High Current Schottky Diodes?

[oatnet]

In the early days, I used an ebikes.ca product that joined two packs together, and prevented one pack from charging the other. I think it had an upper limit of 45a, Which was way more than people were running in 2007. IIRC, it used Schottky diodes to do this.

I just put a 60a 179v a123 pack into the VW Camper Bus conversion, to replace the 30a PSI pack, and it occured to me that it would be nice if I could combine the packs on longer drives, and pull 80% DOD from their combined 90ah. It would have to support a 250a load with 500a peaks, which is higher than I found on my eBay searches. I did find this 46a version, so I am wondering if I could (or should) put (12) of them in parallel for 540a:

C4D05120A, 46A 1200V, Silicon Carbide Schottky Diode, TO-220, 1.2kV
http://www.ebay.com/itm/370605792884

I see the heat sink on the diode above, so I wonder if what I am thinking is wise - would I waste a lot of current by converting it to heat?

Does anyone know Schottky that could handle the full load without paralleling?

-JD

[dnmun]

if both of your packs are the same voltage, and both are lifepo4, then you should be able to use them in parallel without the schottky diode. but if they are just close in voltage and not the same number of cells in series, then you only need for the diode to withstand the difference in voltage between the two packs, so maybe only a 5-8V diode if 2S difference. you will need the current handling capability for sure so it will get hot. i use .4V times the current.

[texaspyro]

Those diodes are a joke... 8 amp continuous... whoopee! Forward voltage drop is three to five times what a silicon diode is!

[Joe Perez]

"Big" diodes are not especially cheap, however they do exist.

http://www.newark.com/powerex/r7s02012x ... 20Products

http://www.mouser.com/ProductDetail/Inf ... i6Ap5N0%3d

[CamLight]

As texaspyro mentioned, those C4D05120A diodes would not be a good choice. You'd need at least 100 of them in parallel for 540A continuous and would have to get rid of about 1000W of heat. Using something like the MBR40250G at 20A each is a little better, but still results in a nasty pile of heat. At this high a voltage it's hard to find a schottky that has a low enough forward voltage drop to reduce all the heat the voltage drop creates.

Assuming 0.75V drop at 20A for each diode (from the datasheet, assuming 120C case temp), you get 15W of heat for each. At 540A you need 27 diodes (handling 20A each) which results in 405W of heat to eliminate. Not impossible, but you'll have a decent-sized, probably fan-cooled, heat sink setup to handle that heat from the 27 diodes.

A really great cooling setup could reduce the number of diodes needed, but you''re still going to have a big sink+fans. Assuming 250A continuous, with 500A peaks, would reduce the number of diodes too. But, it all depends on how long the 500A peaks are. More than a fraction of a second and you're essentially running the diodes continuously (as far has the heating is concerned).

[CamLight]

"Big" diodes are not especially cheap, however they do exist.

http://www.newark.com/powerex/r7s02012x ... 20Products

http://www.mouser.com/ProductDetail/Inf ... i6Ap5N0%3d

Those discs are quite hard to mount properly, but nice low thermal resistance though!

[heathyoung]

Have you considered mosfets as ideal diodes?

[CamLight]

Have you considered mosfets as ideal diodes?

Ooohh...where's the "Like" button?

[full-throttle]

Have you considered mosfets as ideal diodes?

Won't work, the the fets would just sort the two packs together

And paralleling diodes won't work either FYI

[rkosiorek]

Have you considered mosfets as ideal diodes?

Won't work, the the fets would just sort the two packs together

And paralleling diodes won't work either FYI

question: how do you suppose they parallel redundant power supplies in servers, medical devices, aircraft and other mission critical things?

answer: Mosfets in circuits called "Ideal Diodes" they really do work. forward voltage drop is 0V. so no power is wasted in the diode and you don't need a big heatsink, if you need one at all. a typical schottky passing 60A is going to be sweating more than 60W. the ideal diode less than 5W. so of course the damn things work. rule of thumb, really good schottkys use about 1W per amp. normal schottkys are closer to 1.5W

i do agree that parallel diodes will not work. it is tough to match the characteristics close enough so that one doesn't hog the power until it destroys itself. than one by one the rest of the diodes will commit similar suicide. though you could get away with paralleling the diodes built on the same die, like those in a dual diode package.

the usual issue of ideal diode designs is the voltage limits of the driver circuit for the Mosfet. most are designed to be used with standard 48V redundant supplies. so they likely won't be much good with those 100V+ systems.

at one time one of the board members called "Tiberius" was offering ideal diode boards based on the Linear Technologies LTC4357 chip. those have a voltage range of 12V to 80V, so you could use them on a 72V bike.

http://www.endless-sphere.com/forums/viewtopic.php?f=14&t=10471

but at the 179V you mentioned these won't work. you'd need a custom designed driver and a rack of big mosfets. but regular schottkys chewing up 200W+ of power would likely not be a good choice.

rick

[Joe Perez]

The trouble with MOSFETs is their RDS On rating.

