20 ounce PCB!

[firewalker]

https://www.reddit.com/r/electronics/comments/5i9wxd/i_just_received_my_20_ounce_pcb_soldering_is/





Alexander.

[Gyro]

Solder paste stencil and a stepladder! 

[M4trix]

The thickest FR-4 board I've used is 2 oz. You're gonna hand solder those SMD components or use a blacksmith's furnace ? 

[Fsck]

The thickest FR-4 board I've used is 2 oz. You're gonna hand solder those SMD components or use a blacksmith's furnace ? 

wouldn't a pizza oven work out?

[ncoonrod14]

Apparently it's for a Formula E power regulator. Wouldn't it be cheaper to etch as little copper as possible? Also, the trace width isn't even that big? Surely it'd be easier to use less copper and wider traces?

[ataradov]

Surely it'd be easier to use less copper and wider traces?

But they probably care about the weight. Otherwise it would be much cheaper to use chassis mounted components and wires.

[free_electron]

That's not going to work ... the copper will shear off the substrate due to thermal stress. have you done a thermal analysis ? that's going to light up like a xmas tree.

For this kind of chicanery you are better off making a 6 or 8 layer using 4oz copper and plugging it with via's.
With 22 ounce copper your aspect ratio is insane and it requires overblown traces even for tiny holes and the plating there will be non-uniform.

Better solution is to make custom copper bus bars that get pressed onto the board.

[rob77]

that's kind of crazy such a PCB, i think it makes no sens to manufacture such a board out of FR4 with a single extremely thick layer......  i would rather mill the traces (or better said busbars at that thicknes ?  ) out of a copper plate and glue them with a high temp. epoxy to a ceramic substrate. that way it would be able to withstand some serious thermal stress.

[M4trix]

or something like this...

[thm_w]

Apparently it's for a Formula E power regulator. Wouldn't it be cheaper to etch as little copper as possible? Also, the trace width isn't even that big? Surely it'd be easier to use less copper and wider traces?

He explained in the comments. The thinner traces are for higher voltage (400V) and the wider planes are for the high current low voltage outputs.
Bus bars are not possible due to height constraints. SMD bars might be possible, I suspect their wasn't enough space or didn't want to bother with them given reliability and high budget available.

The copper is not etched, its plated on (after the first 1oz or so).

[wraper]

The copper is not etched, its plated on (after the first 1oz or so).

In PCB production, copper is plated after drilling vias, and only after that traces are etched. So they are indeed etched.

[rob77]

The copper is not etched, its plated on (after the first 1oz or so).

In PCB production, copper is plated after drilling vias, and only after that traces are etched. So they are indeed etched.

i doubt they use standard process for such a monstrosity. if it was etched, then it would show signs of significant under-etching (top of the trace wider than bottom). the picture actually shows the opposite the traces are wider at the bottom.

if you would want to make it by plating , then i think you could easilly make them this way:
use standard 1oz copper clad
1. drill with larger diameter
2. fotoresist negative - traces exposed
2. copper plate - freaking long time to build up the monstrous layer
3. tin plate (act as etch resist for alkali etching process)
4. remove fotoresist + alkali etch
5. remove tin plate + do whatever plating is required
6. solder mask
7. drill the holes again with final diameter (they would be clogged by copper)

actually it's very similar to a standard process, just the plating part would be very long + twice the drilling

[firewalker]

http://imgur.com/a/s2zcN

[T3sl4co1l]

Wouldn't the plated copper mushroom over the edge of the resist?  What keeps it growing vertically?

A very thick layer of resist would do, but I don't know that that's even possible.  It would be like a UV-cure 3D printed layer, and take forever to strip away!

Tim

[rob77]

http://imgur.com/a/s2zcN

i'm curious how comes someone who is able to pay 1k+ for a PCB be using a soldering iron that looks like it's salvaged from a trash.

and also i'm puzzled why is he using the same monstrous PCB for the low power stuff ? multiboard design would be better i think, separate board for the high power a separate one for the signal part..  or as free_electron suggested few posts above - multi-layer board with via stitching (actually high power/current PSUs are done that way e.g. HP c700 PSU  gives out 200Amps and the PCB looks quite "normal").

