A question on PCB manufacturing

FR4 is a specification for epoxy-bonded glass-fibre board materials. If you want a board that can survive higher temperatures, polyimide bonded glass fibre can do better.

You need to search on printed circuit baord materials. The Wikipedia entry isn't all that helpful

and the manufactureres web sites aren't always all that intelligible

ndex.html#laminate

-- Bill Sloman, Nijmegen

A typical tin-lead solder paste may have a suggested peak temperature of 210-225 in its reflow profile. So what you are seeing should not happen with any board IMO. Are you certain about your temperature measurement accuracy?

>

Why so many grounds? You're wasting money.

You *must* be going above 220, more like 260 or greater.

We use, I think, standard board stock on a RoHS process without problems. It took some time to learn and we did have some component problems but everything is smooth now. Our temperature profile hits 250C for something like ten seconds. We're looking into a better oven to control this tighter.

Before we got the process tuned we had some board discoloration, mainly a board with a white solder mask came out sorta pink. Still worked, though.

They tested the board and sent you bad ones? Sounds like you need a new board house. Some are pretty bad.

Err..polyimide sops up moisture. Better yet, use Megtron 6; R-5775K and R-5670K, from Matrix USA. I use that material for units that i GUARANTEE reliable operation to

200C.

Or avoiding cross-talk and reflections. With fast signals you really do need buried ground planes between buried signal layers.

-- Bill Sloman, Nijmegen

Is it really a 9 layer board?

I'd talk with your assembly house to see what temperatures the boards are really seeing, then talk to your board house and specify a material that can handle that temperature (and re-work)?

And FR4 != FR4, there are all sorts of variants with different thermal properties (improved for lead free assembly).

Nial.

Idiot.

r

Really? So what's your alternative approach? The stuff you post doesn't suggest that you've ever actually worked with signals fast enough to rub your nose in transmission line problems, and you are obviously too stupid to realise signals with rise times around a nanosecond or faster need this kind of attention.

You do claim to work with current programmable logic devices, and some of today's parts are quite fast enough to make signal integrity problems tolerably salient - obviously not the ones that you are trusted to work with.

-- Bill Sloman, Nijmegen

No, Slowman, you're just too much of an idiot to understand what you spew. There is no odds in a power plane next to a ground plane *AT ALL*. There is no reason, in fact it's dangerous, to have more than one ground plane and for only eight (nine???) planes, three is absurd. Two orthogonal signal planes per power plane (any power) is fine for just about any PCB usage. That means he can easily use a six-layer stackup. Likely no one would touch nine, idiot.

You're completely clueless, as usual. Why don't you try finding a job instead of pretending, Slowman. Make yourself useful and STFU.

polyamide aka nylon does absorb moisture. It's unlikey there are nylon PCBs.

polyimide aka Kapton and is the plastic that doesn't melt, and is used in flex boards. it's browish.

.

re is

for

anes

eans

idiot.

instead

No. It=92s not a nine layer board. Just eight layer (Top | Power Plane | Gnd1 | Mid1 | Gnd2 | Mid 2 signal | Gnd3 | Bottom). The board has over

500 pads and its area is 370 cm2. I need all of them to isolate 600-700Mbps signals in different layers. The GN2 and Gnd3 layers are used for poser plane in some parts of the board that the high speed signals don=92t exist.

ayer

.

re is

for

anes

Not if you are worried about cross-talk. Making the busses orthogonal does help, but not with a worst case transition - the full width of the bus changing state at once. Leaving out the ground plane between the orthogonal busses also makes the individual stripline transmission lines asymmetric and thus dispersive - not as asymmetrical or as dispersive as microstrip transmission lines on the outsides of the board, but not as good as symmetric stripline.

idiot.

Since he uses a eight-layer stack-up - as he specified in his original post - your latching onto the idea that he uses nine layers is the idiotic propositon. There's nothing unusual or difficult in an eight- layer stack-up.

instead

Since you have just made it clear that you are the one who is completely clueless about using very high speed logic, you would have been better advised to STFU yourself. You have made your own idiocy completely obvious, and made it equallly clear that your continuing employment doesn't involve any work that gets anywhere near the state of the art, and hasn't for quite some time.

-- Bill Sloman, Nijmegen

Well then, force yourself to lower your brow.

Sorry, because of the line wrap, I counted "Mid 2" and "signal" as separate planes.

Blind vias? You really don't need a ground between each signal plane. If you insist on that many solid planes, use half for power. If you have *any* other voltages, put them on planes. Ground isn't magic. There is no reason to have a ground and power plane next to each other. Add a couple more planes for low speed signals. You're overkilling it.

"Poser"?

Once again you demonstrate why you're unemployable.

t layer

2

ly

pew.

There is

and for

planes

Once again you demonstrate that you are the idiot. Your claim that the board had too many grounds planes was posted when we didn't know that the board was carrying "600-700Mbps signals" which Javad Benhangi didn't tell us until the 9th June. My reaction to your post was based on my own experience with slightly faster signals - Gigabit Logic's GaAs parts, in a system built around an 800MHz sysnchronous clock - where the extra buried ground/power planes had seemed like a very good idea.

You've seen - presumably slower - boards that worked with fewer ground/ power planes than Javad Benhangi used, as have we all You don't tell us the edge speeds or the logic swing, of the fastest examples of the boards that you know worked, and you seem singularly incapable of understanding that board stack-ups that worked for TTL coud be marginal - or worse - with modern ECL/LVDS/CML singal levels and speeds.

That level of complacent ignorance should render you unemployable, but there are pockets of complacent ignorance in every industry. They are characteristically inhabited by people like you, who confidently assume that anybody who disagrees with them has to be an idiot, and don't bother thinking about the reasons adduced for the disagreement, and don't see any necessity to produce a rational, evidence-based justification for their counter-opinion.

-- Bill Sloman, Nijmegen

The reason of using this stack up here is to isolate 600-700 Mbps signals that are in different layers. Ground planes are partly used as power planes because there is five different voltage levels and a single power plane isn=E2=80=99t enough for handling the power distribution= .

Anyway, how do you choose your stack up? Which software do you use? Would you please introduce some good document/ application note to help me on designing the stack up on such high speed board? Many thanks

snipped-for-privacy@att.bizzzzzzzzzzzz wrote:

yer

2

ly

pew.

=BF=BDThere is

and for

signal planes

=BDThat means

uch nine, idiot.

e

ng a job instead

te

f you

other

have

r low

You don't need a ground plane between planes to isolate them. If the signals are orthogonal there will be little or no coupling. If the signals are a fairly even density across the board the impedance will be fairly tight. A plane on one side is almost always good enough.

Your English isn't very clear here, so perhaps we're talking past each other... I would suggest having only one GROUND plane. The other full planes can be used for power. The signals don't care whether the plane next to them is power or ground (assuming sufficient bypassing, obviously). In any case there is *no* reason to have a power and ground plane adjacent to each other.

A number of ways to do it (it would have helped here if you hadn't top-posted). Of course one would have to know a lot more about the board (power, packages, layout).

If you're scared (and have deep pockets ;) and insist on 8-layers Top: Signal-1 (most component footprints perhaps low speed wiring) Signal-2 Ground Signal-3 Power-1 Signal-4 Power-2 Signal-5 (decoupling opposite components)

Going cheaper (six layer): Signal-1 Ground Signal-2 Signal-3 Power Signal-4

Top | Power Plane |

I don't do layout, but the layout guy uses Allegro. The job I'm in now doesn't have nearly the frequency requirements you do. I've been there, though.

I've never seen anything useful. There's a lot of junk out there, though.

You never answered the question about what a "poser plane" was.