Bolt-on copper on standard PCB

[ArdWar]

I'm wondering with this kind of bolt-on copper design. Can I just simply open up soldermask on a standard PCB construction and literally bolt on milled copper busbar on top of ENIG? Are there any important considerations that I should be aware of?
Do I need plating for the busbar? I'd guess there would be potentially corrosion/oxidation concern especially with the PCB-busbar interface not completely airtight.
I'm not looking for volume production here, so design for manufacturability is less of a priority. Beyond not wanting unnecessary bother for myself that is.

I was looking at embedded cooper and/or extra heavy plating before, but the setup cost alone is not exactly worth paying for couple of fast changing one off prototype.

[T3sl4co1l]

Yes, but make sure the bolted connections are adequate.  It may be worthwhile for example to use solder or press-in terminals, and bolt to that.

PCB itself isn't great, because it cold-flows under pressure, and fasteners can come loose.  Hardware to maintain tension is suggested -- NOT lockwashers*, much more force and compliance range is required -- conical (Belleville) washers are the best choice.

*Split washers in particular, have very little spring force (it's a single turn of a coil spring, like who cares), and once collapsed, provide no ratcheting action either.  In fact they're basically as good as flat washers -- which is to say, more susceptible to unthreading under cyclic shear than without washers at all; washers tend to lubricate a joint, making it easier to come undone.  Star washers help, but not as much as you'd think.  There are special ratcheting washers (NORD-LOCK, etc.) that do actually improve cyclic shear resistance.  Otherwise, a plain dry bolted joint with tight fitting parts and no washers tends to be the best route.

Now, both shear and creep should be pretty small here, in a prototype situation, but keep in mind the effect of vibration and creep when moving the design into production.

Might also be able to order (or hand form) pieces with thru pins, for example waterjet a forked end then CNC bender or pliers or clamp and hammer the ends over so they can fit into holes or slots, then solder in.

Soldering at all, of course, is its own adventure... preheating to nearly the melting point is likely required, a chunky solder iron, and localized hot air might not hurt too.  Lap joints are also possible, just keep in mind the thickness of metal and its effect on the board: FR-4 and Cu have different shrink rate so it'll warp a bit on cooling.  Be sure you have enough clamping force to flatten it for thermal interface purposes, and avoid leadless components that might be overstressed or solder improperly as a result.  Or perhaps, consider a design with all the power stuff on one board, and riser/stacker boards for drive and control.

Tim

[ajb]

It may be worthwhile for example to use solder or press-in terminals, and bolt to that.

I would second this suggestion, since it keeps the mass of the bus bar out of the soldering process and sidesteps concerns with material creep, and the mating holes can be designed to allow for arbitrary assembly tolerances quite easily.  There are threaded SMT terminals ready for PNP assembly (example), which makes assembly easy and can keep the bottom of the PCB flat (for example if you wanted to mount the board to a heatsink like the photo in the OP).  You also get the option to space the bus bars up off of the PCB, and use the space underneath for components -- or get really crazy, and have two (or more!) layers of bus bars that can cross over/under each other. 

Do I need plating for the busbar? I'd guess there would be potentially corrosion/oxidation concern especially with the PCB-busbar interface not completely airtight.

You can can make any connection gastight if you tighten it hard enough  

If you were to directly bolt the bus bar to the PCB, you'd probably benefit from controlling where the bus bar contacts the PCB, so that it's limited to the area directly clamped by the fastener.  You could do this with a washer between the PCB and the bus bar or with a boss machined into the bus bar.  Either way, that will keep the contact pressure between the copper and the PCB higher, although it does mean you have to worry about the PCB creeping again. 

I don't think I've really seen useful numbers for how much a PCB will actually creep under a bolted connection, though.  It's something that often comes up in these discussions as a concern with directly-bolted connections, but without some idea of the magnitude of the problem it's hard to know how difficult it might be to engineer around. 