Take this little guy: http://search.digikey.com/us/en/product ... ND/2126325

19 milliohms at 250A is 1.2 kilowatts of heat, and that's at 25°. RDS derates by 2x at 100°, giving you nearly 2.4Kw of heat.

[chilledoutuk]

call me stupid but whats wrong with one pack charging the other?
If you start at the same voltage then the pack that discharges quickest will most likely not mind being charged back up a bit.

[rkosiorek]

the big problem is not during normal operation. but if one pack is considerably weaker or has failed it would prematurely drain the other pack. diodes would stop that.

but it comes at a price. a schottky diode will drop about 1V to 1.5V at high current. which will also be a lot of wasted heat at high currents. Mosfets can be paralleled and this would reduce the effective Rds resistance. but that would mean a lot of expensive fets in parallel.

likely a more sensible solution would be to put in an A-B switch to switch between battery banks, like switching between fuel tanks, and a pair of coulomb counters/battery fuel gauges. to gauge the status of each battery bank.

rick

[oatnet]

Wow, rkosiorek, Joe Perez, full-throttle, CamLight, heathyoung, texaspyro thank you so much that is really helpful. Schottky idea dustbinned for converting too many watts to heat, and I'll look into Ideal Diodes. I did see mention of 'ideal diodes' while I was searching, now I have a better understanding of what they are, and see if I can find something to match the volts/amps I'll be using.

dnmum:if both of your packs are the same voltage, and both are lifepo4, then you should be able to use them in parallel...
chilledoutuk: but whats wrong with one pack charging the other...

dnmum, it also needs to be the same LiFePO4 chemistry, or the discharge curves don't line up. I've paralleled 10ah and 15ah packs of the same chemistry, and pulled 22ah from them.

chilled, the problem is when I lose a cell in pack A, then pack B dumps unlimited amps into pack A, until both are at the same voltage. The unlimited amps could be too much for the power wiring, and matching the voltage can severely deplete pack B, especially since I am most likely to lose a cell when the pack is mostly discharged. If I parallel at the cell level, I'll only lose that cell group in both packs, but when I parallel at the pack level, this problem arises.

I was also considering paralling some of these a123 28s3p modules, but it sounds like it would be better if I did it at the cell level...

-JD

[full-throttle]

@ rkosiorek: I know what the "ideal diodes" are http://www.ti.com/lit/an/slva139/slva139.pdf
I've used the P-ch circuit as reverse polarity protection

It won't work in this case since each pack will turn the fet on and subsequently short the two packs together

A circuit based on fets would work as long as there's logic in place to turn the highest pack on only - but that is not an "ideal diode" anymore

[rkosiorek]

@ rkosiorek: I know what the "ideal diodes" are http://www.ti.com/lit/an/slva139/slva139.pdf
I've used the P-ch circuit as reverse polarity protection

It won't work in this case since each pack will turn the fet on and subsequently short the two packs together

A circuit based on fets would work as long as there's logic in place to turn the highest pack on only - but that is not an "ideal diode" anymore

well "true ideal diodes" with a proper controller do not allow reverse current under any circumstances.

http://cds.linear.com/docs/Datasheet/4357fd.pdf

that TI appnote is not a full implementation of an ideal diode. it just a partial implementation that uses some features of an ideal diode in a charger. in a true ideal diode.

rick

[full-throttle]

well "true ideal diodes" with a proper controller do not allow reverse current under any circumstances.

In the event of a power supply failure, such as if the output of a fully loaded supply is suddenly shorted to ground, reverse current temporarily flows through the MOSFET that is on. This current is sourced from any load capacitance and from the other supplies. The LTC4357 quickly responds to this condition turning off the MOSFET in about 500ns, thus minimizing the disturbance to the output bus.

page 6 in your link

Otherwise it's essentially what I wrote

A circuit based on fets would work as long as there's logic in place to turn the highest pack on only

Look, I don't want to argue - it appears our definition of "ideal diode" is different. Others got to understand they are not really diodes so should not be treated as such.