FCDEV3B update

[Mychaela Falconia]

Hello everyone,

A quick update on where things stand:

* According to the email from pcbcart, the bare PCBs should be done in
  the week of March 13; accounting for shipping delays, I'll probably
  get them in the week of March 20.

* BGA reballing: I sent the Spansion BGAs to the reballing company
  last week, along with a check for $600 to reball 20 BGAs at $30 each.
  The reballing company should be receiving my parcel on Monday: the
  postal tracking system says it got there today (Saturday), but most
  businesses aren't open on Saturdays, so it'll have to wait till
  Monday.  I'm anxiously waiting for the reballing company to examine
  the physical parts I sent them, along with the accompanying notes,
  and tell us if it's all good or if there are some complications -
  I'm hoping to hear from them early next week when they receive the
  package.

The PCBs and the reballed Spansion flash+pSRAM chips are the two last
remaining pieces we need before we can go into board assembly.  Earlier
today I sent an email to Technotronix (the assembly shop I'm going to
use) outlining the preparatory steps we'll need before the actual
assembly (a laser-cut stencil for the solder paste will need to be
made, and some of the parts will need to be subjected to a slow bake
prior to reflow) and asking how much we'll need to pay them for all of
this.  Once again, I'm hoping to get some answers next week.

Finances: after paying an extra $289.45 to pcbcart for the use of
materials that are qualified for GHz RF operation, the $600 to the BGA
reballing company and some miscellaneous expenses (parts for making
the special serial cables etc), we currenly have $400 USD left in the
FreeCalypso account.  The BGA reballing people warned me that there
may be an additional cost of about $200 depending on what they need to
do - I assume we'll know the answer this coming week once they receive
my package and look at the parts - thus we'll have between $200 and
$400 to apply toward the assembly step costs.  And I still don't know
how much these assembly steps will run - again, I'm hoping to get some
answers this coming week.

Serial adapters and cables: I plan on using FT2232D generic breakout
boards from pldkit.com as the serial interface to the FCDEV3B in the
initial bring-up once we get our first boards made.  These FT2232D
adapter boards are currently out of stock at pldkit.com - it looks
like I bought the last ones - but I emailed them and they told me that
they'll be making more of their boards, and they should be back in
stock at about the same time when we start building our FCDEV3B
babies, thus anyone who receives an FCDEV3B from me should have no
problem getting an FT2232D adapter board to go with it.

When using a generic FT2232x breakout board (as opposed to one
customized for FCDEV3B interfacing, see below), the cable connecting
the FCDEV3B to the FT2232x adapter will be a little messy: it will
split into individual single-pin female sockets on the FT2232x end,
and the user will need to connect them to male pins on the breakout
board per a wiring color code chart.  The making of these special
serial interface cables will also be a little tedious with all of the
required crimping.  I already have the necessary parts, but I have yet
to assemble the first cable - I will need to do it some time between
now and the late March timeframe when we'll get the first FCDEV3Bs
assembled.

I also told the gentleman at pldkit.com that further down the road we
would like to have a custom FT2232x adapter board with a header pinout
matching our FCDEV3B, so the two boards can be connected with a simple
straight-through ribbon cable.  He was receptive to the idea, and I
told him I'll revisit that discussion once we get our FCDEV3B boards
built and prove them at least minimally working using the generic
FT2232D breakout board as the serial interface.

Once we reach the point of making the first board assembly attempt,
barring any problems with the BGA reballing or with the still-unknown
cost of the assembly steps, and if no problems occur in the assembly
process itself, the first moment of truth will be whether or not the
Calypso boot ROM responds to UART code download, i.e., to fc-loadtool.
If this step succeeds and we load our loadagent into the Calypso chip's
internal RAM and jump to it, the next moment of truth will be whether
the flash+XRAM chip is alive or not, i.e., can we get a response to
flash chip ID queries and then program the flash.  If we can get this
far, we'll program a firmware image, and the big moment of truth will
come when we try to boot it.  If we can get this far, we will then be
able to branch out into SIM, radio and audio peripheral tests.

If all of the basics work, as well as the SIM interface, I'll put in a
real SIM card from Operator 310260, connect a real antenna and try
connecting to that operator's live network, and I will make this
attempt before the long and steep learning curve for RF calibration,
i.e., with an uncalibrated board.  We run uncalibrated when we run our
FreeCalypso firmware on Mot C1xx phones and it's been working for at
least me and David, so it would be worth a try.  I am not afraid of
the modem acting as a rogue transmitter: the firmware starts the
network connection process by listening for cells in the receive mode,
and it needs to find at least one cell whose BCCH it can successfully
receive before it will transmit the first RACH burst, thus if the
radio tract doesn't work, it won't get as far as turning on the
transmitter.

As far as RF calibration goes, there are 3 parts to it:

1. VCXO calibration, as in the AFC DAC setting to get it onto the
   initial frequency for the cell search and the parameters related to
   how it responds to adjustments;

2. Some Rx path fine-tuning which I still haven't wrapped my head
   around;

3. Putting the Tx power levels exactly where they need to be per the
   spec.

Out of these three, the first one (VCXO calibration) is the most
critical, as it affects the modem's ability to find and connect to a
cell.  We could get lucky and connect to a cell with an uncalibrated
modem, but if it fails, we'll need to reserve judgment until we've
attempted at least this part of the calibration process.  The standard
way to calibrate the VCXO is to disconnect the antenna, connect a coax
going to test equipment in its place, command the modem to transmit
continuously on some fixed ARFCN (obviously something that MUST NOT be
done ever with a real antenna connected), and use the connected test
equipment (TI used R&S CMU200) to measure the actual frequency of the
modem's Tx signal.  I have now implemented all of the necessary
commands in fc-tmsh to send all of the necessary l1tm command packets
to the firmware for this procedure, but I have yet to acquire a CMU200
unit, and more importantly, I have yet to begin climbing the learning
curve that comes with it.

So this is where we stand currently.  Next week I'm hoping to hear
some answers from the BGA reballing company and from the board assembly
shop (Technotronix), find out if we have enough money to cover the
assembly, and take the appropriate next steps.

Hasta la Victoria, Siempre!