GTM900-like FreeCalypso modem module idea

[Mychaela Falconia]

Hello FreeCalypso community,

I have a new hardware idea: we can create a FreeCalypso modem module
in the form factor copied from Huawei GTM900.  If you look at the
GTM900 module which measures 56.5x36 mm overall, you will see a big
metal shieldcan in the middle, containing the guts of the modem - the
only components outside that shieldcan are the main FPC interface
connector (power, UARTs, SIM, analog voice, misc) and the little
antenna connector.  The shieldcan containing the guts of the GTM900
measures about 41x33 mm.  For comparison, Openmoko's triband modem
core which we have replicated on our FCDEV3B is about 40x30 mm - just
a tad smaller than Huawei's version.  So here is my idea: what if we
produce a module in the same form factor as GTM900, put *our* core in
the middle instead of Huawei's, put in a GTM900-mimicking FPC connector
and route our interface signals to it, and bring out the antenna
interface either on an SMA like on the FCDEV3B or on the same kind of
miniature connector as on the GTM900?

The first obvious question is how would any such product be any better
than the already existing GTM900-B modules from surplus which are
available dirt-cheap.  Aside from triband RF, the main advantage of
the proposed semi-clone would be the addition of a digital voice
interface via MCSI, which is not offered by Huawei's historical
version or indeed any other historical Calypso modem.  Where can we
squeeze in MCSI when Huawei's original 40-pin FPC interface doesn't
have such?  My idea is to reassign pins 32 through 40 (LPG and analog
voice in Huawei's original) as follows:

Pin 32: MICBIAS
Pin 33: MICIP
Pin 34: MICIN
Pin 35: EARP
Pin 36: EARN
Pin 37: MCSI_CLK
Pin 38: MCSI_RXD
Pin 39: MCSI_TXD
Pin 40: MCSI_FSYNCH

This way the most essential analog voice pins are still brought out,
so we are not artificially taking away chipset-inherent functionality,
but the 4 MCSI signals are brought out too.  The rest of the signals
(power, UARTs, SIM and a little bit of misc) will remain mostly
unchanged from Huawei's version.  The choice of SMA vs. miniature
connector for the antenna interface would be up to whoever pays for
it, or if I end up doing it with my own money (*after* my big surgery,
not before, so not any time soon), I will probably go with SMA.

Compared to my earlier and still-standing idea of producing a
FreeCalypso modem in the form factor of an SMT module like BenQ M32,
this GTM900-like module idea would entail far less development cost
and uncertainty.  The development cost for this idea would definitely
fit within 10 kUSD from start to finish, compared to my estimate of
30 kUSD for the SMT module, and there would be other major
simplifications, namely:

1) The modem core would be the one from GTA02/FCDEV3B, avoiding the
much more adventurous path of trying to fit into the quite different
layout floorplan from BenQ M32.

2) The challenges of designing a fancy test fixture with a spring
socket for the production testing and calibration of SMT modules, plus
the challenges of passing the antenna connection through a castellated
interface, would be completely avoided.  The production test setup
would consist of a little test board that would connect to the FPC
interface, providing power, UARTs and a test SIM socket - simple
enough that I can produce it myself - and the calibration setup would
connect to the SMA right on the module or to an SMA-terminating
pigtail which would in turn plug into the miniature RF connector on
the module.

The project name for this proposed FC modem module is FCM40, and I
have already created the MCL and the netlist for it in the
freecalypso-schem repository - look in the fcm40 directory.  (It was a
very simple job, less than 2 hours: I copied the FCDEV3B V2 design as
the starting point, removed everything outside of the core, and added
the new FPC interface.)  Creating a ready-to-fab PCB design can be
done in two ways:

1) If someone wants it fast and is willing to pay, I can give the
layout job to my trusted Altium contractor (the same gentleman who did
FCDEV3B V1 and V2), and I expect that he can probably do it in one
month.  I am still waiting for a price estimate from him, but I expect
a few thousand USD.

2) If no one pays for it and I end up doing it with my own money after
my big surgery, I would take the long road and migrate away from PADS
and Altium toward a FOSS solution.  I would need to create my own fork
of either geda-pcb or pcb-rnd with the PADS-like data model which I
absolutely require and which neither geda-pcb nor pcb-rnd maintainers
agree with, then convert the GTA02 modem core from PADS to the new
FOSS format, then do the new layout for the FCM40, which I expect to
be much simpler than FCDEV3B.  This is the same approach which I wish
to take for my desired FC handset board, so the same work would serve
both objectives.

I would appreciate any thoughts from community members.

Hasta la Victoria, Siempre,
Mychaela aka The Mother