In a 4-layer PCB, is the outer copper etched before pressing or after?

[TheUnnamedNewbie]

So most 4-layer PCBs are made with a prepreg-core-prepreg stackup. The outer copper is foil. Is this foil etched before the prepreg is cured, or after?

[eugene]

A 4 layer board is just two double sided boards glued together, right? The only order of operations that makes any sense (to me) is performing all of the etching before gluing together with prepreg. Followed by drilling, plating of holes, ...

Doing the etching in two separate steps has no advantage that is visible to me.

[free_electron]

that's not how it works.

4 layers start off as a double sided board.
0 ) take double sided board , apply dry-film photoresists , expose (mask or LDI or maskless imager) , develop.
1 ) etch the copper structures ( we don't drill holes ! )
2 ) apply isolation material ( prepreg ) on the outsides , then cover in copper foil and bake it in a press.
3) drill all holes that need plating
4 ) apply photomask (negative image ! copper that needs to remain is visible) , expose ,develop
5) plating. this is called selective plating. you only grow copper where needed (where the photo mask is open , and in the holes
6) tinflash. a layer of chemical tin is applied. the tin only sticks to the visible copper (traces and in the holes )
7) remove photomask
etch. only copper that is visible is eaten. the tin coated copper is not etched. the etchant only eats copper, not tin.
9) strip tin (if needed )

there are other processes like non selective plating where they plate before photomask. but that is wasteful: you grow copper to eat it afterwards. so it is not really done anymore
they always use negative mask. for reworkability reasons.
- negative mask HARDENS in light , positive mask weakens in light. so a negative mask only gets better (harder) when exposed to ambient light.
- negative mask remains can easily be removed (retouched). assume there is a speck of dust on the film. that area did not get light so the mask is not hard. after developing you see pinpricks (very easily spotted in the AOI as copper is refelctive. it lights up like a christmas tree ! . simply fill them with some resin and cure. Worst case you will have a tiny little hole in the copper. with positive mask a speck of dust becomes copper ! risk of shorts !
opens are easier to find during electrical test. you only need to test end-to-end on a single node. a sort is hard. you need to test any trace against any other adjacent trace.

there are many more steps that i have skipped. planarisation, backbuttering ,copper rough-up  wall activation before plating. post plating ENIG , ENEPIG , HASL. soldermask , designator

you want to reduce runtime.

a six layer is different.
you can do a core and have two laminations. but that means two times  in the oven. that makes the board brittle and takes a lot of time ( 4 to 6 hours per run)
so the best is to make two double sided etched boards , put some insulation between them , apply oute foil , dril , mask , plate ,tinflash, etch
you only have to do the oven step once.

with 8 or more layers all bets are off. you can do 3 cores (6 layers) and do a single oven run. but when you start playing with blind and buried vias or layer vias the whole thing goes out the window.
most costly is ELIC (every layer interconnect). that uses laserd, stacked vias and requires an oven run per layer you add. but it is extremely flexible as you can make exactly what you need. gor from layer 2 to layer 9 ? no problem. 2 to 8 , 2 to 7 3 to 6. anything goes. you are not bound by symmetry or pairs. you also have no stubs

things get even more complicated as you start doing via in pad or backdrilling and plugging. there are interconnections between layer 2 and four but they are not visible on the outside. they have been selectively drilled and capped by the outer foil. that can be done for signal integrity , but also for creepage reasons. high voltage boards do not want access to the internal nodes , or even have the ability for copper to zap through. those are special techniques and you CAD software needs to support that. it requires specials rule sets. and the tool needs to understand that process.

and we have not touche resistive or capacitve embedded layers or even embedded components , where elements are actually buried inside the board.
and then there is flexible board, rigid-flex, metal core , embedded wires and copper coins (for thermal or current reasons) . boards with thick copper inside and thin outside.
3D printed boards (FDM)

There is a lot of technology in a pcb.

[Feynman]

Yeah, in a Prepreg-Core-Prepreg configuration etching (as well as drilling, plating, ...) is done after pressing.

[langwadt]

that's not how it works.

4 layers start off as a double sided board.
...

afaik it can also start as two double sided boards, just not so common

[Doctorandus_P]

Eurocircuits - how to make a 4-layer PCB (full version)


Several other PCB manufacturers have also made youtube vids about their process but the above from Eurocircuits has quite a lot of detail.

Some of the big chinese factories have much more automation in their process.
Instead of humand pushing around  carts with PCB's and loading them into machines, they just have conveyor belts  to transport the PCB's from machine to machine. It's a bit like a row of SMT placement machines, but instead it's the PCB manufacturing itself.

[free_electron]

that's not how it works.

4 layers start off as a double sided board.
...

afaik it can also start as two double sided boards, just not so common

nope.

you would need 3 etching steps for that.

etch bottom of top core , etch top of bottom core. four masks (you need to preserve the outer copper foils. two masks , two individual panels.
laminate
photomask again
drill
plate

so you waste two masks, one etching step and you run the risk of pinpricking the outer copper foils. not done

[free_electron]

if you want to see a real factory look up Tripod or TopSearch
www.tripod-tech.com

Companies like Topsearch easily produce 1.5 million square feet of finished PCB a month.

The largest one by far is Unimicron
this is one of their factories in germany :

[TheUnnamedNewbie]

so you waste two masks, one etching step and you run the risk of pinpricking the outer copper foils. not done

it is done if you ask for it and have very good reasons (eg, I've had a 4 layer board made like this because we required very specific core materials on the outer layers, in a weave type that was not available in prepreg. In hindsight, it might have been possible to do it with prepreg after all, but at the time the choice was made to do it that way, and the manufacturer didn't complain)

[Doctorandus_P]

The largest one by far is Unimicron

They may be good at making PCB's but it's a lousy video.
It's not much more than some voice telling how great they think they are themselves while showing panels hiding machines.
The EuroCircuits video shows how they worked (about ten years ago) and it shows much more of the actual manufacturing process.

[free_electron]

so you waste two masks, one etching step and you run the risk of pinpricking the outer copper foils. not done

it is done if you ask for it and have very good reasons (eg, I've had a 4 layer board made like this because we required very specific core materials on the outer layers, in a weave type that was not available in prepreg. In hindsight, it might have been possible to do it with prepreg after all, but at the time the choice was made to do it that way, and the manufacturer didn't complain)

yes, of course the fab will do it if needed, but it is not common to go that way. I've done selective stackups where the outer core was made from a special RF material. Getting it to bond to copper is problematic and you need super smooth copper so the factory only provides it as a ready made "core" so in that case you do need special processing which you will pay for.
Due to the smoothness requirements they need to install special machinery to replanarize the outer layer post-plating. Radar applications. (phased array antenna)