Why would China mfgr grind #'s off of FETs?

[John in CR]

My motors came with stock 60V30A 15 FET Infineon controllers, but for some reason they filed or sanded the numbers off the face of the FETs. The only info I could get out of the motor factory was that the maximum off the charger voltage was 83V despite having 100V caps in the controllers. I've run one for over a year at 81V off the charger with a shunt modified to about 50A operation with no issues whatsoever.

Now that attempted mods have blown FETs on two of my controllers, I'm interested in putting one back to stock, but why the hell would they sand the numbers off of the FETs making correct replacement impossible to know? The same maker of the controller even put a 15 FET board in a case meant for an 18 FET board, apparently in some effort to cut corners, so why sand the FETs?

Any consensus on what the original FETs are likely to be? What's the best FET for our purposes with a mid 80's voltage limit?

John

 

[rhitee05]

Sometimes they do that to make reverse-engineering hard. Seems a little pointless, but admittedly effective.

At 83V max voltage you're probably looking at 100V FETs. Even if you could find say a 90V FET from somewhere (IRF doesn't make them), a little extra headroom is good to have. As per our previous discussion 4110s would be good or 4310s a close second. Those seem to be the best options for TO-220 package (the standard).

 

[SilverSurfer]

It may also be that they were miss printed at the factory, they removed the numbers and sold them as factory seconds. The mfg. bought them pre sanded for a discount price. Just a posibility. SS

 

[liveforphysics]

If you're willing to make the clamps to mount them, IXYS makes a sick 85v fet. I emailed you the datasheet about 2 weeks ago in my offer to do your 24fet controller with them.

http://ixdev.ixys.com/DataSheet/b45fb52d-f2a5-4237-a04b-a87d0203d941.pdf

Identical 25degC RDsOn typical of the irfb4110's.... yet, this package with roughly double the silicon to case back area in the package has some monster advantages in cooling.

The thermal characteristics for the IRF4110 is 0.904W/C going from silicon to heatsink, with a max Pd of 370w (which it could never see with that low of Rth of course)
The thermal characteristics for the IXTV230N085T 0.52W/C going from the silicon to heatsink, with a max Pd of 550w (which it could also never see, but would get a lot closer lol)

For something with a footprint on the board limiting you to a TO220 package footprint, this is the very best fet option I could find for you 2 weeks ago after searching through every mfg for hours of looking. It doesn't have the handy screw-hole in the top, so you would have to make your own clamping method. For the 24-fet, it is setup to easily make your own clamping method, for the side-mounted case controllers, you would likely need to make your own head spreader bar that is about 1/4" wider, so you can drill and tap the area a bit above the old fet mounting hole location to mount little cantilever style clamp for each fet. I was thinking just using some cut-up extruded aluminum window frame pieces or something similar when planning how to do the clamps for the 24-fet to mount this package type.

More of a PITA to build, but you would have about a ~60-70% improvement in your continuous power levels from the hugely improved thermal properties of this modified package.

 

[rhitee05]

Another slight mod that I think would improve thermal characteristics of multi-FET controllers.

The Rth of the isolation material is normally the main bottleneck for heat dissipation (kapton tape, silicone pad, etc). Rth of this material is related to area, the problem is the main thermal path on the back of the package is fairly small. Maybe 1/4"x1/4" (1/16 in^2) for a TO-220. A good interface material might have an Rth of 0.05-0.1C/W-in^2. That doesn't seem like much, but for a spot that small you're talking 0.8-1.6 C/W, which is as big or bigger than the junction-to-case and case-to-sink combined (~0.9 C/W). If you can spread out this area, the Rth will go way down.

I'd propose mounting each bank of FETs on it's own little heatspreader. Something like 1/8"x1" aluminum bar would probably work well, cut to whatever length necessary for each bank. The FETs attach directly to this bar with nothing but a tiny bit of thermal compound - NO electrical isolation. Then place your isolation material of choice on the back of the bar and attach to the main heatsink (obviously the bars must be separated from each other as well) with a little more compound. Aluminum is a very good thermal conductor, so the bar itself has negligible Rth and now you're looking at an area of probably ~ 1 in^2 for the isolation material, so the Rth there is now 0.05-0.1 C/W. Assuming a pretty large heatsink, say 1.5 C/W, the total thermal resistance just went from ~3.5 C/W to ~2.5 C/W.

As a side benefit, the aluminum bar ensures that all the FETs in each bank are very close to the same temperature. This helps make sure they share the current evenly.

 

[dnmun]

most likely were the 75fn75's with 75A 75V specs. and the caps come in standard voltages, going from 63V to 100V as the next step up.

they sand them off so their competitors in china cannot reverse engineer an identical controller, or BMS.

 

[CamLight]

For FET mounting, I highly recommend Aavid's MAXCLIPS (http://www.mouser.com/ProductDetail/Aavid-Thermalloy/MAX07G/?qs=sGAEpiMZZMttgyDkZ5WiuvoQ07e5IX8g%252bgq3%252bKVy%2fnc%3d)

They apply even pressure across the MOSFET which is critical for lowering the case-to-sink thermal resistance. And if used with TO-220's, they prevent lift-up of the one end of the MOSFET case if the tab mounting screw is overtightened (which significantly increases the thermal resistance).