[T3sl4co1l]

Oh, I forgot to go back and complete my thought. Edited.

Not actually sure how much creepage FR-4 exhibits; it doesn't seem to be a normal documented parameter, and there's oodles of academic papers on this-and-that customized application for composites, but, I just want to know what bounds to design to, okay?..!

This,
https://journals.sagepub.com/doi/10.1177/0021998319845045
seems to suggest it's fractional percent, which feels about right.  Granted, I have no idea what resin they used, in relation to, whether it's representative of what PCBs use.  The metal-laminate structure is probably not a big factor in that, but of course it will be stiffer and stronger than nearly-metal-free PCB material.

This might also be of interest,
https://nvlpubs.nist.gov/nistpubs/jres/045/2/v45.n02.a06.pdf
you could presumably load up the board with a lot of metal -- use 4+ layers, at least heavyish copper (>2oz), stitching vias in the screw-mount pad, etc. -- and gain some stiffness that way, but keep in mind copper itself is far from the strongest metal out there.  Now, it's not dead-soft in PCB form, the purity is good but does contain some oxygen I think??, and there is some strain imposed from the plating process.  So it will be a bit stronger than annealed Cu, but under fastener loading and given the small fraction of metal in the clamped joint, probably the copper will creep as well, would be my guess.

Put another way: you can give it a better chance by maximizing metal in the joint, but it probably still creeps.

Solder creeps some as well, not to forget.  Lead-free alloys do creep less; the downside is, that strain can be transformed into whiskering, depending on alloy, treatment and time.

Tim

[David Hess]

Soldering at all, of course, is its own adventure... preheating to nearly the melting point is likely required, a chunky solder iron, and localized hot air might not hurt too.

I have done things like that with this soldering iron, which is still temperature controlled:

https://www.amazon.com/Weller-W100PG-Farenheit-Soldering-Degree/dp/B002I7X7ZS/

Drill small holes along the length of the copper to provide access to add solder.  Maybe put a machine screw through every other hole to hold the copper flat?

[ArdWar]

It may be worthwhile for example to use solder or press-in terminals, and bolt to that.

Makes sense to use terminals to add some spacing. No more worries with moisture entrapment, probably lessens creep concern, and I guess I can just use standard busbar grease to protect the joints. Although I initially thinking with direct contact I can get it to work as extra heat spreader too.

Hardware to maintain tension is suggested -- NOT lockwashers*, much more force and compliance range is required -- conical (Belleville) washers are the best choice.

*Split washers in particular, have very little spring force (it's a single turn of a coil spring, like who cares), and once collapsed, provide no ratcheting action either.

I agree, Belleville is a staple here. Not a fan of split washer either since even if it works it gives uneven contact pressure. Maybe fine with metals and composite, but definitely not with plastics.

[law]

I remember seeing this converter - it was pretty cool.
In this TI post the designer talks about that exact approach not being entirely ready for a commercial product as there was no interface material:
https://e2e.ti.com/support/tools/simulation-hardware-system-design-tools-group/sim-hw-system-design/f/simulation-hardware-system-design-tools-forum/802484/tida-01168-details-on-top-side-interface-material
Annoyingly, the design they go on to reference has no further information about it!

Quite a few moons ago I noodled with a high current motor controller design that used a laminated bus bar approach.
The aim was to take laser (or waterjet) cut flat copper bus bar and bend small fingers that would press or solder into PCB.
I designed it so that I could pass longer copper fingers through holes in the plate above, which would then go into the PCB.
Its probably easier to just look at the pictures.
The actual bus bar idea worked pretty well, but was fiddly. I had to make jigs to bend things nicely.
Copper also may not have been the best material as it was pretty soft.
I ended up shelving the project. The unit was small enough that just using eg, 6 layer PCB with 8oz copper pour was much cheaper and easier to build, and the clamping action from the 6 bolts did not provide an even clamping force. This led to the unit bending....

