Need some help designing an ideal diode circuit

[Lukas52]

Hey,

for a solar installation i need some really powerful diodes to combine a battery and if that goes out a power supply. Big diodes have big loses tho, so an ideal diode would be ideal (funny).

I haven't found anything of the shelf that can handle my loads tho. Battery is a 8S LifePo pack (~ 24V), the PSUs are also 24 Volt (22-28, can be set). Design load is about 200A max, peaks up to 250 for a second or two.
Since i have never designed a PCB before i thought this might be a good place to get some input

What i have done already:
Lots of google search, seems like making a ideal diodes isn't to difficult on the surface. I have picked the LTC4357 as my controller and the beautifully named IRL40SC228 as my mosfet of choice.
I went with this controller because it looks very simple to use in the datasheet.
The mosfet was chosen based on the following criteria:
 - n channel
 - 40 Volt Source Drain
 - 360 A Current
 - 20 Volt Source to Gate
 - low RDSon (relatively speaking)
 - available and in stock at LCSC (i kinda need to get my stuff from there since they will be making my pcbs and i'm trying to dodge shipping costs)

I have also started to design a PCB, but messed up a little, so that isn't ready yet.

I have attached the Datasheets and my Circuit plan.

Given that i am not entirely sure about what voltage will be applied to the gate pin, i am not sure if the 20 Volt Source to gate voltage is enough...

[Zero999]

The LTC4357 will apply between 10V and 15V to the MOSFET's gate.

The MOSFET will still need a decent heat sink, as it will disispate up to 24W continiously and nealy 38W peak.

[Lukas52]

Those 10 - 15 Volt are a delta correct?

Since the datasheet specs the absolute (as in relation to ground i would assume) Gate Voltage at Vin - 0.2 V to Vin + 10 V

If that is correct i would never have to worry about the Source/Gate Voltage since even 15 is well withing the +- 20 V spec.

Heatsinking is a good call. It will be mounted with a much Copper connected to it as humanly possible, and the copper planes will be connected directly with screws to my busbars, but i still think the heat transfer wont be good enough...
Will just sticking a heatsink on top of the mosfet even make a difference? Since its only touching the plastic package. I was hoping to get away with a tiny 40mm fan blowing directly down onto it but that is a bit of a stretch.

[Smokey]

https://www.infineon.com/dgdl/Infineon-IRL40SC228-DS-v01_02-EN.pdf?fileId=5546d462566bd0c701567ece08d03664

Junction-to-ambient is 62C/W.  That's a temp rise of 1488C at 24W if you don't have good heatsinking

[Psi]

One of the best things for good heatsinking is using thicker copper layers in the PCB.
Or you can use an aluminum PCB.

If you need you have it a one-sided PCB so you can thermally couple one side of the PCB to a metal heatsink.

Another thing that works well is paralleling mosfets. It can end up cheaper and have lower RDson (so less heat) when you put more fets in parallel. The ultra low Rds top-of-the-range mosfets tend to have a premium price. So if you can avoid them you save money you can spend on having more mosfets.

For example,
Your IRL40SC228 mosfet is 557A and 0.5mR and costs $4.62 in 1 off price.

TPH1R204PB,L1Q costs less, $1.55 but is 1.2mR and rated 150A.
If you used 3 in parallel it would cost $4.65 and be 0.4mR and work up to 450A or so
Having the heat generated in many separate locations really helps cooling a lot. Hot spots are the enemy.
It adds a little complexity in terms of PCB layout as you now have to check all fets will sharing the load equality. But that is usually not too much of a problem

On the other hand, if you don't care about heat because you will be adding a metal heatsink/fan to it then a single mosfet might be a simpler solution to get running quickly. 

But keep in mind, adding extra mosfet of the same type in parallel makes a huge difference over a single mosfet because you immediately have less heat to get rid of due to lower Rdson and on top of that the heat is not all in one spot.

If it's a one-off project and cost isn't much of a problem you could have 4 of those IRL40SC228 mosfet in parallel and then you only have to deal with 5W of heat instead of 20W and that's 5W over those 4 locations. so really its 4 locations that need to get rid of 1.25W Probably don't even need a heatsink now, just a PCB with thick copper.

[Lukas52]

I don't care much for the price, since i only need two, but have to make five.

I do want to use 2 Oz Copper on both the top and the bottom layer of my 2 layer PCB

Most things you see in the wild use parallel mosfets, so you might be on to something

Is there anything i would need to watch out for when using them in parallel? I remember hearing that mosfets need to be matched pretty closely, otherwise some will burn out quicker than others.

I also attached a picture of my prototype. Im not done placing all the vias, since i don't know how to make grids in EasyEDA...

My Layout could accommodate multiple mosfets, depending on whether or not i have to match things like track lengths to the gate.

[Psi]

You'll need a bigger PCB than that, not just to fit them on but to heatsink them and get rid of the 5W or whatever you end up having to dissipate. Unless you glue a finned heatsink to the other side of the PCB to compensate. etc

Add ~10 ohm resistors in series with each of the fet gates before you join them all together. It helps them all turn on at the same time.
It's probably a good idea to throw in a TVS diode from your combined gate point to source to protect the gates from +/- 20V or whatever your gate max is.

Since you're application is an ideal diode its just switching DC, its not like you are PWM'ing them.

You also want to be feeding DC in from one side and removing it from the other.
eg.
IN-----------------
        m1  m2  m3
        -----------------OUT

Or this is ok too, you have a large ground plane for In and Out to keep the resistance to each fet low
       IN
|||||||||||||
--------------
m1  m2  m3
--------------
||||||||||||||
      OUT


But NOT like this

IN-----------------
        m1  m2  m3
OUT---------------

This is bad because M1 will take more of the load due to all the extra copper resistance to get to M2 and all the way down to M3

[Zero999]

Those 10 - 15 Volt are a delta correct?