Lastly, since the mounting screw for the clip is away from the hot case of the FET, there's less of a chance of thermal cycling loosening the screw. In fact, even if the screw loosens, there's still pressure on the FET. If you're using the tab mounting method for a TO-220 and the screw loosens even a little bit, that FET is doomed (even if you use split washers, which are totally, completely useless). Use properly tightened Belleville washers if using TO-220 tab mounting screws!

 

[liveforphysics]

Hey! Thanks for the link to the clips camlight that will save quite a lot of hassle for me if John decided he does want to go with those 85v fets. For his motor combo, I really can't think of a way to get a more powerful controller using that 24fet infinion platform.

Anyone see any objections with those fets (aside from the price)?

 

[rhitee05]

Personally, I'd be nervous about running 85V FETs in a 83v system. Not much headroom.

Not as important, but these FETs will be more susceptible to dv/dt issues than 4110s. The figure-of-merit is the Qgd/Qgs ratio. These are ~1, 4110s are ~0.3. May not be an issue, but this makes it more likely.

I'd be inclined to stick with 4110s despite the thermal advantage.

 

[John in CR]

Hey! Thanks for the link to the clips camlight that will save quite a lot of hassle for me if John decided he does want to go with those 85v fets. For his motor combo, I really can't think of a way to get a more powerful controller using that 24fet infinion platform.

Anyone see any objections with those fets (aside from the price)?

I thought it was a go on whatever you think is best on that controller. Sorry you were waiting on me. You should know that I always defer to your better judgment on this electronic stuff even if I don't always listen. I may have to go actively blown on the ventilation, and figure out how to go way bigger on the phase wires, but it sounds like a great controller for Colosus or a souped up ventilated multi-kw scooter hub.

 

[amberwolf]

I'd propose mounting each bank of FETs on it's own little heatspreader. Something like 1/8"x1" aluminum bar would probably work well, cut to whatever length necessary for each bank. The FETs attach directly to this bar with nothing but a tiny bit of thermal compound - NO electrical isolation. Then place your isolation material of choice on the back of the bar and attach to the main heatsink (obviously the bars must be separated from each other as well) with a little more compound. Aluminum is a very good thermal conductor, so the bar itself has negligible Rth and now you're looking at an area of probably ~ 1 in^2 for the isolation material, so the Rth there is now 0.05-0.1 C/W. Assuming a pretty large heatsink, say 1.5 C/W, the total thermal resistance just went from ~3.5 C/W to ~2.5 C/W.

FWIW, you could cut it down even more if you use separate heatsinks for each bank, and no mounting bar. Electrically isolate the heatsinks from each other, rather than the FETs, and let the FETs directly contact the actual externally-exposed heatsink, so there are then two (three, really) less thermal barriers (the mounting bar, the paste for the bar on the FET side, the electrical isolation between the bars and the main heatsink). Now there's only the FET-to-heatsink barrier, with whatever paste you use.

I don't know how to calculate the difference in Rth, but it should be significantly lower?

 

[liveforphysics]

If you wana get fancy, skip paste, hot plate solder them right to a copper heat spreader.

 

[rhitee05]

FWIW, you could cut it down even more if you use separate heatsinks for each bank, and no mounting bar.

Might make a small improvement, but if you use a heat spreader the Rth due to the insulator is almost negligible. And then you don't have to worry about getting hit with 100V if you touch your heatsink. You'd have to enclose your heatsinks inside a plastic box for safety, and then you'd have to blow air over them because now they're enclosed. Works fine, but it's nice if you can keep the heatsink isolated!

 

[rhitee05]

If you wana get fancy, skip paste, hot plate solder them right to a copper heat spreader.

Even better! If you want to make it double-duty, use the heat spreader as a bus-bar, since it'll be soldered directly to the drains.

 

[liveforphysics]

If you wana get fancy, skip paste, hot plate solder them right to a copper heat spreader.

Even better! If you want to make it double-duty, use the heat spreader as a bus-bar, since it'll be soldered directly to the drains.

Did you see the battery tester I made in this style?

It took getting hot enough until the solder melted to kill it, amazingly, I tested each FET, re-soldered it, and it worked fine! lol

 

[rhitee05]

 

[JEB]

Wow great looking thermal image Liveforphysics, where did you purchace the "camera", and about how much did it cost? Thanks JEB

 

[amberwolf]

Might make a small improvement, but if you use a heat spreader the Rth due to the insulator is almost negligible. And then you don't have to worry about getting hit with 100V if you touch your heatsink. You'd have to enclose your heatsinks inside a plastic box for safety, and then you'd have to blow air over them because now they're enclosed. Works fine, but it's nice if you can keep the heatsink isolated!

Pshaw...that's what the warning signs are for:


Besides, if I did it the way I wanted, to, it'd have an active fan cooling the heatsinks anyway, and I'd have it enclosed in a tube to pull the air past them; that'd also prevent touching them.

 

[liveforphysics]

Wow great looking thermal image Liveforphysics, where did you purchace the "camera", and about how much did it cost? Thanks JEB

Thanks. It belongs to Microsoft, it's a fluke Ti40, price is 5-digits.