[Whales]

Woah that looks impressive.  I wonder if at some point you can skip the PCB and put the SMD parts directly on the copper busbars, "macro PCB".

There are threaded SMT terminals ready for PNP assembly (example),

I recently used some of these in a project: https://www.lcsc.com/product-detail/SMD-round-nut_Sinhoo-SMTSO3060MTJ_C2916383.html

They're ferrous, not sure if that's a dealbreaker for conductivity.  EDIT: Just cut one open and it's solid brass -- perhaps just the nickel plating was magnetic. 

The spec'd hole size has some slop in it, so keep in mind your tolerances when designing the holes & misalignment clearances of your busbars.

Paste + hotplate reflow wasn't enough to solder them on its own, but briefly touching them with a soldering iron whilst the boards were on the hotplate worked magic.

[uer166]

I remember seeing this converter - it was pretty cool.
In this TI post the designer talks about that exact approach not being entirely ready for a commercial product as there was no interface material:
https://e2e.ti.com/support/tools/simulation-hardware-system-design-tools-group/sim-hw-system-design/f/simulation-hardware-system-design-tools-forum/802484/tida-01168-details-on-top-side-interface-material
Annoyingly, the design they go on to reference has no further information about it!

Quite a few moons ago I noodled with a high current motor controller design that used a laminated bus bar approach.
The aim was to take laser (or waterjet) cut flat copper bus bar and bend small fingers that would press or solder into PCB.
I designed it so that I could pass longer copper fingers through holes in the plate above, which would then go into the PCB.
Its probably easier to just look at the pictures.
The actual bus bar idea worked pretty well, but was fiddly. I had to make jigs to bend things nicely.
Copper also may not have been the best material as it was pretty soft.
I ended up shelving the project. The unit was small enough that just using eg, 6 layer PCB with 8oz copper pour was much cheaper and easier to build, and the clamping action from the 6 bolts did not provide an even clamping force. This led to the unit bending....

This is not a bad way of doing it if you can do it as a stamped part (probably needs to be thinner copper). Reminds me of a Tesla driver inverter and how the FETs and DC link solder into the main PCBA.

[ajb]

This,
https://journals.sagepub.com/doi/10.1177/0021998319845045
seems to suggest it's fractional percent, which feels about right.  Granted, I have no idea what resin they used, in relation to, whether it's representative of what PCBs use. 

I only have time for a quick skim atm, but it sounds like that was testing for in-plane movement, versus the normal clamping forces we'd be more concerned about with a PCB clamped by a bolt or similar.  That's been the case for basically all of the academic work I've found the last few times I've looked into this as well.  I don't know enough about composites to know how much of a difference that really makes, but intuitively it seems like compression across the layers would cause less movement than the in-plane tensile loading -- after all, the material is fabricated under normal compression, so presumably the fibers are already well packed in that direction anyway.  That specific test also mentions that the test coupons were cut at 45° to the fiber axes, which is the worst case scenario in terms of the fibers being able to resist elongation of the material (you can see this in all kinds of woven materials, including your denim jeans -- try stretching the material at 45° to the weave, and it will stretch a lot more easily than if you pull along the warp or weft).  So, yeah, my intuition is also that a bolt through a PCB is probably not going to experience any more creepage than a belleville washer can handle.

I recently used some of these in a project: https://www.lcsc.com/product-detail/SMD-round-nut_Sinhoo-SMTSO3060MTJ_C2916383.html

They're ferrous, not sure if that's a dealbreaker for conductivity.  EDIT: Just cut one open and it's solid brass -- perhaps just the nickel plating was magnetic. 

Yeah, the same basic parts (or basic shape anyway) are sold as both electrical terminals and mechanical standoffs, and often the mechanical ones are steel.  But with a cross section in the ~tens of mm² versus a length of a few mm, the resistance from the bulk material is basically in the nano Ohms no matter what metal you use. 

