Linear Mode of IGBT FSOA/hotspots? - feasable?

I'd feel much better with a MOSFET.

I don't think I've seen any IGBTs rated for any kind of DC aside from "off".

I don't know the coefficients or how you'd decide that an IGBT is going to be unstable; you're on your own here, in "test to destruction" land. It should be at least broadly similar to BJT behavior, so if you have a BJT of the same die area that's rated for such-and-such, then...

Thermal limits are by die area anyway, so you're wasting no more money on a 20A IGBT than a 10A MOSFET, or a 5A linear MOSFET.

Is this supposed to be turned on solid, after the fact, as an SSR? A real relay might start looking competitive, too...

Or maybe something a little stranger, like a bidirectional current limiting diode on the AC side, to be bypassed with a TRIAC. Or use a dumb resistor...

Tim

--
Seven Transistor Labs 
Electrical Engineering Consultation 
Website: http://seventransistorlabs.com 

"Klaus Kragelund"  wrote in message  
news:6e0312b6-5b9c-4a36-a5f3-f833fc2f95eb@googlegroups.com... 
> Hi 
> 
> For an application I am looking into using a large IGBT for inrush  
> (charge of DC link capacitor from mains) 
> 
> To reduce BoM costs, I would like to drive it in linear mode, monitoring  
> the current and adjusting the gate voltage accordingly 
> 
> There are however some talk about hotspots of a linear mode driven IGBT. 
> 
> The IGBT is a 20A device. My intention is to drive it at 10mA, so 2000  
> times less than the max current 
> 
> Does anyone know the workings of these hotspots, and if the issue is  
> real or just academic? 
> 
> Regards 
> 
> Klaus

It is generally not recommended. Some datasheets don't even state a forward SOA, some like this one do but then it's for single-shot:

IOW the only way to find out if it is legit for a product is probably to obtain the blessing from the manufacturer for brief linear operation. Most likely they will not give it. Why not use a FET? I've never had much use for IGBT because I could obtain much lower conduction losses with FETs and they are cheap.

--
Regards, Joerg 

http://www.analogconsultants.com/

I've asked the manufacturer about this only for reference, since our internal expert on IGBT's has provided details that this is a very risky path to take.

The reason for doing it was to use a part that was in the circuit in advance, so now we have to add either a PTC in parallel with the IGBT or a MOSFET/resistor to do the limited inrush

Cheers

Klaus

It's a four-layer device and I think that's what makes it dicey. One engineer (Jim T?) once said that using an IGBT in the linear range is like walking at the edge of a cliff.

Maybe you can replace the IGBT with a FET? If not I guess the added cost has to be swallowed in order to achieve a robust design.

How does a PTC in parallel work in this scenario?

--
Regards, Joerg 

http://www.analogconsultants.com/

I don't remember those exact words... but they're accurate.

I think maybe the solution is a combination. I'm sure the IGBT is there for control once the loop gets going. In-rush could be handled by a paralleled FET. (Likewise I don't see a PTC working.) ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142     Skype: skypeanalog  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.

Reminds me... On one of the GenRad portable supplies I simply used a relay and a resistor to bring up the caps until a POR released the relay and activated the loop. That's how I thwarted the wise-ass who killed my supply by rapidly toggling the ON/OFF switch >:-}

... ON from too short an OFF inserted a delay in the POR. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142     Skype: skypeanalog  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.

That's how I used to test ultrasound machines, rocking the circuit breaker back and forth. Initially they'd evenutally blue-screen. "But you can't do that!" ... "Can you tell the power grid in Romania that it shouldn't do that?" ... a few months later the system was nearly bullet-proof.

--
Regards, Joerg 

http://www.analogconsultants.com/

It is not "some talk". I was advised by IR, years ago, you must NEVER allow an IGBT to run in linear mode. You have to drive it hard on and hard off. When in linear mode, they have the typical positive temp coefficient, so that the hot spot hogs all the current, and goes into thermal runaway. As the transistor is really a parallel array of thousands of tiny transistor elements, they don't current share.

When hard saturated, they get into a region where the tempco is much less positive, and may actually turn just a bit negative, allowing much better current balancing across the device.

It is NOT academic, I've had the silicon die pass me like glowing bullets when they failed. Why don't you just use a FET? They are always negative tempco, and can work well in the linear range. The only advantage to an IGBT is the saturation voltage can be quite low for high voltage devices. Makes no sense to use an IGBT for linear operation.

Jon

The IGBT is part of the power module, used for inrush. But, we need slow co ntrolled inrush (by adding a PTC in parallel), to charge the DC link slowly in order to reduce peak ground currents via the Y capacitors to earth. The y would otherwise trip the ground fault relay

Cheers

Klaus

That is a common solution, but the relay does not like vibration

Cheers

Klaus

Use a FET or a TRIAC. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142     Skype: skypeanalog  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.

