/gsm/com/rfcap: tri-band GSM modem believes itself to be quad-band

Michael Spacefalcon msokolov at ivan.Harhan.ORG
Thu Jan 16 05:57:14 CET 2014


joerg Reisenweber <joerg at openmoko.org> wrote:

> Bottom line: even 850MHz FR can do 900MHz but the reception is pretty poor=
> though just sufficient in usual urban environment.

Hmm, interesting!  Of course given the geographical theater I operate
in, my interest would be in going the other way around.  The funny/sad
thing is, I do not currently have *any* GSM (or GSM+newstuff) device
which has "proper" support for the 850 MHz band (the secondary GSM
band in the lands I operate in) and which I can hack in any way at all:
sure, there are 3G+ smartphones all around me, and they all claim to be
quad-band when they go into GSM fallback mode, but of course they are
the typical Android etc crap, and I don't know how to hack them (if
that can be done at all), so I don't even know whether that SGS2 (as
an example) is connected to an 850 MHz network or a 1900 MHz one!

As to the hackable Calypso devices I have (one FR and a few Pirelli
DP-L10s), they are all 900/1800/1900 MHz, so to this day I don't even
know what GSM services might exist in the 850 MHz band around me...
I do use one of these devices as my everyday phone though (and the
non-Calypso Mot V66 I used previously was/is 900/1800/1900 MHz as
well), so basically all my cellphone-using life I've been using what
is effectively a 1900 MHz only device.  I roam over a pretty wide area
and have yet to encounter a spot without good coverage - and it is
kinda neat when I realize that I get all this great coverage despite
my phone supporting just GSM1900 and absolutely nothing else!
(Meaning nothing else relevant to my geographical area.)

But given what you just said, I am now tempted to try running cell_log
(from OsmocomBB) set to the [128,251] ARFCN range (corresponding to
the 850 MHz band), and see if it can pick anything up.  I'll try it
when I get some time to play.

> I don't think network/BTS can "direct" the mobile to another channel, it can
> only adveritse other channels but it's up to mobile to determine if that=20
> channel actually works in the particular situation

Indeed I think that the scenario I hypothesized in my original post is
very unlikely in reality, but as far as the network directing a mobile
to a channel, I was thinking along the lines of what happens in call
establishment, channel hopping etc - granted, I still have a lot of GSM
learning ahead of me as I set out to reintegrated that code from
<http://scottn.us/downloads/peek/> in the place of the binary libs in
leo2moko, but I was under the impression that whenever the BTS basically
tells a mobile "let's continue this conversion on channel so-and-so",
that direction contains not only the new timeslot number, but also the
new ARFCN, and at least in theory the network is free to pick any ARFCN
which it serves and believes the mobile to support as well.

But of course I could be entirely off-base here, as I haven't really
learned that part of GSM yet.

> there may be other=20
> problems to receive 900MHz band than just a SAW filter, like local interfer=
> ence=20
> or whatever.

I see what you are saying: the scenario of a tri-band mobile device
advertising and believing itself to be quad-band (what Om modems do)
is indistinguishable from that of a genuinely quad-band device having
the 4th band rendered unusable by some unpredictable external factors
like local interference.  A valid point indeed.

The fact that, as you say, an FR made for 850 MHz could receive 900 MHz
and work so-so (limp along) does make for an argument in favor of
keeping the "I am QB" setting in that /gsm/com/rfcap file - if that
file were changed to reflect just the bands "officially" supported by
a given FR unit, it would not even attempt tuning its radio to the 4th
band.

Has anyone else in the community ever used an FR successfully in the
"wrong" band?

> When the mobile can't use a certain frequency, the BTS will not
> "force" the mobile to that frequency.

OK, I need to learn some more about what happens in "channel hopping"
scenarios, and how (and by whom) the frequency channel is selected in
those cases.

Oh, and while we are on the subject of GSM frequency bands, I guess I
need to give everyone an update on my efforts to build my own quad-band
Calypso phone.  Right now that project is bogged down in mechanical
design issues which have nothing to do with the number of GSM bands or
even with the Calypso chipset.  What I really want is something as
nearly-identical as possible to the Pirelli DP-L10, and I won't feel
satisfied with anything less than that.

Back in 2013-11 I was in communication with some seller (German
apparently, but advertised on a Chinese site) who claimed to have
around 700 of those Pirellis for sale.  I wanted 100, and the price
would have been some 3-4 months worth of my spare budget.  But just as
I raised the money for the first partial order, the guy told me that
someone else bought all of those phones, they had no more left, and
whoever bought all 700 or however many they had, was not willing to
resell any of them to me.  Talk about bad luck!  If that guy really
had about 700 of those phones, and if indeed someone bought them all
at the price he was asking for, that must have been over 20 kUSD!
Someone out there must have some really deep pockets...

Well, the way my psyche works is that the more someone denies something
to me, the more I want it.  Having given up on the idea of being able
to obtain any surplus Pirelli DP-L10s in any decent quantity (after
having lost a bunch of money to Chinese "sellers"/scammers who claimed
to have them, that is), I am now refocusing my efforts (and my cash)
toward producing a new "candybar" phone, completely from scratch, that
would be mechanically identical or nearly-identical to the Pirelli, no
matter how much it will cost to make the necessary molds etc.

Of course the original 128x128 CSTN LCD module used in the Pirelli is
just as unobtainium as the complete phones containing those LCDs, and
as much as I often do things out of irrational passion (heart ahead of
brain), I still had enough sense to realize that switching to some
currently-available LCD module would be a more sensible approach than
trying to get the original LCD custom-remade.

So I found a currently-made, sold and supported TFT LCD module with
the needed interface (16-pin parallel, framebuffer integrated in the
LCD module itself, so I can hook it up to the Calypso MCU data bus
just like in the Pirelli) and of approximately the right size: it is
128x160 rather than 128x128 (OK, I can fit 2 more lines of text on the
screen when it's time to design the UI), has almost exacly the same
width, but of course it is a few mm longer.  Mounts differently too.

So now instead of copying Pirelli's PCB outline verbatim, millimeter
for millimeter, I have to pay a mechanical designer to look at the
overall mechanical design of the Pirelli DP-L10, see how my new LCD
differs from the original, and create an all-new mechanical design
that would be *conceptually* identical or nearly-identical to the
Pirelli, but shaped around the new externally-sourced parts: the new
LCD I've almost settled on, and whatever currently-available components
I find for the GSM antenna, speakers, vibrator etc.

Only after I get this new mechanical design created will I be able to
start designing my quad-band Calypso "dumbphone" PCB: I need to know
what outline to fit my PCB into, and because of the LCD difference at
least, we already know that it won't be identical to Pirelli's
millimeter for millimeter...

So that's where that project stands currently.

VLR,
SF



More information about the community mailing list