Hi, all - I'm struggling with a recalcitrant mail host at the moment,
so have changed hats to gmail - it's still the same old
steve@arbury.com , though, for those who don't know me. (struggling
with broken remote mailhosts is even more tedious than LPDDR timings).

Some responses:
Hector - that supplier of Micron parts - have you used them / got
prices? That looks suspiciously like the full list of Micron orderable
parts. We definitely do need to sort out which shape LPDDR chips we
want / can get, and quite soon. If they're a real vendor, that's
super!

(An LPDDR point of note - I've got a pile of Beagle-xMs that have -166
memory (in PoP package, so tiny short buses). They don't work at all
well at 200MHz. There's evidently not a lot of slack in this system.
This may, or may not, bode ill for stacking two multi-die 4Gbit x32
chips on B4).

Hector - STM32 H107 Yeah, can't see why not. How much current is it
taking in real sleepy-modes?
A good PMIC would indeed be handy, but there seems to be no such
thing. Forcing the PCB onto laser-drilled vias just to fit a PMIC
designed for cellphones seems like a terrible waste. There are other
manufacturers of PMIC-type devices. Exar, Maxim, and no doubt others.
They don't (as far as I know) do the SmartReflex thing we need to get
the best speeds out of our OMAP. There's also the horrible fragility
of the TI PMICs. I've got a pile of blown-up Beagles & XMs, which
haven't survived contact with software types. Because the PMIC is on
such a fine process, there's no possibility of just whipping it off
and replacing it.

Hector - Yes, PWM is mildly handy for LCD backlight, but it's easily
available elsewhere. Not a showstopper.

RTC: There's a crappy one in the OMAP, but I'd rather use something in
the micro. (Using an MSP430 for this sort of thing is my normal
tactic, but, apart from my bitter experiences with ST, I'm OK with
trying something of theirs.

David: Yes, the lack of SPIs is a bit of a concern. We do have a
decent pile of i2c, though, for slow IO. SPI is only really necessary
for fastish ADCs or audio, both of which I think are adequately
served.

Patrick: Hi, and...
Yes, I'll package up some progress PDFs for you, probably tomorrow now.
The reason I'm trying to cram some useful IO on this board is that,
historically, expansion boards don't get built. I (/we) hope to make
it so painless this time that there will be loads, but, for a decent
subset of customers, a Balloonboard s a balloonboard - what it ships
with, it stays with.
There are also issues with presenting the IOs in a less user-hostile,
and silicon-fragile, manner. 1.8V OMAP IOs, and crazy-fast Xilinx IOs,
are OK, but terribly inconvenient to bolt up to many things, and very,
very ESD-prone. Hiding them behind a veneer of breakout board feels
liek the right thing to do, given a choice. High density connectors
are, likewise, a pain. If you're using 2-layer bucketshop PCBs,you
don't want to be trying to hand solder 0.5mm connectors, or break out
a grid of high density stuff. Getting a solder short and wiping out
your Balloon is distressing.
However, I'm definitely open to any and all suggestions. (except, maybe, HDMI :)

Also, you people being so dismissive of S-video. While it'll not make
a good 1600x1200 programming desktop, I think you're being
unnecessarily negative. A well-done S-video display makes for a pretty
tolerable UI / webpage. Far better than a 320x240 smartphone, and
people get by on those. So, what's the use case? I shall bang on about
this until I get an answer I feel is acceptable.

Steve