something which is not needed?

If I don´t take into regard the offensive tone of your comment, for everyone posting here and the zero offer to make this world more knowledgeable (f*** the world,
for the people in this forum at least), then I should ask you what´s the point of an iPhone when you only need to make phone calls. And there are many more
examples, I assure you. Furthermore, it doesn´t mean that if someone finds no use in something, some other guy/gal won´t.
Trying new things is part of experimentation, without experimentation there is no progress. (And before you say something about that, just take a look in the past
2 centuries to see how many inventions were made "by accident").

So, yeah, please post something useful next time. I might seem new on this forum, but I assure you I spent almost 10 years in forums and electronics. I am just
testing something new to me and I´m looking for opinions...

Best Regards,
Lefteris

[TrickyNekro]

So... I got it working, in the sim at least but still after the sim, materialization always follows.
In the materialization, I´ll use different components, still the philosophy of the circuit remains the same.


Thanks for the help, guys/gals.
Best Regards, Lefteris
Greece