FCDEV3B component quality issues

Mychaela Falconia mychaela.falconia at gmail.com
Tue May 2 21:13:58 UTC 2017


Serg wrote:

> Interesting! Possibly components quality issues or other variables...

In the case of component quality issues, the one component which is
most suspect is the flash+pSRAM chip Spansion S71PL129NC0HFW4B.
Openmoko's modem design which we've copied for our FCDEV3B supports
the option of using several different flash+RAM chips on the same PCB
footprint, Om's mass-produced units used Samsung K5A3281 (4 MiB of
flash, 1 MiB of SRAM), but I wanted to use the same high-capacity chip
found in the Pirelli DP-L10 (16 MiB of flash, 8 MiB of pSRAM), and it
fits the PCB footprint.  (Another issue with the K5A3281 - Openmoko's
choice - is that I wasn't able to find a datasheet for this exact part,
only for some of its relatives, and I am not too comfortable with using
parts sans proper documentation.)

The Spansion S71PL129NC0HFW4B chips that have been populated on our
current 8 FCDEV3B boards have been bought from Jotrin in China back in
2013, together with the Calypso, Iota and RF PA chips on our boards.
The other chips I bought in that initial order (Calypso, Iota, Rita
and RF PA) all came in proper packaging (Calypso chips on a tray, the
other three on cut tape from reels), but the Spansion chips came loose
in a thin plastic bag, not even ESD or moisture-protected.

All of these parts sat in my storage for a little under 4 y until a
few months ago when the time finally came for us to build our first
FreeCalypso hardware product.  All other components went directly to
Technotronix (the assembly shop) as the BGAs already had proper SnPb
balls and for non-BGA packages the lead material doesn't matter, but
the Spansion chips went through a BGA reballing step first: they came
with RoHS balls (or at least I presumed so, see below), and we needed
SnPb balls in order to do the traditional SnPb solder reflow without
the crazy high heat used for RoHS.  I separated (manually with tweezers)
20 out of the 100 Spansion chips I bought from Jotrin back in 2013 and
sent those 20 to the reballing company; when the reballed chips came
back, they went to Technotronix, and 8 of them ended up on our FCDEV3B
boards.

The next big surprise happened three weekends ago when I was checking
out all 8 boards as I needed to ship one to Harald Welte and another
to Serg.  On 7 out of the 8 boards the initial state of the Spansion
flash was fully blank as expected for NOS parts, but the chip on one
of the boards had previous content in it!  It was some MTK firmware
image; I saved it before erasing the flash, and I just posted it on
the FTP site:

ftp://ftp.freecalypso.org/pub/GSM/unsorted/spansion-flash-mtk-image.zip

Unless that MTK firmware image was somehow magically (etherically?)
programmed into an on-the-shelf, never-populated flash chip through
telekinesis, the presense of that MTK fw image in one of the chips is
pretty damning evidence that at least some of these Spansion chips
we've got are not NOS but depops.  And from what I understand about
BGAs, a depopulated BGA must be reballed before it can be repopulated,
so the reballing which I put these chips through was probably not
their first.

Going forward, we should probably acquire another batch of these
Spansion chips from a better source, but at the present stage I feel
that we can continue with the bring-up of our FCDEV3B boards as they
are.  Out of the 6 boards that are still in my hands only two exhibit
issues which may be related to the Spansion chip: S/N 001 has the high
strangeness when booting from flash in mode 1, but the workaround of
using flash boot mode 0 works perfectly well, and S/N 004 is unable to
boot Magnetite fw at all (I haven't tried Citrine), neither from flash
nor via fc-xram - it could conceivably be a problem with the pSRAM
part of the Spansion chip.  I plan on looking into the issue on S/N 004
a little later, after higher priorities like finishing the automated
calibration software have been taken care of.

BTW, the board on which the flash chip came with previous MTK content
is the one that went to Serg (S/N 003).

Hasta la Victoria, Siempre,
Mychaela aka The Mother


More information about the Community mailing list