Cross-hatched planes in a flex circuit consist of latticework of copper on the surface. Flex PCBs with cross-hatched planes involve openings at regular intervals. The use of cross-hatched planes is common in flex circuits.
The hatch pattern offers dual benefits to flex PCB. The special latticework provides ground benefits and also offers structural support.
Interconnects on a hatched ground plane have a characteristic impedance from 30-40 Ohms along the straight runs. In the curved region of the flex cable, the impedance variation is much larger with characteristic impedance ranging from 30-60 Ohms.
The wide linewidth (W/H = 4) creates regions of very low odd-mode impedance over the copper, while regions between the copper are much closer to a 50 Ohm target. The exact variation appears to be approximately 28-62 Ohms, or an average of 45 Ohms odd-mode impedance. The differential impedance comes to approximately 78 Ohms.
Based on the cross-section impedance values shown above, it becomes clear that the actual return loss will vary along the interconnect. You can run an S-parameter simulation from this geometry to determine the return loss up to very high frequencies. The return loss is acceptable and the insertion loss is low due to the width of the traces. One problem at higher frequencies is mode conversion, due to the hatched ground plane.
So if the layer above/below the pre-preg has lots of missing copper in one area, but not in another, then the resin will flow to the empty area, and the resin/glass percentage will be different between those areas? And for a new material, you might not know how much of an effect this would cause?
Is there something else that would cause problems (or new variables?) with the stackup?
Well, that’s probably more of a manufacturing question, so maybe we’ll get one of those guys to jump in; but here’s how I understand it. Your description is basically what happens, It’s not a lot, as you remember inner copper is typically 0.0007" so we’re not talking a lot of material here, and even without that most dielectrics change DK over distance, temperature, frequency…you get the picture. Often times a board like that might not even be a high speed board. Heavy ground pours occur just as often on power control boards and they are pretty insensitive to impedance shifts.
Other things that could shift in a new material (sometimes even a new batch of material) are things like water absorption, and all kinds of shifts due to temperature.
Fortunately we’ve made boards out of almost everything, and stay on top of the process control aspects to a degree it’s hard to believe. It’s the only way we can reliably make some of the ultra high tech boards.
Where would I find that “what we can hit without too many errors”? Do you mean the 10% (5% for advanced or micro) listed at Rigid PCB Capabilities | Sierra Circuits ?
Yes and no. Clearly we must hit specs as much as possible. One hundred percent is not to good. So specifying 10% for instance one needs to consider measurement error for instance. So you add a guard band. I am not privy to either our actual failure rate or the failure rate used for profitability of jobs, costs, etc. But somewhere someone has determined that we can offer a 10% spec and count on no more than xyz failures, and I deduce this because somebody has to set prices. If any of this becomes important for some issue relative to your company I’m sure management and QA would be glad to assist you. But you reached the limit of my knowledge of that type.
If the Lamination Press is set up correctly the resin will flow/fill evenly across the entire surface even with differences in copper density but it would likely take more resin to fill than a solid plane layer. For new materials we always run a First Article panel to check the actual press-out thickness. Un-balanced and heavy copper weights are stack-up variables where again an F/A panel would probably be called for. We make a lot of F/A panels.
I’m only trying to get a feel for what expectations are reasonable to just assume, vs which ones are asking for heroic measures. To come at this from another direction, is it common to be at +9% in one part of the board and -8% in another part of the same board, or is more that some boards will be at +9% and other boards at -8%?
Is the full ±10% still part of normal (unpredictable) variation, or is it more that you can choose between discrete steps, and ±10% will mean that at least one or two of those discrete steps is within design tolerance?
With drills, I’ve seen designs specifying multiple “different” drill sizes that are all much closer to each other than the hole-size tolerance. Supposedly the manufacturer would consolidate sizes when substituting in physical drill sizes that they actually had on hand. It wasn’t clear to me whether the differences were a sort of tag (like marking a width of 6.0001 instead of 6 to signal special treatment) or reflected the actual ideal size, or maybe even indicated that the design was still sort of abstract. I’m trying to get a handle on the Controlled Impedance equivalent of those overly precise drill sizes.
Well hopefully they’ll all be at +/- 1% or less. Not being in manufacturing, and not have ever seen any statistical process control data I couldn’t guess. There are different issues which would give a different skew, but I don’t know what ones they have to deal with. My hunch is all minor deviation from ideal is a Gaussian distribution of all the tertiary and quaternary factors making up the PCB fab process. Major problems I believe are almost always traceable to some misstep.
Choosing between discrete steps sounds digital and I think the whole process and most steps are all essentially analog – that is, continuously varying. Not sure I completely understand your question. Keep in mind the goal is that every PCB be +/- 0 variation, which we aim for but due to the sheer number of variables can’t be achieved except randomly.
I strongly suspect that if you see drill sizes separated by a tenth or hundredth of a mil its for identification purposes. You can bet the same bit is used on each one. It’s mostly an easy way for the designer to keep track of what he has and pass that information to the fab house.
For any more detail on your first two items you’ll have to reach out to manufacturing.