This is not a bad way of doing it if you can do it as a stamped part (probably needs to be thinner copper). Reminds me of a Tesla driver inverter and how the FETs and DC link solder into the main PCBA.

Hard to tell from the pictures, kinda looks like they have tabs on the copper bus bars that mate into PCB-mount quick connect receptacles?  That's certainly not a bad way to go if you can fabricate the geometry and have some way of applying the required insertion force evenly.  The assembly in the picture seems like you'd need to chuck the whole thing in a press with suitable support fixturing to get the parts together.

[langwadt]

https://www.dmtl.co.uk/images/galleryImage/cropped/20171023165516.png

[Whales]

https://www.dmtl.co.uk/images/galleryImage/cropped/20171023165516.png

SW as in Whitworth?

I must buy these now.  Many hijinks are to be had.

[langwadt]

https://www.dmtl.co.uk/images/galleryImage/cropped/20171023165516.png

SW as in Whitworth?

I must buy these now.  Many hijinks are to be had.

nope standard metric, https://www.mac8japan.com/uploads/entry_meta/file_value/2945/en_2022a_sw.pdf

[law]

Hard to tell from the pictures, kinda looks like they have tabs on the copper bus bars that mate into PCB-mount quick connect receptacles?  That's certainly not a bad way to go if you can fabricate the geometry and have some way of applying the required insertion force evenly.  The assembly in the picture seems like you'd need to chuck the whole thing in a press with suitable support fixturing to get the parts together.

Thats right. There are small tabs on the flat bus bar. I bent the tabs using a 3d printed jig. The jig was two parts to make sure the bend was in the correct place and the right radius.
It was then assembled in stages. You can see in the 6th picture that the bottom plate was installed, then soldered to the FETs. A Kapton separator was installed, then the top plate installed.
The top plate should be soldered to the FETs, but wasn't so I could disassemble. Then the top capacitor PCB was pressed onto the tabs of the bus bar. I used a small jig with long fingers and a normal press.
The aim was that the copper tabs would deform the through holes on the PCB, much like those Wurth style M5 (etc) press-in mounting lugs.
I found that the copper was a bit soft and the tabs tended to deform a bit. Being so soft, I'm also not so sure how well the tabs bit into the through holes. Some further analysis would be needed.

In general I think the idea had legs, it just needed more work. If I was to do it again, probably would use surface mount FETs and work toward eliminating all soldering to the bus bar. That might mean tabs going down into a FET board or something.
Though, these days it's probably cheaper just to get huge copper embedded into the PCB directly by the PCB manufacturer....

The laminated bus bar approached is used a lot in high power applications, but not typically with the fingers etc. If you have a search around the laminated bus bar area, you might find some suggestions on what fasteners they use.

For small volume or one-offs, other folks just go bananas and solder copper bar straight to the board. I've also done this on a prototype - Paste on the board, copper on the paste, then straight in the toaster oven.
An example, not mine:
https://build-its-inprogress.blogspot.com/2016/08/introducing-gigatroller.html

[ajb]

washers tend to lubricate a joint, making it easier to come undone

...and also making it easier to achieve proper tensile preload on the screw!  Or at least *more consistently*, anyway, which is especially important with smaller fasteners where the thresholds of 'tight enough' and 'there goes the head' are both low and close together    A proper preload also helps defend against creep, as well as thermal cycling, by allowing the elastic deformation of the screw to take up those dimensional changes without losing enough tension to allow the head to turn.  Longer screws are also better in that respect, since there's more length to stretch and the total elongation is higher.  That makes belleville washers especially valuable on shorter screws, like the ones we typically use to fasten PCBs, where the screw is too short to stretch that much.   

related PSA for anyone doing any kind of production:
https://www.wihatools.com/search?options%5Bprefix%5D=last&options%5Bunavailable_products%5D=show&q=torquefix&sort_by=relevance&type=product
https://products.wera.de/en/torque_tools.html
(et al, not to mention power options for higher volumes)

We do a lot of M3/4-40 up to M5/1/4-20 (curse of being in the US) and I would never be without a set of drivers like the above.  For washers, we use serrated belleville washers all over the place.  The ones we get from McMaster are designed to work with adjacent metric and UN screw sizes, so we can cover our needs with only a handful of PNs.  I don't know how much of a difference the serrations make in practice, but we've never seen one come loose 

Reminds me of a Tesla driver inverter and how the FETs and DC link solder into the main PCBA.

