Power plane thickness for low voltage high current supply.

I'm design a PCB with low voltage (1.0V) high current (up to 50A) power supply for a single chip. But I'm not sure about the thickness of copper for the power plane. How do I calculate this parameter? Is 1oz enough for it? I saw gigabyte has been using 2oz on its motherboards, but I guess the highest current on a motherboard is much higher than 50A?

50A ! Thats a lot of current for a PCB and an IC
How far is the chip from the supply, what temp range are you working to etc etc How many layers...
Using IPC-2152 current calculations with a 20 deg C temp rise, ambient 22 deg C, for 50 A 2oz Copper you require a track 40mm (that is over 1.5 inch wide).
I suspect somthing is not quite right with the current calculations, some more information is required, as apart from motor control boards or similar I have never had to cater for 50A... and never for a single chip.

you can go with more copper thickness say 8 or 12 oz for your current requirement

Thank you for the replies. Yes, it's 50 A and the estimation is reasonable. It's an FPGA with the static current higher than 10 A. I'm working with 12 layers, and the dc-dc is placed right aside the chip. A track over 1.5 inch wide is possible, acturally the current is on the whole plane (of coz there're holes such like vias). 12 oz sounds great for the current, but would be too much for the board. As many intel cpus consumes such high current, do they require very thick power planes on the motherboards?

You need to use a copper wieght that will also allow the manufacturer to etch the features required, vias etc, so even 2oz may be to thick to allow fine features to be etched.
Which Intel IC is it, I'd be curious to what PSU controller you are using, 50A still sound like an awful lot of current.

---------- Post added at 08:42 ---------- Previous post was at 08:27 ----------

Had a look around some of the designs we have, and at a few sites. The 50A figure I believe is a miscalculation or an instantaneous requirement, that would be supplied by tha capacitors on the board. I have also talked to a couple of friend who do boards and some engineers, and none of us have ever had to do a 50A supply for an FPGA, ever.
Is this figure based on worse case switching withh all gates switching at once.

I once implemented a power electronics design (150 A output current) with 6 oz (210 µm) outer layers, but it hasn't much to do with regular PCBs. Minimum feature width was about 700 µm, allowing standard SO-ICs as smallest package technology. There are many detail issues that can't be discussed in a brief. The design finally served it's purpose.

Most PCB manufacturers are suggesting a multi layer sandwich as an alternative, and I would definitely go for it in a new attempt. Besides the said structure size problem, switched mode high current designs are usually suffering from circuit inductances, and a sandwich is the best way to beat it.

Had a look at Nat Semi- workbench tool, highest current 1V circuit I could get from them was 30A, this takes up an area 35 x 35mm.
I have done boards with high current, but not 50A where you have a BGA device on the board. I still believe the 50A is not a constant demand, but peak levels. Designs I have done that are in 10-30A range not only required careful layout but also thermal engineering, and careful EMC engineering due to the high dI/dt switching currents. The copper thickness for current is to reduce resistance and thus heat generated, for short term pulses the caps can supply the instantaneous requirements, and thinner copper on multilayer boards will handle pulses of current during switching cycles without over heating (a 10mm trace will withstand 100A for 2 seconds)...
But the origional post seems to refer to a constant 50A, this requires clarification.

Standard house breaker is 15A. So it is impossible for a CPU to run a t 50A.
Intel Core 2 is rated 55watts.
FYI:
for every 35um copper thickness, you need 0.075mm trace/space.
So, for a 4oz copper board you will need 0.3mm trace/space.

marce said:

You need to use a copper wieght that will also allow the manufacturer to etch the features required, vias etc, so even 2oz may be to thick to allow fine features to be etched.
Which Intel IC is it, I'd be curious to what PSU controller you are using, 50A still sound like an awful lot of current.

---------- Post added at 08:42 ---------- Previous post was at 08:27 ----------

Had a look around some of the designs we have, and at a few sites. The 50A figure I believe is a miscalculation or an instantaneous requirement, that would be supplied by tha capacitors on the board. I have also talked to a couple of friend who do boards and some engineers, and none of us have ever had to do a 50A supply for an FPGA, ever.
Is this figure based on worse case switching withh all gates switching at once.

Click to expand...

Thank you for reminding me. I was using the FPGA provider's power estimation tool for the calculation based on the information from my FPGA guy. I'll double check with him to see if it is accurate enough.

---------- Post added at 18:10 ---------- Previous post was at 18:08 ----------

FvM said:

I once implemented a power electronics design (150 A output current) with 6 oz (210 µm) outer layers, but it hasn't much to do with regular PCBs. Minimum feature width was about 700 µm, allowing standard SO-ICs as smallest package technology. There are many detail issues that can't be discussed in a brief. The design finally served it's purpose.

Most PCB manufacturers are suggesting a multi layer sandwich as an alternative, and I would definitely go for it in a new attempt. Besides the said structure size problem, switched mode high current designs are usually suffering from circuit inductances, and a sandwich is the best way to beat it.

Click to expand...

What does sandwich mean? Providing a single voltage with two power planes?

---------- Post added at 18:51 ---------- Previous post was at 18:10 ----------

loosemoose said:

Standard house breaker is 15A. So it is impossible for a CPU to run a t 50A.
Intel Core 2 is rated 55watts.
FYI:
for every 35um copper thickness, you need 0.075mm trace/space.
So, for a 4oz copper board you will need 0.3mm trace/space.

Click to expand...

I think you found a green cpu :-D I found some evil ones, e.g. Core2 Extreme Processor QX9775, Vcc about 1V, maximum Icc 150A. AMD also has such evil cpus (Athlon 64 FX-74, Phenom II X4 965 Black Edition, ...) I'm quite curious how they made the motherboards.

What does sandwich mean? Providing a single voltage with two power planes?

Click to expand...

Two or more, depending on the current requirements. Powerplanes can be e.g. 100 µ (3 oz) each.

Standard house breaker is 15A. So it is impossible for a CPU to run a t 50A.
Intel Core 2 is rated 55watts.

Click to expand...

Present I7 core is 130W @ 1.3V = 100 A. Don't see how it's related to house circuit breakers.

FvM said:

Two or more, depending on the current requirements. Powerplanes can be e.g. 100 µ (3 oz) each.

Present I7 core is 130W @ 1.3V = 100 A. Don't see how it's related to house circuit breakers.

Click to expand...

If the output is 50A DC, what do you think the input is? The power comes from somewhere.
This doesn't even take into consideration the other +12V and +5V for other devices in your computer.

I do layout boards with 12V 100A power. Mainly for battery charger units. That roughly correlates to a 110VAC 30A input.

If the output is 50A DC, what do you think the input is? The power comes from somewhere.

Click to expand...

100 A @ 1.3V refers to maybe 140W DC input power, provided by 12V CPU power connector of a standard ATX power supply. Even a small 350 W ATX power supply can easily deliver it without overloading a wall socket.

I guess, you didn't notice, that the discussion is about low CPU voltage supply.

Recent PC power supplies are required to use PFC and don't consume much more input current than you can calculate from it's real power.