really thick copper

[Simon]

I need to control 60A of current on a PCB so can't really see any option but nice thick copper like 4-6 oz. I'll probably use 4 layer so that signals can get in and out more easily.

I am aware that thick copper creates problems for track width as at aome point you etch down this traces. But do i need to care ? thin traces are just for signals so if they end up thinner that does not matter, or are there manufacturing issues with an inconsistent thickness. Would the inner layers be less than 6oz and say just 1 oz ? What is the standar way of doing it ?

[matseng]

Another option is to use PCB Busbars - they are meant to transfer a lot of current between points on a PCB.....

[Simon]

I'm wondering how much the busbars cost and how much manufacturing compications putting something with so much thermal inertia on the board might cause. I can't attach a bus bar to a mosfet leg so still need decent copper like 4oz to get from the bar to the pins. As it is I am using more than one mosfet just to make the PCB's life easier.

Bus bars might be the way to go if they come with screw points for wires to screw to as that solves another canundrum.

[IconicPCB]

Problem is not variation in thickness.Problem is variation in width/undercut.

We have pressed and milled 1mm thick copper for multilayer pancake transformer design.

Post prototype work was done a by a chinese shop which was selected after some surveying of various performance capabilites.

[Aodhan145]

Why does it need to be on a pcb?

[Simon]

Why does it need to be on a pcb?

Good point, thus far I assume that pick and place is the fastest method of assembly and cheaper, if i take them off the board I'll have someone hand wiring 6 mosfets and more and having to screw it all down to heat sinks etc.

[matseng]

I'm wondering how much the busbars cost and how much manufacturing compications putting something with so much thermal inertia on the board might cause. I can't attach a bus bar to a mosfet leg so still need decent copper like 4oz to get from the bar to the pins. As it is I am using more than one mosfet just to make the PCB's life easier.

The busbars could start a millimeter from the pin of the fet. The power losses (and the corresponding heat) should be minimal compared to having a track many cm long - or?

The Corsair 1200W supply that can push out 100A continuously uses busbars.  If it's good enough for them....

http://media.bestofmicro.com/7/1/523837/gallery/in_modular_front_w_600.jpg

[Simon]

The thing is I also need a method of attaching my power input and output wires and soldering them to large planes is not going to be practical and again going to cause a power distribution problem. So it would make sense for any bus bar to automatically incorporate threaded holes possibly into a turret that can be used to screw ring terminals down to. I don't want to be in a position where I am using a bus bar and a separate screw PCB terminal. Basically I'm looking at is it now practical to get a piece of brass machined to the size and shape I want so that I have everything in one component feature. I think the answer is starting to be yes.

As you say I can run the bar past the MOSFET terminals and then use a short length of copper which would not generate that much heat and be connected to a nice cool bus bar.

[langwadt]

The thing is I also need a method of attaching my power input and output wires and soldering them to large planes is not going to be practical and again going to cause a power distribution problem. So it would make sense for any bus bar to automatically incorporate threaded holes possibly into a turret that can be used to screw ring terminals down to. I don't want to be in a position where I am using a bus bar and a separate screw PCB terminal. Basically I'm looking at is it now practical to get a piece of brass machined to the size and shape I want so that I have everything in one component feature. I think the answer is starting to be yes.

As you say I can run the bar past the MOSFET terminals and then use a short length of copper which would not generate that much heat and be connected to a nice cool bus bar.

lots of stuff like shield cans are made by etching sheets, might even be possible to reflow to the pcb

for connections something like press fit nuts should work

[daqq]

You can have multiple types of copper on a PCB - see: http://www.epectec.com/articles/heavy-copper-pcb-design.html ... though I don't think there is a more exotic process known to man

Also, I always wanted to try out having water cut pieces of brass/copper sheet metal cut out and have them pick and placed an reflowed onto a PCB... you might try that

[DenzilPenberthy]

I spoke to a lady from Wrekin Circuits at the Southern Electronics show the other week about heavy copper boards. They do up to 16oz if I remember correctly. She seemed very knowledgeable so maybe give them a shout? I think the main issue is that the acid doesn't etch 'straight down' so the tracks don't end up rectangular in cross section. The acid etches the side of the track under the photo-resist as well so the traces end up somewhat trapezoidal in shape but I think it can be accounted for to get a reasonably accurate CSA.

http://www.wrekin-circuits.co.uk/

[ovnr]

You know, 60 amps isn't that much? Run 2oz copper on both sides of the board so you effectively end up w/ 4 oz; if you can accept a 25 deg C rise, that ends up at a 12mm wide track. Scale it up a bit - I don't know how space-constrained your application is - and you should be good to go.


(As a point of reference, consider high-end video cards with GPUs eating 150-200A. They're simple (well, "simple") multi-layer boards with no extra fuss.)

[IconicPCB]

TGhey wouldnt be doing 16oz copper etching.

They might go for pattern plating on a thin base so that the etching process does not etch too heavy a layer.