I was curious about that inverter and went looking, found this video: https://youtu.be/argrHjADn8g (has good english subtitles).  There's a good shot at 3:11 showing exactly what connects to the PCB, and it's only the S&D terminals of the FETs, which just look like extra-long DIP pins.  The entire 3-phase bridge is made out of stamped copper welded to the tabs of the FETs, because why not at that point?  Don't need chunky copper-billet-to-board connections when there's no board! 

[Psi]

Consider the extra cost of getting bus bars made to size and the bolts to hold them on and the labour to install them.
Then compare that to just getting your PCB fab to put thick copper on the PCB.
Maybe crazy thick inner layer to spread the heat out, or if you need external copper then maybe one side of the pcb can be super thick. Just be aware the thicker the copper is the wider your track width and clearances have to be.

Some fabs can also plate thick copper on only some traces. But i'm not sure what this costs.
https://www.epectec.com/articles/heavy-copper-pcb-design.html

[tszaboo]

If you look at that TI design, you see that the entire input current flows through L200, which will barely heat up 30~40 degrees. I tend to think they went a bit overboard with those busbars, the input current is barely 50A. Similarly, I don't think there is too much need for the busbar for the output terminal. I mean it looks impressive, and maybe you need these for automotive, where the ambient temperature can be quite high already, maybe they are struggling to keep the heatup low.

When I was doing high current, we had copper busbars. They were nickel plated , cut and bent to shape and soldered to the board.

Woah that looks impressive.  I wonder if at some point you can skip the PCB and put the SMD parts directly on the copper busbars, "macro PCB".

There are threaded SMT terminals ready for PNP assembly (example),

I recently used some of these in a project: https://www.lcsc.com/product-detail/SMD-round-nut_Sinhoo-SMTSO3060MTJ_C2916383.html

They're ferrous, not sure if that's a dealbreaker for conductivity.  EDIT: Just cut one open and it's solid brass -- perhaps just the nickel plating was magnetic. 

The spec'd hole size has some slop in it, so keep in mind your tolerances when designing the holes & misalignment clearances of your busbars.

Paste + hotplate reflow wasn't enough to solder them on its own, but briefly touching them with a soldering iron whilst the boards were on the hotplate worked magic.

I was wondering about these parts as well. I tried sourcing similar parts from MAC8 or Würth, but they were very expensive. The Chinese parts seems to be priced  with the cost-plus model, and that is a game changer.

What surprises me is that nobody seem to offer nickel plated Aluminium busbars as a service. Aluminium PCBs are very cheap, and their base material cut to shape would be perfect for busbars.

[ANTALIFE]

Consider the extra cost of getting bus bars made to size and the bolts to hold them on and the labour to install them.
Then compare that to just getting your PCB fab to put thick copper on the PCB.
Maybe crazy thick inner layer to spread the heat out, or if you need external copper then maybe one side of the pcb can be super thick. Just be aware the thicker the copper is the wider your track width and clearances have to be.

Some fabs can also plate thick copper on only some traces. But i'm not sure what this costs.
https://www.epectec.com/articles/heavy-copper-pcb-design.html

30oz... that reminded me of this beast:
https://imgur.com/a/having-fun-soldering-20oz-700um-pcb-s2zcN

[thm_w]

hah that section of the board makes no sense, and look at the vias they can't even stack to the top.

