Derating MOSFETs

[salty9]

The thread "Greentime controllers" started me wondering about the wisdom of derating components, MOSFETs in particular, and rather than jumping into that discussion opened a new thread.

I would really appreciate feedback on derating and percentages for maximum life.

[bigmoose]

Search my user name and derating, I believe I have posted links to the open standards for derating FETs, capacitors, resistors for high reliability space applications. DeRating is how you enhance reliability. The high flyers have spent a lot of money determining how much to derate, and how much it buys you in reliability. There are standards with all the numbers filled in...


Edit: I had some time to find it:
If you want an easy to read derating guide for high reliability complex electronics, take a look at DERATING FOR EEE-INST-002:
http://snebulos.mit.edu/projects/crater/docs/engineering/EEE-parts-derating.pdf

Also:
Those that are serious about designing a more reliable widget should acquaint themselves with MIL Handbook 217. It is available for a free download here: http://www.quanterion.com/Publications/MIL-HDBK-217/index.asp

[neptronix]

The thread "Greentime controllers" started me wondering about the wisdom of derating components, MOSFETs in particular, and rather than jumping into that discussion opened a new thread.

I would really appreciate feedback on derating and percentages for maximum life.

Great question, but i doubt you're gonna get a proper scientific answer.
How about this general rule - don't run things at 100% or above their maximum absolute rating, ever?

Unless we are talking about motors, in that case they can go over their constant rating for certain periods of time. But electrical components are far less tolerant. maybe they'll tolerate it for a few hours, few weeks.. or maybe they'll just go bang instantly.

Every controller manufacturer worth their salt says that you should derate mosfets and capacitors just a little to add some headroom. That amount is usually 10% - 20%.

Did you know that the military typically derates things by 25%-50%?

[liveforphysics]

Before you can derate, you need to find the real capability. For example, a FET with 20mOhm of resistance, that you know needs an insulating thermal transfer pad, and the silicon die can't exceed 150degC, you might find your total heat dissipation budget for the part is something like 10watts, and your 10watts gets used up with just 22amps on the part, not counting the flyback diode heating and switching loss heating.

So, the part might say 5,000,000amp rating (or whatever BS they put on the datasheet), and the reality is, at 22amps you're fully maxed out with no margins. If you wanted a 70% margin on that, you're looking at 15.4amps for that device, and this isn't a full 70% margin because you have other heating contributors, this just covers RdsOn.


This is one of my favorite BS datasheets of all time. I love the 375amp continous rating for the part, that later down... it says something about 11amps, then in practice when mounted to a PCB they get so hot they melt the solder and float away at 9amps...

So, in this case, you would take the datasheet number of 375amps continuous, and derate it by 98% to get in the realistic usable ball park.

http://www.irf.com/product-info/datashe ... 9l2pbf.pdf

[texaspyro]

This is one of my favorite BS datasheets of all time.

Spec sheet BS? Say it ain't so, Luke. Say it ain't so...

[dnmun]

all the mosfets seem to be knockoffs, chinese manufactured, of IR parts. do not expect they will have the same device parameters as the real thing.

there is a guy on here now with blown mosfets in his C'lyte controller. they are chinese knockoffs of the irf1010e which has a 60V breakdown voltage and his mosfets blew up at the 56V or so level of his 48V A123 lifepo4 pack. making good mosfets is not like making bluejeans.

[Jeremy Harris]

When looking at derating it's worth remembering that the voltage that the FETs see can be greater (in all probability will be greater) than the battery voltage. Few controllers have decent commutation capacitors, and few seem to have paid any real attention to decent low inductance layout. You can be pretty sure there are going to be spikes on the supply close to the FETs that will be a fair bit over the supply voltage.

[dnmun]

agreed, but he had his bike static on the stand, not in motion or under throttle, just exposed to the full pack voltage when they blew up. so i assumed it was from the poor quality of the chinese mosfets, not able to withstand the breakdown voltage.

