Fairchild IGBTs OK? 1

[Renovator1]

Basically just looking for feedback here on the quality level and how far said IGBTs deviate from the datasheet specs (ie - says it will handle 60A at Tc = 100C but lets out the magic smoke at 40A, etc...). I'm mainly interested in non-punch-through (NPT) or "field stop" IGBTs for easy paralleling and several Fairchild parts keep coming up in my searches at, eg, half the cost of IXYS/Infineon/IRF/etc and since I totally believe TANSTAAFL I'm wondering what gives. So, any feedback would be most welcome. (this will be my first use of IGBTs for something besides tinkering, btw).

-Jeff

 

[Renovator1]

electricuwe - you must have missed where I wrote this about the Fairchild IGBT under consideration:

Ic = 40A @ 100C (I never bother with the 25C rating)

The rest of your post assumes such a level of cluelessness on my part it is hard to interpret it as anything but rude. Suffice it to say, I did not choose a 600A module over a 40A single IGBT "just to get the loss down"; I choose the 600A module because it will need to switch 200-300A continuously with peaks approaching 600A.

***

itsmoked - the way I look at the price of the module is this: it offers the equivalent ratings of, say, 20 single IGBTs in one convenient package and someone else went to the trouble to make sure they are all thermally coupled and dynamically matched - that's worth a premium, in my opinion, but to get it for a discount is a no-brainer!

 

[electricuwe]

Sorry Renovator1, I didn't intend to offend you. But unfortunately there are more and more participants in this forum who lack a sound engineering background.

On your part you could do lot to avoid confusion, if you clearly state what your requirements are. Starting a discussion on 40 A devices and than later changing to 600 A modules without providing background information is quite misleading.

By the way, if you want to compare devices at 100°C case the Semikron Module is a 490 A device. Assuming this your intial design must have involved paralleling more than 10 Discrete devices, which is not an eays task. The pitfalls you can get by wrong arrangement or driving easily override differences in quality you will find at different manufacturers.

But basicly both approaches are viable. Benfetits of the discrete approach are lower device cost, better heatspreading and, if done properly, lower total stray inductance.

Advantages of the module are integrated isolation of the power circuit to the baseplate, avoiding the difficult external paralleling and the lower manufacturing cost for the inverter.

 

[sreid]

In power modules, the free wheeling dide is matched to the die size-speed. Same for IGBT descrete Co-Packs. I always buy the short circuit rated devices even thourh the losses are higher. With (almost) all modern FETs, the body diode is rated for use as the free wheeling diode. Regulate the gate voltage to 15VDC if you want your Short Circuit protection to work.

 

[Renovator1]

electricuwe - no problem, I appreciate your last response. I did not state the application up front but I did mention that "easy paralleling" was a must for the discretes. At any rate, the application is a series wound DC traction motor controller that does regen (a 2 quadrant drive, in other words). I know from experience that even getting more than a few mosfets to switch on and off at the same time (despite the positive Rds[on] tempco!) is a real trial and usually requires hand-matching for transconductance... Not something you want to have to do for anything that will be produced in volume! I expect the same issues with IGBTs except with the added bonus that most also have a negative tc for Vce[sat]. The exception here seems to be the NPT technology which trades off excellent paralleling (positive Vce[sat] tc) and low on drop for slower switching speed and lower voltage rating. Well, 600V is plenty for this application (192V is the practical maximum) and if I can switch 600A in 250nS that's plenty fast enough at 16kHz, so being able to do so in 77nS is just icing on the cake. Oh, and the module only costing $30 more than the equivalent amount of discretes (prior to derating for parallel operation) pretty much made it a no-brainer choice, I thought. Though, if I missed something that's what you folks are usually good for...

So, yeah, I was looking at somewhere between 15 and 20 of the 40A (@ 100C) Fairchild IGBTs in parallel for each side of the half bridge. I'm sure some day I'll find a use for the hundred I bought but for this app I think the Semikron module is the best choice.