[brabus]

Those fuses look definitely inadequate for automotive use; moreover, there seems to be quite a lot of them. I wonder if a more rational design could have allowed for a lower number of proper blade fuses.
Those stacked ceramic capacitors make me cringe. I wonder how they will tolerate the thermal expansion of PCB and copper, plus the vibration of the automotive environment.
Let's not say anything about the bodged (ceramic) capacitor on the optocoupler.

Regarding the PCB, my guess is that it was milled and then covered in a thick layer of soldermask.

I wonder how higher is the efficiency with this monster-thick copper solution, in comparison with the same circuit on a normal PCB.

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

[T3sl4co1l]

Those stacked ceramic capacitors make me cringe. I wonder how they will tolerate the thermal expansion of PCB and copper, plus the vibration of the automotive environment.
Let's not say anything about the bodged (ceramic) capacitor on the optocoupler.

They're both called "automotive grade", meaning flexible terminations I think.  I don't know how flex that actually allows for, but the stacked kind have an extra thing going: they're soldered in a lead frame, which takes up a bit of flex.

Regarding the PCB, my guess is that it was milled and then covered in a thick layer of soldermask.

Nah, they can etch and plate stuff like that.  You can get milled boards like that, though.

You can also get laser cut (or sheared), stamped/embossed, and laminated bus bars, which contrary to a thought voiced elsewhere: they aren't thick at all, indeed, thinner than you see here, because they can be bonded with a thin film (a few thous) dielectric, instead of fiberglass.  They aren't much cheaper, though.

Tim

[ataradov]

Would it be easier to make a standard 2 oz board, but keep high current tracks without solder mask, and then solder thick copper wire to those tracks?

[thm_w]

Would it be easier to make a standard 2 oz board, but keep high current tracks without solder mask, and then solder thick copper wire to those tracks?

Depends who its easier for,
Which would you rather do, manually mask off the high current areas and spend a few hours soldering down copper. Or telling the PCB manufacturer "plate 20oz copper please".

[DerekG]

Those stacked ceramic capacitors make me cringe. I wonder how they will tolerate the thermal expansion of PCB and copper, plus the vibration of the automotive environment.

I'm thinking about how hot the solder paste oven will need to be & how long the board & all the components will have to sit in that heat for reliable connections.

You can see why there are no SMD electrolytic caps designed into this board.

[batteksystem]

All I can say is, this is definitely not the most efficient solution, or the most well thought solution. After all, I think it is possible to deliver the same performance without using 20 ounce copper.

[rob77]

All I can say is, this is definitely not the most efficient solution, or the most well thought solution. After all, I think it is possible to deliver the same performance without using 20 ounce copper.

fully agree , looks like he built two power blocks which output 150Amp 13,8V each, that's roughly 2 times 2100W and yet he's using a water block to cool it => not great efficiency.
i must again compare to the HP C7000 power supply which outputs 200Amp 12.25V , that's 2450W, achieves end-to-end efficiency of 94% and the PCB fits into your palm (ok it's a little longer than a average human palm). the PCB looks quite normal and the high current path (200Amp) is handled by bus bars which also act as a heat sink. you can find a tear-down of that PSU on the "Play with junk" youtube channel.

[cowana]

This board is a good example of the fact that you can have almost anything made if you pay for it, whether it's the best solution or not.

20oz copper and waterblock cooling sounds like it's been designed by someone who is scaling up the low-power knowledge they have, rather than actually being experienced in high (and high efficiency) design.

[rob77]

and also i'm puzzled, what the use could be for such a DC brick in a electric race car... 4kW is definitely not the main drive...  probably powering a DC motor to drive a hydraulic pump (brake power assist) ?
but in any case the efficiency is critical for a race car - every saved Watt-hour on efficiency can be spent on more aggressive accelerations and be faster.

