Y'know, now I wonder what exactly drives wire bonding. So, we have CSPs now of course, and that's probably just because it took a long time for PCB-level density to get high enough to demand it. At the same time, you're paying a premium for the fine pitch PCBs (probably HDI as well) to use them. So it's not used in much quantity, except when there's no choice (i.e., cellphones yes, washing machines no).
Which have in turn been available in many forms over the decades, whether as laminated interposers (who makes those anyway? they're basically PCBs, but not...), or fired ceramic-metal bodies (classic IBM). Obviously the latter isn't exactly affordable, but the former is essentially any old PCB, if not a full sized one. Were they just not able to hold tolerances over whole-PCB distances, or something with multiple layers (getting copper, soldermask, vias, etc. to line up) or what?
And why not interposers all the time, or at all? Lead frames must suffer some tolerance issues, perhaps? Otherwise, why not flip-chip directly to them? (They sort of do this with some power transistors, with wide top-side connections rather than bondwires.) Or if interposers are superior, then why not use them instead of lead frames? Well, obviously lead frames must be cheaper, at least for that middle space where the number of leads, and density, isn't too crazy, and who cares about flipping.
On the other hand, clearly wirebonding isn't a painful process; it's done fast and automated, and adapts the fine pitch of the chip to the coarser pitch (and looser tolerances) of the lead frame. The question of course is, how much can an alternative save, if any? Evidently, not much in the average case, but I just wonder what the quantitative breakdown is.
Tim