FCDEV3B project status

[Mychaela Falconia]

Hello FreeCalypso community,

It is time for an update on where the FCDEV3B project stands.  So far
we have only built the first trial batch of 8 boards back in early
April, and 5 out of these 8 boards are still with me.  The other 3
boards have been sent out to Das Signal, Harald Welte and Serg.

Of the 5 boards that are still with me, only one (S/N 007) appears to
be completely free of defects, whereas the other 4 exhibit the
following flaws:

S/N 001: This is the first board I brought up and the one I've been
doing most of my work on.  It exhibits the following two currently
known defects:

  - The strange flash boot mode 1 issue that baffled me originally,
    but was later found to not occur on any other board;

  - Just discovered today while working on calibration: the Tx output
    does not work in the 900 MHz band, the maximum Tx power I can get
    out of it (as measured with my CMU200) is 1.6 dBm, whereas the
    *lowest* Tx power in this band per the spec is supposed to be
    5 dBm.

S/N 002: SIM socket does not work and the RF tract may have problems
too (not sufficiently tested).

S/N 004: Unable to boot FC firmware via either flash or XRAM,
suspected problem with the Spansion flash+pSRAM chip.
SIM and radio minimally tested via OsmocomBB and appear to be working.

S/N 008: SIM socket does not work, but the RF tract appears to be OK.

In the case of S/N 001 I can live with the defects, i.e., this board
is good enough for development use in 1900 MHz lands, in the case of
S/N 008 I will need to have Technotronix re-inspect the solder joints
under the SIM socket and try replacing that part, in the case of
S/N 004 I will need to do some more tests via JTAG once I have the
proper wiring harnesses, and in the case of S/N 002 I will need to do
some more RF tests in order to decide if this board is worth trying to
repair.  All of the above require time, which I severely lack.

We still have enough unpopulated PCBs and parts to build another 12
boards, but before we do that, we need to fully resolve all problems
on our current batch, and that goes back to the time shortage (or is
it womanpower shortage?) problem.

And I am still working on the RF calibration.  One good news is that I
got automated VCXO calibration working.  Of the 3 parts in need of
calibration (VCXO, Rx bands and Tx bands), the VCXO is the most
critical one (required to be able to connect to networks), the Tx power
levels are next (the uncalibrated ones are usually too low, but might
also be too high), and the Rx per-band GMagic calibration is the least
critical, as the differences between the hard-coded GMagic values and
the properly calibrated ones are very small.

As I just wrote, the most critical VCXO calibration is now automated.
My original manual procedure was a royal pita, but with my new
automated calibration tools I just connect the RF coax from the CMU200
to the board to be calibrated and run fc-rfcal-vcxo; the latter
program talks to both the DUT and the CMU200 (via rvinterf and
fc-cmu200d, respectively) and automatically performs the entire
procedure in a couple of seconds.

I also discovered that my CMU200 is partially defective in that the
main RF generator doesn't work, but my unit happens to be equipped
with the secondary (or Aux) RF generator option, and the latter does
work.  For VCXO and Tx power level calibration steps the CMU200 is
used in the RF analyzer mode, which has been working fine all along,
but for the Rx AGC (GMagic) calibration it needs to work in the signal
generator mode, and that was when I hit a snag with the RF generator
not working.  Fortunately the optional Aux RF gen works, and I can use
it as a substitute until I figure out which internal part needs to be
replaced to get the main RF gen working.

My next step on the calibration front is that I need to buy one of
these adapters:

https://www.digikey.com/product-detail/en/murata-electronics-north-america/MXHS83QH3000/490-7708-ND/4358148

This adapter appears to be the right one (won't know for sure until I
have one in my hands to try) for connecting to the RF test port on
Openmoko-made GTA02 units, and I need to connect my CMU200 to an
Om-made and Om-calibrated GTA02 in order to see what the Tx power
levels are on a properly built and properly calibrated unit.  I know
what these power levels are supposed to be per the spec, but the
highest spec power levels could not be attained in my poking at an
FCDEV3B with fc-tmsh, and I don't know if it is some issue with my
measurement setup (loss in the connecting cable and connectors maybe)
or a limitation of the RF PA or other components we are using, hence I
would like to compare against the "golden" reference of Openmoko.

To make matters worse, I am currently broke as all hell because of
REcon.  Back in February I submitted a proposal to give a talk about
FreeCalypso, and in early May I got an email from them saying that my
talk has been accepted.  Two weekends ago (the weekend of May 13-14)
Shannon and I bought our non-refundable plane tickets to go to this
REcon, expecting to be reimbursed for my portion of it, but now those
REcon people have gone silent: they don't respond to my emails, and
there is no update on their website regarding speaker selections even
though they were supposed to be posted in the first week of May.  Thus
I don't even know whether or not I will get to present about our FC
project at the conference, but meanwhile I am $2200 USD in the hole
because of the plane tickets for me and Shannon.  It is really not
cool the way these REcon people effectively screw those they invite as
speakers.

Because of having been screwed by REcon as described above, I am
currently unable to buy the GTA02 RF adapter mentioned above or the
Molex crimp tool I need for making proper wiring harnesses at least
until my next payday...

Until next time,
a very overworked and frustrated Mother