Closing the loop on different Calypso chip versions

[Mychaela Falconia]

Hello FC community,

Our collection of Calypso DSP ROM dumps is now complete:

ftp://ftp.freecalypso.org/pub/GSM/Calypso/dsp-rom-3311-dump.txt
ftp://ftp.freecalypso.org/pub/GSM/Calypso/dsp-rom-3416-dump.txt
ftp://ftp.freecalypso.org/pub/GSM/Calypso/dsp-rom-3606-dump.txt

Yesterday (Friday) I got two reworked FCDEV3B boards back from my
assembly and rework people, these boards had the Calypso chip replaced
(on a BGA rework station followed by X-ray exam) with older version
D751774AGHH (the 751774 number refers to the older die revision with
DSP mask ROM version 3416 instead of the final 3606), one of these two
reworked boards came back perfectly working, and I used it to produce
the DSP ROM 3416 dump.

In evolutionary terms, the change from Calypso C05B to Calypso C035,
accompanied by DSP ROM change from 3311 to 3416, introduced AMR support
for voice calls and possibly other features: it appears that AMR voice
memo recording and playback (apparently intended for MMS) was
introduced at the same time, although we don't have a setup yet for
exercising this feature in FreeCalypso.  The same C05B to C035 change
also upped the maximum clock frequency from 39 to 52 MHz for the ARM7
core and from 91 to 104 MHz for the DSP, although I am once again not
sure if this speedup is required for AMR or not.  But then TI made
another die change and the associated configuration management burden
just to change the DSP ROM code once again - thus I can only reason
that whatever was wrong or suboptimal with version 3416 was significant
enough to warrant those costs.

I have exercised this experimental FCDEV3B-751774 (DSP 34) board to
the extent of connecting from it to Operator 310260's live GSM network
and testing a few voice calls.  All calls were connected by this
operator on half-rate traffic channels in AMR mode, and the voice
passed successfully in both directions, so this older DSP does have
working AMR - or at least working under "easy" conditions.  Our FC
Magnetite firmware runs with patch version 4130 (extracted from TSM30
source) when built for a DSP 34 target.

Going forward, we are going to use D751992AGHH Calypso chips (DSP 36)
in all of our FreeCalypso products, not this older 751774 version with
DSP 34.  TI must have had some good reason for making the DSP ROM
change from 3416 to 3606, our L1 version reconstructed from a certain
vendor's TCS211-20070608 semi-src delivery officially supports only
DSP 36 (our support for DSP 34 and for the even older C05B should be
regarded as an experimental backport), and we are not going against TI
on this issue - we'll keep using DSP ROM 3606 chips at least until we
either recover the DSP source from TI or undertake a major effort to
reverse it via disassembly, at which point we'll be able to see for
ourselves what the diffs really are.

Oh, and while we are on the subject of Calypso chips, let me clarify
that all of current FCDEV3B boards are made with real SnPb solder, not
RoHS, using Calypso chip versions with SnPb solder balls (that's what
the GHH suffix means, the other option is ZHH for RoHS), and the same
is planned for all future FreeCalypso hardware products.  We have
tried the RoHS alternative twice now (once with a batch of 4 boards
made in RoHS from the beginning, then again with a Calypso chip
replacement on two boards where the new Calypso chips were the RoHS
version), and all of those boards came out with fatal defects in the
RF Tx path, with the symptoms pointing to the PA.  (It could also be
the antenna switch module after the PA, but the PA itself seems more
likely.)  It appears that the PA (or perhaps the Tx paths in the ASM)
is very sensitive to heat damage, and the higher heat of RoHS reflow
kills it.

Obviously the mainstream GSM phone and modem manufacturers who were in
business over a decade ago had this issue solved somehow, as all of
them used RoHS in the second half of the 2000s decade, but I personally
do not consider compliance with assinine RoHS directives to be worthy
enough to chase after this RF PA RoHS heat damage problem with my own
money - I would rather just build everything in SnPb.  (These
experiments aren't cheap: every failed experiment means having to
throw away boards that cost a lot of money.)  The whole idea of
keeping lead out of electronics makes sense only for consumer crap
that gets thrown away, not for precious FreeCalypso devices which must
never be thrown away regardless of which solder alloy they were made
with.  If you decide that you are no longer interested in playing with
your FreeCalypso device, you must NOT throw it away or destroy it in
"recycling", instead you must pass it on to someone who will give it
the eternal love it deserves.

Hasta la Victoria, Siempre,
Mychaela aka The Mother