^ In the resistive region, of course!

All three (IGBT, MOS and BJT) exhibit NTC in linear/CC range, for the same general reason as far as I know (charge carrier concentration varying with temp).

Observing ratings, of course.

The historic claim of "freedom from 2nd breakdown" I think was only ever a matter of degree: ancient lateral MOS was simply so ludicrously space-inefficient that it would melt before reaching runaway.

Curiously, SuperJunction MOS usually meets full DC SOA. Not sure if that's to do with a different tempco, or what.

A little curious what the application/goal and driving constraints are... a very beefy MOSFET could be chosen here, which would have more than enough die area to handle the precharge power, and low enough Rds(on) to drop much less than the IGBT. A'course... such MOSFETs aren't cheap, either. :) Based on the IGBT (initial?) choice, I'm guessing price is the bigger motivator?

Tim

--
Seven Transistor Labs 
Electrical Engineering Consultation 
Website: http://seventransistorlabs.com

I have a portable NiMH battery charger that runs on unleaded gasoline. To control the inrush when the battery pack (274 V DC nominal) connects to the inverter, it uses a small ceramic-housed resistor, switched in and out by a relay at startup. It was built in late 2000, and I don't know the exact cycle count on it, but it's probably on the order of a few tens of thousands; no complaints from the relays, resistor, or inverter. As far as I know, the latest models do it the same way, and this design has sold about 2 million copies. On the other hand, it does have the luxury of a pretty high BOM cost, probably US$10k-$15k or more.

Matt Roberds

** That is best done with a suitable NTC thermistor. ** Is there a more expensive and idiotic way you might try instead ? ** This is not April 1, is it ?

Cos 10mA is not an inrush surge of any kind.

** I know your posts are not for real.

... Phil

Contrary to what others have said, I have successfully used IGBTs in linear mode. The device in question was an IXYS 32A high voltage NPT-IGBT running at a few mA current with a max CE voltage of up to 2kV.

No problems at all, running for several thousand hours without problems. So it is feasible, at least at those low currents. As a side note, I was surprised back then how well the SPICE model reflected reality also in linear mode. Obviiously, SPICE models don't tell you anything about SOA etc. and of course the mentioned concerns are for real.

Another example of IGBTs running in linear mode were IGBT audio amplifiers (don't ask me what that shall be good for, though). Although some of them gained a bad reputation due to failing IGBTs - there were also many which actually worked. Google for Alexander (spelling??) amp.

So in summary, you are in uncharted territory and you do it at your own risk, this is, you have to do a lot of tests to figure out the safety margins. But there are examples of working IGBT designs in linear mode.

Klaus

See Analog Devices AN-211. Alexander's amplifier used a pair of special 250V IGBTs that aren't available now. In particular Toshiba offered complementary IGBTs, gt20d101 and gt20d201.

The high Vgs voltages caused some extra trouble for Alexander. He used a separate higher voltage for the VAS and driver and a TL431 plus a thermally-coupled BJT, with gain for biasing.

--
 Thanks, 
    - Win

It's hard to know exactly what it'll do when you attempt a linear region, doing so means you need close loop to maintain it and yes, depending on the inrush time it can destroy it if you are pushing it to the max.

A simpler solution, pulse it when inrush is detected.

Jamie

Am 21.04.2015 um 16:53 schrieb Joerg:

I'm currently designing a constant current load and I found that BJTs are the better choice. It is easier to prevent oscillations and in linear mode, they can carry more current than even really big MOSFETs. All MOSFETs I considered allow only about 12V at 4A while a cheap TIP102 allows 20V. And yes, hot spots are real. I have seen transistors (both types) dying without obvious reason and without getting feelable hot. The process seems to be very fast.

If you are interested, a preview of this (hobby-) project is available at

(unfortunately, only in german for now) Under "Safe Operating Area" you can see a rare picture of the SOA in a linear scale and how dramatically it affects the allowable current at higher voltages.

Cheers

Robert

If you are having trouble with that consider a ferrite bead right in front of the gate. FETs are often faster which is why they can be dicey. The major upside of FETs is that they need very little drive power without having to resort to stacked arrangements such as Darlington.

they can carry more current than even really big MOSFETs.

That I doubt.

Old American saying: Then you need a bigger FET. For example this one:

Nice. When I semi-retire some day I want to learn how to create a better web site. Not so easy for an analog guy with next to nothing in programming experience.

Yep, got to watch that SOA and also not push it to the limits. John Larkin has some impressive photos of the results when one does push the limits.

--
Regards, Joerg 

http://www.analogconsultants.com/