[Warhawk]

I bet this is a university (government) sponsored bullshit.

[nctnico]

For this kind of chicanery you are better off making a 6 or 8 layer using 4oz copper and plugging it with via's.

I agree. Also the inner layers will need much less isolation clearance than the outer layers so you can have much more copper on the inner layers.

[free_electron]

The copper is not etched, its plated on (after the first 1oz or so).

In PCB production, copper is plated after drilling vias, and only after that traces are etched. So they are indeed etched.

not necessarily. for this thick copper they don;t do this. they drill , plate, mask , plate again. as copper growd they may mask again. if you look sideways you will see 'layers' of copper. there will be clearly visible ridges.

[PCB.Wiz]

For this kind of chicanery you are better off making a 6 or 8 layer using 4oz copper and plugging it with via's.
With 22 ounce copper your aspect ratio is insane and it requires overblown traces even for tiny holes and the plating there will be non-uniform.

Yes, but you have to admit, the pictures are really cool !! 

I'm impressed it can even be done at all, just not something I'd use in production.

Any price indications on that PCB shown ?

[ataradov]

Any price indications on that PCB shown ?

Reddit thread says ~ $2200.

[wraper]

but you have to admit, the pictures are really cool !! 

For me it looks defective as hell, those ugly vias, SMD pads of irregular size, ugly tracks (the most narrow ones).

[rob77]

Any price indications on that PCB shown ?

Reddit thread says ~ $2200.

wow ! a proper multi-layer board would cost approx 1000$, so half the price

just for fun entered such a board to pcbshopper 30cmx40cm 8 layers of 4oz copper...   $1,148.08 including shipping via DHL .

but you have to admit, the pictures are really cool !! 

For me it looks defective as hell, those ugly vias, SMD pads of irregular size, ugly tracks (the most narrow ones).

yes, that's a proof of manufacturing via plating and not milling... and a proof of bad PCB design... all those ugly vias are because of too thin annular rings - there was not enough base for copper deposit of that height.

[T3sl4co1l]

but you have to admit, the pictures are really cool !! 

For me it looks defective as hell, those ugly vias, SMD pads of irregular size, ugly tracks (the most narrow ones).

What the heck do you think a regular PCB looks like, magnified to the same scale where the traces look that tall?

Tim

[wraper]

but you have to admit, the pictures are really cool !! 

For me it looks defective as hell, those ugly vias, SMD pads of irregular size, ugly tracks (the most narrow ones).

What the heck do you think a regular PCB looks like, magnified to the same scale where the traces look that tall?

Tim

It would look much better than this, without vias with broken annular rings (some are "eaten away" almost completely), and without traces with 2x thickness variations. It this case, design rules were ignored, quiet obviously.

[T3sl4co1l]

IPC standard for acceptable vias (as drilled) is 0 mil minimum annular ring, or even less (i.e., the drill cut into the outer perimeter of the via pad, but >80%, or whatever the figure is, remains intact).  You might be surprised.

With so much excess copper around, I would imagine the acceptable limits on annular ring are just as relaxed.  You don't usually think the in-barrel plating amounts to anything, but it takes up real board space here!

Tim

[wraper]

IPC standard for acceptable vias (as drilled) is 0 mil minimum annular ring, or even less (i.e., the drill cut into the outer perimeter of the via pad, but >80%, or whatever the figure is, remains intact).  You might be surprised.

With so much excess copper around, I would imagine the acceptable limits on annular ring are just as relaxed.  You don't usually think the in-barrel plating amounts to anything, but it takes up real board space here!

Tim

I know that IPC class 1 accepts almost any crap that will work.

[MaximRecoil]

I would like to know how the printing was done. You can't print on a surface as irregular as that one with traditional silkscreen printing, which involves forcing ink through a screen with a squeegee, and typical digital printing methods such as inkjet or laser can't do that either.