[RetireMeNow]

Consider the extra cost of getting bus bars made to size and the bolts to hold them on and the labour to install them.
Then compare that to just getting your PCB fab to put thick copper on the PCB.
Maybe crazy thick inner layer to spread the heat out, or if you need external copper then maybe one side of the pcb can be super thick. Just be aware the thicker the copper is the wider your track width and clearances have to be.

Some fabs can also plate thick copper on only some traces. But i'm not sure what this costs.
https://www.epectec.com/articles/heavy-copper-pcb-design.html

30oz... that reminded me of this beast:
https://imgur.com/a/having-fun-soldering-20oz-700um-pcb-s2zcN

I've had this priced out and the heavy plating option is crazy expensive. Best suited for defense/military budgets.

I've used SMT bus bars similar to the last link but if over ~50mm you *MUST* allow for different thermal mass during expansion/contraction during reflow (Hint Bus bar holds the heat much longer). This leads to buckled boards. Use some 90 bends or hoops in the bus bar design to circumvent this.

[Psi]

I've had this priced out and the heavy plating option is crazy expensive. Best suited for defense/military budgets.

That's good to know, I always wondered what it costs.

Although, maybe they gave you an inflated price because they simply didn't want to do it. Did you get a quote from multiple places?

Also, how expensive is 'crazy expensive'. If a typical PCB costs say $4 each, and you added a 30oz trace, would you expect that to become roughly $8, $20 or $50 each? (based on what you were quoted)

+$8 or +$20 might still be workable in some cases. Since high current products usually have a good amount of profit in them.
So it's really down to what advantages it gives you over a normal bolt-on metal plate, and if that matters to you.
For a medium/big company it's probably always better to design a copper plate and reduce BOM costs, but for a small home based side-business it might be worth it to pay +$20 and keep things really simple for design/assembly.

[tszaboo]

I've had this priced out and the heavy plating option is crazy expensive. Best suited for defense/military budgets.

That's good to know, I always wondered what it costs.

Although, maybe they gave you an inflated price because they simply didn't want to do it. Did you get a quote from multiple places?

Also, how expensive is 'crazy expensive'. If a typical PCB costs say $4 each, and you added a 30oz trace, would you expect that to become roughly $8, $20 or $50 each? (based on what you were quoted)

+$8 or +$20 might still be workable in some cases. Since high current products usually have a good amount of profit in them.
So it's really down to what advantages it gives you over a normal bolt-on metal plate, and if that matters to you.
For a medium/big company it's probably always better to design a copper plate and reduce BOM costs, but for a small home based side-business it might be worth it to pay +$20 and keep things really simple for design/assembly.

It was a decade ago, but as I recall, for me it was a lot more expensive than adding soldering-in 4 custom bent/cut/plated busbars.
The price was of course a lot better when talking about thousands of boards. If the PCB is complicated with many weird shape bubars, then probably this method is cheaper. Also PCB size as I recall doesn't matter too much, it's the extra work that you have to pay for.
But if you want to be price competitive, system design will help a lot more. Try to get the board to standard 4 layer board with increased copper thickness, and don't use these exotic techniques.

[temperance]

Long ago when 4 layer boards were extremely expensive and only for the happy few you could buy trough hole bus bars from various manufacturers to build power hungry computer systems with TTL IC's. Of course many disappeared. But I needed something like that not so long ago. Here is what I found:

https://www.e-fab.com/capabilities/products-stiffeners-bars/

https://stormpowercomponents.com/bus-bar/circuit-board-stiffeners/
http://static6.arrow.com/aropdfconversion/40996864df4b45b81ade7fbba2138962cafe263a/rigidbus.pdf

https://schroff.nvent.com/en-ba/products/enc30925-001

An other option is nickel plated strips like those used to connect batteries in battery packs. Companies making stencils for solder paste application can also cut those materials to any shape you envision.