FCDEV3B loudspeaker

[Mychaela Falconia]

As I just mentioned in my previous post regarding the sleep mode
hardware bug investigation, I got to play with connecting a loudspeaker
to our FCDEV3B a couple of days ago.

When I originally designed the FCDEV3B two years ago, I included a
provision for an on-board loudspeaker; the intent of this provision is
strictly for the purpose of using the FCDEV3B as a development board,
rather than a GSM gateway or anything along those lines: it is for the
purpose of exercising voice calls and Calypso DSP audio features
(ringtone melodies etc) from a lab bench.  With TI's own D-Sample kit
one can connect the power and the serial cables, boot it up, register
to a GSM network, dial a voice call with AT commands, and hear the
downlink audio come out of the loudspeaker built into the handset part
of the D-Sample kit - and it is loud enough that I did not need to
lift the handset out of its cradle and hold it up to my ear.  I wanted
the same capability on our own development board with the right chipset
for which we can build our own firmware, the ability we lack with the
D-Sample which has an earlier chipset version for which we lack L1
support.

But when I did the original design for the FCDEV3B 2 y ago, I did not
select a specific loudspeaker part, instead I punted on that decision
by bringing the speaker connection out on a two-post header (a pair of
pins sticking out of the board) and leaving the actual speaker
selection and connection for later.  And now that "later" is now...

Looking through the very overwhelming array of loudspeaker part
choices at Digi-Key, I have pretty arbitrarily and almost randomly
picked this one:

https://www.digikey.com/product-detail/en/soberton-inc/SP-1510/433-1140-ND

Note that this speaker already comes with little wires coming out of
it.  If we stick with this choice of speaker, the most sensible way to
put it on the FCDEV3B would be to attach the speaker itself to the
empty area on the PCB that is reserved for the speaker (as marked on
the silk screen) with double-sticky tape or some similar ad hoc means,
and solder the little wires directly into the PCB holes where the J312
header is now.

But my initial speaker experiment was different: instead of soldering
anything, I used the same Molex crimper which I previously used for
the FCDEV3B-to-FT2232D dual UART custom cable to crimp female
terminals onto the wires from the speaker (30 AWG), and inserted these
two terminal-equipped wires into a 2-position Molex crimp housing,
resulting in a speaker assembly that ends in a female connector that
can mate with a two-post header.  I then slipped this speaker assembly
onto the J312 header on my existing FCDEV3Bs.

At first it didn't work on one board, so I tried another board.  This
time I was able to elicit a beep from the speaker with my AT at SND debug
command, so at first I thought it was a board-dependent issue.  Then I
realized that slightly wiggling the speaker assembly with my finger
would make it start and stop working, so there must be a bad connection
somewhere; the crimp connections are my primary suspect at the moment.

Satisfied with being able to make a beep with AT at SND, I tried to play
some E1 melodies from TSM30 - see my previous post for the fc-fsio and
sleep mode issues encountered along the way.  I was able to get the
melodies to play after disabling sleep and getting a successful upload,
but they were coming out of the speaker very quiet, barely audible.  I
thought perhaps something is wrong with the volume level in the
melodies themselves, so I tried bringing up GSM and making a test call
to WWV, a time of day service that also has regular tones - I use it
as my go-to voice call test number all the time.  The flaky wire
connections to the speaker were acting up again, and I only got the
speaker to work for a brief moment during my test call.  The audio I
got out during that brief moment was also very quiet, barely audible,
but it was too brief to really tell, and the speaker quit working
afterward: I could no longer get it to cooperate no matter how I
wiggled it.

My planned next course of action is to remove the J312 header from one
of my boards and solder the wires from the speaker directly into the
holes, thereby eliminating what I suspect to be the unreliable
connection in my current non-working arrangement.  If I can get the
speaker to work reliably without depending on wiggling, I will then
retest both the DSP-generated E1 melodies and the WWV test call, and
get a better idea of the true sound level coming out of the speaker
under real-life conditions.  If it is still too quiet, I will try
increasing the gain of our loudspeaker amplifier by changing a pair of
resistors on the board: see TI's Leonardo schematics and TPA6203A1
audio PA datasheet.  My reading of the TPA6203A1 and Iota datasheets
tells me that we still have some headroom for increasing the amplifier
gain before we run out of the available voltage swing range.

Hasta la Victoria, Siempre,
Mychaela aka The Mother