[Simon]

I spoke to a lady from Wrekin Circuits at the Southern Electronics show the other week about heavy copper boards. They do up to 16oz if I remember correctly. She seemed very knowledgeable so maybe give them a shout? I think the main issue is that the acid doesn't etch 'straight down' so the tracks don't end up rectangular in cross section. The acid etches the side of the track under the photo-resist as well so the traces end up somewhat trapezoidal in shape but I think it can be accounted for to get a reasonably accurate CSA.

http://www.wrekin-circuits.co.uk/

I have spoken to spirit circuits before and they will also do anything, but I am trying to stay away from exotic. I don't really care if the signal tracks are trapezoidal or even triangular but I wonder if it causes a problem for processes.

[Simon]

You know, 60 amps isn't that much? Run 2oz copper on both sides of the board so you effectively end up w/ 4 oz; if you can accept a 25 deg C rise, that ends up at a 12mm wide track. Scale it up a bit - I don't know how space-constrained your application is - and you should be good to go.


(As a point of reference, consider high-end video cards with GPUs eating 150-200A. They're simple (well, "simple") multi-layer boards with no extra fuss.)

I have allowed for 40C rise and it's still at 25A on 4mm wide 6-8oz. Running traces in parallel is an option but you still need to get that power into the mosfet pins and vias also have limits. Also paralleled tracks will have limits because the calculator assumes that the heat is going through the board as well and not expecting it to hit another trace generating as much heat.

[ovnr]

I have spoken to spirit circuits before and they will also do anything, but I am trying to stay away from exotic. I don't really care if the signal tracks are trapezoidal or even triangular but I wonder if it causes a problem for processes.

The issue with thin traces and thick copper is that the tracks are trapezoidal, but the narrow end is toward the PCB. So if you go too far, the traces will simply not adhere well enough - or at all - to the board. If it was the other way around, it wouldn't be that much of a problem.

I have allowed for 40C rise and it's still at 25A on 4mm wide 6-8oz. Running traces in parallel is an option but you still need to get that power into the mosfet pins and vias also have limits. Also paralleled tracks will have limits because the calculator assumes that the heat is going through the board as well and not expecting it to hit another trace generating as much heat.

Parallel traces are no better than a single wide trace. Are you space-constrained in the width of your trace? If not, make it 20mm wide; it does not matter. The localized heating around the MOSFET pins is unlikely to cause an issue, as the copper plane acts as a heatsink, as does the pin.

[Simon]

Oh I see yes of course it woulk do that. Hm maybe stick with 4oz. I want to use multiple mosfets anyway and they are two 30A channels so just do 3 mosfets per channel. Yes i can use a really wide trace and then branch to each mosfet with as wide a trace as I can. might be interesting to see the resistance of the traces as with a 6mohm on resistance of a mosfet the track resistance could actually cause load balancing problms.

[forrestc]

I need to control 60A of current on a PCB so can't really see any option but nice thick copper like 4-6 oz. I'll probably use 4 layer so that signals can get in and out more easily.

I am aware that thick copper creates problems for track width as at aome point you etch down this traces. But do i need to care ? thin traces are just for signals so if they end up thinner that does not matter, or are there manufacturing issues with an inconsistent thickness. Would the inner layers be less than 6oz and say just 1 oz ? What is the standar way of doing it ?

I've seen a design for a high current board which was constructed of a multilayer board with multiple layers and via stitching throughout the length of the copper, with all of the vias either copper filled or at least heavily plated.   Very impressive.   You might want to consider this as an option.

If I was going to consider this, I'd just give up doing it myself.   I'd probably call a company which does this type of stuff, and know how it works.  I'm about 99% sure the design I mentioned above was done by divsys.com - who apparently does this type of stuff as part of their PCB fabrication service.  I can't vouch for them specifically but every time I run into them at a trade show I love to see what they've brought as an impressive PCB - often related to power.

[mongo]

One crazy idea,

A long time ago I was playing with electroplating vias at home, if you can electrically isolate the trace, and if your fab can keep it clear of solder mask it would be trivial to electroplate just that trace.

I tested it on a junk board, in which I set my Z-axis way too deep on my cnc PCB router.  It is a very easy to control process and I have added an ancient link to an old google+ post where I documented how you can get everything you need at the local hardware and drug store.



https://plus.google.com/114378443997641555027/posts/DVigHqanmnu

[ROBOT]

Check out figure 1.

http://www.epectec.com/articles/heavy-copper-pcb-design.html

My guess would be that they start out with 1oz for the low current portion then mask and plate up the High current traces.

[Simon]

Yes sounds like they plate sections up. Probably expensive. I'll probably try and stay with 4oz but will be interesting to see what the PCB manufacturer can do.

[uncle_bob]

Hi

There are a number of other odd things with "narrow" traces and very high currents. Copper thickness is never quite uniform ...

The normal way to do this is to go to wide(r) traces. 10 to 20 mm is a pretty typical target width. If your design is not able to tolerate that sort of geometry ... go for buss bars. Stitching a 10 layer board gets you into the through hole plating process. It's not bad with laser drilled holes but it can get nutty with deep mechanically drilled holes. You can buy a *lot* of copper bar stock for what a fancy board will cost you.

Bob

[nuno]

Is that for a BLDC motor controller, or 3 half-bridges?

[SL4P]

Where does the policeman come into this?   

[Simon]

No it's to switch brushed motors, I'm thinking 10-20A per switch device and be able to scale up or down.