[rob77]

I would like to know how the printing was done. You can't print on a surface as irregular as that one with traditional silkscreen printing, which involves forcing ink through a screen with a squeegee, and typical digital printing methods such as inkjet or laser can't do that either.

silk screens are thing of past.... either UV cured ink/film (similar process as the soldermask) or direct inkjet imaging is used nowadays. considering the irregular surface i would assume direct inkjet imaging was used.

[DerekG]

I've just received an unsolicited email from www.whxpcb.cn

The email states:

Our capabilities
Layers 1-28 layers,
Board thickness 0.125 to 6.0mm
Maximum copper thickness  18oz
Minimum mechanical drilling of holes 0.15mm.
Min line width /spacing is 3/3mil. 
quick turn prototype(2L-4L,2/3day)

[tszaboo]

Apparently it's for a Formula E power regulator. Wouldn't it be cheaper to etch as little copper as possible? Also, the trace width isn't even that big? Surely it'd be easier to use less copper and wider traces?

He explained in the comments. The thinner traces are for higher voltage (400V) and the wider planes are for the high current low voltage outputs.
Bus bars are not possible due to height constraints. SMD bars might be possible, I suspect their wasn't enough space or didn't want to bother with them given reliability and high budget available.

The copper is not etched, its plated on (after the first 1oz or so).

There are bunch of ways to put busbars on PCBs. One of them is to just laser cut the busbar and laminate it with plastic where you need it, and screw mount it flush with the PCB or on standoffs. I guess this could have been solved on a semi-standard 35/70/70/35 4 layer board.
For unknown resons, the desinger decided to do this. Probably he designed his ever first PCB in his life in Eagle, or maybe in MS Paint, and after the tracks were decimated by the current, he just ordered the thickest copper he could get.

[MaximRecoil]

silk screens are thing of past....

Not for everyone. But as I said, that can't be done with screen printing anyway.

either UV cured ink/film (similar process as the soldermask) or direct inkjet imaging is used nowadays. considering the irregular surface i would assume direct inkjet imaging was used.

I mentioned inkjet, and inkjet machines can't do that, not unless the nozzles were mounted on a 3-axis setup like with a 3D printer, and the coordinates corresponding with the substrate's irregularities were programmed into it. Otherwise, there's no way for it to spray a focused line at varying distances from the nozzle without the nozzle "knowing" what those distances are, nor would there be any way for a conventional inkjet nozzle to rotate 90 degrees in order to spray up the side of the "wall" of a tall PCB trace.

Whatever they used, I would like to see it operating so I can see how it works.

[rob77]

silk screens are thing of past....

Not for everyone. But as I said, that can't be done with screen printing anyway.

either UV cured ink/film (similar process as the soldermask) or direct inkjet imaging is used nowadays. considering the irregular surface i would assume direct inkjet imaging was used.

I mentioned inkjet, and inkjet machines can't do that, not unless the nozzles were mounted on a 3-axis setup like with a 3D printer, and the coordinates corresponding with the substrate's irregularities were programmed into it. Otherwise, there's no way for it to spray a focused line at varying distances from the nozzle without the nozzle "knowing" what those distances are, nor would there be any way for a conventional inkjet nozzle to rotate 90 degrees in order to spray up the side of the "wall" of a tall PCB trace.

Whatever they used, I would like to see it operating so I can see how it works.

why do you think the ink must be "focused" ?  ink jet printing is by shooting small drops of ink down to the media you print on... if you drop the same droplet of ink from 3mm height or 10mm height the result will be the very same point of ink on the target... it's droplet of ink not a beam... you don't have to "focus" it.  the walls of the traces on the PCB are NOT perfectly vertical so it's possible to print on them from above.. and actually if you have a close look to the images you'll see the distortion caused by printing on the walls from above.

[julianhigginson]

well, you gotta hand it to him - that's a really unique PCB!