Since the datasheet specs the absolute (as in relation to ground i would assume) Gate Voltage at Vin - 0.2 V to Vin + 10 V

If that is correct i would never have to worry about the Source/Gate Voltage since even 15 is well withing the +- 20 V spec.

It's specified relaive to the power supply voltage.

External N-Channel Gate Drive
(V_GATE – V_IN)

Heatsinking is a good call. It will be mounted with a much Copper connected to it as humanly possible, and the copper planes will be connected directly with screws to my busbars, but i still think the heat transfer wont be good enough...
Will just sticking a heatsink on top of the mosfet even make a difference? Since its only touching the plastic package. I was hoping to get away with a tiny 40mm fan blowing directly down onto it but that is a bit of a stretch.

Assuming it's a one-of, I would solder the MOSFET tab to a piece of copper, screwed to a large heatsink. The PCB can be connected via short, flying leads.

[MrAl]

Hello,

There are also clamping hold-down methods for attaching semiconductor packages to heat sinks they are not always screwed down or soldered.
You can clamp the device down to a heat sink large enough to handle the power, and if you attach the PCB to the heatsink the leads of the device can be soldered to the PCB in a normal fashion.  The PCB and the heatsink becomes one integral construction.
That's the way i would do it when dealing with some 20 to 30 watts of heating because that's a hell of a lot of heating for a transistor.
You'll still have to be lucky you don't get any hot spots.  You should use adequate thermal paste or glue.
In some cases you can even use thermal glue to hold it down.

The clamps used for this kind of thing are often just spring steel, but you can improvise that part.  Of course you can't apply too much pressure either or the package will crack.  The spring steel allows for thermal expansion and contraction.

Just to note, cooling from the TOP of the plastic package is only good for low power devices that need the slightest of extra cooling.
Plastic has a lousy thermal resistance spec.

[PeteH]

Don't forget that ideal diodes have a fundamental regulated forward voltage drop (depends on the ideal diodes controller used).

So you get improvements in power loss by paralleling until you hit the forward regulation voltage limit (25mV * 360 to 55mV * 360) - thay sets the "minimum" achievable loss for the ideal diode network (9W - 20W).

There are newer ideal diode controllers with lower forward regulation values.

[Psi]

Also, thermal epoxy is good at permanently attaching heatsinks to to220/dpak's.
Not so good thermally since it's attaching to plastic, but every bit helps.

Give the top of the package a quick sand with like 320grit sandpaper and the same for the heatsink bottom.
After it cures it will be very well attached.

[Lukas52]

Right now there is nothing on the bottom side of this PCB, so putting a heatsink there is no problem, also those big holes are for screws that mount directly to 60x6mm copper bus bars.

If i use multiple mosfets i would probably need to put some passives on the bottom side as well, so a thermal pad is the best i can do in that scenario. Given the more spread out heat that might not be an issue tho.

Aluminum pcb would be an option too, but they don't seem to be available with 2 Oz layers.

My Current setup uses some enormous schottky diodes. At the rated 200 A they drop more than 100 Watts, so the limits of a ideal diode look very acceptable in comparison

[Lukas52]

I did some more parts searching and came to the following conclusion:

Going with 2 Mosfets in Parallel is my best bet because:

 - At that size you can still get packages that can be soldered with a iron (i don't have hot air, yet)
 - I go from 27 Watts in a single package to 18 Watts in two packages (so only 9 Watts each)

When using 4, i would only drop another 6 Watts cause they rise in RDSon faster than they fall in current, given the tiny packages they come in it's hard to justify that little bit extra

I will modify my PCB design to accommodate that and report back once I'm done. Maybe i can get it done with just an aluminum heat sink clamped onto a thermal pad on the backside.

[Lukas52]

I'm mostly done with the PCB (i ended up using 4 bigger fets anyways).

According to my calculations this should end up somewhere between 0,16 and 0,2 mOhm or 6-8 Watts @ 200 A, and therefore a little bit more at peak 250.
That should be manageable by clamping a heatsink to the bottom of the PCB i think.

One question i found myself asking while starting to place vias: Are they even necessary? I thought i need them for current handling so i can use both sides of the PCB. I know how to calculate the resistance and therefore amperage of a track, but i'm not sure how to do this on a copper plane.

[Psi]

I would just have all layers copper filled and use 0.3mm via's in a line around the outer area of the fets on 3 sides.

Your via's on the pin side look fine.

[Lukas52]

Final Version Attached.

If it actually works as intended i will post the gerber files so someone else could reproduce them, just in case there are more people out there with a need for a very big low voltage DC diode

Compared to my first design i did add some more mosfets, but also switched the Controllers VDD Pin to be powered from the OUT side, rather than the IN side, reason being that i was only thinking about the one for the battery while making the design which would always have battery voltage no matter what. The one for die PSUs however would not (since they turn off above a certain battery soc).

Big Thank you to everyone!

[Zero999]

Don't forget that ideal diodes have a fundamental regulated forward voltage drop (depends on the ideal diodes controller used).

So you get improvements in power loss by paralleling until you hit the forward regulation voltage limit (25mV * 360 to 55mV * 360) - thay sets the "minimum" achievable loss for the ideal diode network (9W - 20W).

There are newer ideal diode controllers with lower forward regulation values.

Thanks for pointing that out. It's only important when the voltage drop across the MOSFET due to it's on resistance is under 55mV. It wouldn't have made any difference with a single 0.55mR MOSFET, or with two, but it might do with four, as the voltage drop will be 25mV.