[Jeremy Harris]

agreed, but he had his bike static on the stand, not in motion or under throttle, just exposed to the full pack voltage when they blew up. so i assumed it was from the poor quality of the chinese mosfets, not able to withstand the breakdown voltage.

Possibly, but also possibly something arising from the turn on surge, especially if the power switch or connector "bounced" electrically at turn on (pretty strong probability of this, I believe, as I've yet to see a mechanical switch that doesn't bounce). It's another good argument in favour of an electronic power switch, with a slow turn on and off. It's hard to know for sure what happens on the rails of these thing every time they are powered on with a supply so close to the FET absolute max rating.

[dnmun]

you can tell him that, he said it was sitting on the bike stand powered up, long after the controller was turned on.

my point was that it is silly to derate these mosfets since they are not IR parts so the derating is from an unknown quality standard to begin with, but IR builds to a known standard they establish to warrant their parts and the chinese have no such legal obligation.

viewtopic.php?f=2&t=40818

[Arlo1]

Thanks for the links Dave. Its funny becsause one of the charts at the bottom shows 40% for comercial use as the max and I found 40-50% of the irfb4110 ratings is what I can get with my moded chineese controllers. So I have been close to what the chart sugests.

[zombiess]

Thanks for the links Dave. Its funny becsause one of the charts at the bottom shows 40% for comercial use as the max and I found 40-50% of the irfb4110 ratings is what I can get with my moded chineese controllers. So I have been close to what the chart sugests.

Many people blow up controllers because they run them at their absolute maximum ratings (me included at times). It's something I'm trying to get away from because I know it's a bad practice. Just because you can run your 12 FET IRFB4110 controller at 100V 100A doesn't mean it's the wisest thing to do. It may survive for a while, but a hot day, a voltage spike, a steep hill etc is all it takes to push it past it's breaking point.

My 18 FET controller with 4115s should be able to do 210A per bank of 3 FETs at 100C if it had none of the typical losses and went straight from the datasheet specs, but I know that's not realistic. That's why I've decided through temperature monitoring and experimenting to not exceed 105A battery amps / 130A phase amps and I normally only run it at 85A battery, 115A phase. Before anyone says it, yes I know the FETs only feel phase amps, I'm just listing both settings here since it's what most people are use to seeing on these controllers. With temperature monitoring I make sure I never exceed 80C either and I rarely ever see over 65C at 85A/115A unless climbing a steep hill at low throttle.

Even if my controller was built with IRFP4110 FETs and I kept the voltage around 82V, I still would not go more than about 10% past these settings because the reality is the TO-220 package is what causes the limit with it's small legs and poor thermal transfer characteristics.

It is possible to burst numbers higher than what I'm posting as many people have done, but if you want reliability you should size your controllers more appropriately and monitor them as well. Modding them to improve cooling / thermal transfer doesn't hurt either.

Someone here can correct me if I'm wrong or have observed something different, but I've noticed wheel size plays a big difference in what you can get away with when pushing the limits when using a hub motor. The smaller wheels allow you to accellerate more easily meaning you are demanding power from the controller for a shorter amount of time. Once up to speed and the amps fall off the controller has time to cool off again. You also need less phase amps to get you going.

[Arlo1]

Wheel size will effect the longevity of all components. The system will be on phase amp limits untill a higher speed which meens you will be having chopy current going though the controller to the motor more often and not to mention it is from a guess which the china controllers often get wrong and over shoot causing them to be stressed. See farfels dyno sheets from a sevcon vs china controller!

Now as for my 12 fet I managed to lay down 4.5 HP at the rear wheel with a HS35 and could have tried for more but Im not stupid I then road the bike and was able to see 100 battery amps at 82-84v I did not set the bike up at max voltage. I also then detuned the controller ~10-20 phase amps and a coupel battery amps (it was last summer I dont remember excact numbers) I then detuned it further with using the torque throttle feture in the CA!

Now this lets me know if I want to set up a 12 fet with the proper fetures and some active cooling I could run 4.5 hp on that bike for ever. But For my brother I tuned it down to ~ 2.5 hp.