Now, figuring out the most economical way to get rid of the waste heat without resorting to liquid cooling... I'm seriously considering embedding heat pipes into a standard aluminum extrusion heat sink... perhaps this should be a topic for another thread?

sreid - yeah, that's the nice thing about IGBTs compared to MOSFETs - IGBTs don't have an intrinsic (parasitic) freewheeling diode (FWD) so one has to be added to the package if needed. Thus, it is much easier to dope it properly for whatever optimizations are desired (low Vf, fast Tfr/Trr, etc.). The poor FWD in a MOSFET is pretty much along for the ride. That said, one not so good aspect of the Semikron module is that the Rth of the diode dice is higher than the Rth of the IGBT dice so the FWD ends up at a much higher below a certain duty cycle... The thermal modeling is driving me crazy as a result.

 

[electricuwe]

Regarding cooling:
For a single 62mm Module, as long a you do not operate it to close to it's limit, you should be able to find a suitable combination of heatsink and fan. However, the module you have choosen is the maximum rating available in that package, so this makes it a tough job.

Take care that the most critical part thermally may be the diode when driving at low speed (low duty cycle for the IGBT, high duty cycle for the diode)

On IGBT and Mosfet-Diodes:
Intrinsic Diodes of Mosfets are really good in the lower voltage classes, but still very poor for the higher voltage ratings. Even the Diode in Mosfets with fast diode is poor compared to a dedicated diode.

On the choice of device itself:
For long time there has not been significant development in Mosfets in the range from 100...300 V. Therefore a 600 V IGBT outperformed a 15 year old Mosfet-Chip design in many cases. But now semiconductor suppliers recognising a new market for such devices increase their development effort for Mosfets in that range. Check what is announced.

 

[Renovator1]

electricuwe - yep, the true Achille's Heel is the anti-parallel diode: it has a much higher Rth[j-c] than the IGBT die does. I suspect that, practically speaking, this will not be a big factor (i.e. - the motor will only draw maximum current when accelerating from 0 RPM and as the rotor RPM increases, BEMF will automatically decrease the current that can be drawn for a given supply voltage).

That said, it is still true that IGBTs achieve a higher current density for a given die size than MOSFETs, or at least that's my understanding of the current state of the art of both devices.

 

[sreid]

I like R-Theta Fab Fin Heatsinks with direct impingment Fan Cooling. Ball Bearing DC Fans can have MTBF's in the 100,000 Hour range.

http://www.r-theta.com/

 

[electricuwe]

To avoid oversizing the power electronics and to ensure smooth accelleration you should ramp up duty cycle slowly so that the speed is expected to follow or use a current loop. Even if you do not allow the current to exceed nominal value (which is done in most cases to get faster accelleration), the low speed region is critcal for the diode. And thermal time constants of the semiconductor chips are quite small compared to a mechanical system.

 

[Renovator1]

Thanks for the R-Theta link, sreid - their FabFin product is ideal for this application. Hope their prices are reasonable.. some of the heat sink quotes I've gotten so far have been nuts (it's hard for me to accept the heat sink costing more than the module it's supposed to cool...)

Anyway, I gave you a star for that one - thanks!

And you read my mind concerning the fan. Any particular manufacturer you prefer? I've had mostly good experiences with Comair/Rotron (formerly EG&G, wasn't it?).

electricuwe - You read my mind, too. Might be time for me to make a foil hat...

I'll be using local desat-detection to protect against abnormal events more or less instantly but an AVR microcontroller (optoisolated, of course) will keep an eye on the average current during motoring (buck) and regeneration (boost) modes and not allow them to exceed either pre-programmed absolute limits or user-programmed limits, whichever is lower.

One thing I'm not sure I can get away with but will be trying out soon: placing the gate driver/sensor board right on top of the IGBT module's bus bars. Besides the obvious benefits like minimizing the loop area between driver and gate, I'm going to see if mounting a couple of Allegro's linear Hall Effect Devices on the top side of the PC board directly centered over the relevant bus bars will allow me to monitor the current in them with reasonable accuracy... The app engineer at Allegro thinks it's an interesting idea but was somewhat skeptical about its accuracy/repeatability. Well, that's what prototypes are for

 

[sreid]

The Fan Manufacturer I go with for Industrial Applications is Mechatronics.

http://www.mechatronics.com/

Get it customized for you environment.

For general purpose applications (office, Lab), NMB.

http://www.nmbtech.com/cgi-bin/start.cgi/nmb/fanstart.html

And the Ultimate; Magnetic Bearing Fans.

http://www.xoxide.com/enermax-marathon-enlobal-fan-120mm.html