Hello Marcel,
On Thu, 02 Apr 2009 22:33:01 +0200, "Marcel Klein" <marcel(a)koeln.ccc.de> wrote:
>
> Okay, second try didn't work anymore... so the problem still exists. I
> even tried 250mW.
I received an important hint that the problem is most certainly related
to a clock issue.
Some background: The BS-11 synchronizes its internal clock to the
clock of the E1 line. Its does not have a high stability clock which
is needed for GSM. The official networks use a very stable and exact
clock, the requirement for a BTS according to GSM 05.10 is +/-0.05 ppm.
(There was a variant of the BS-11 sold which had a versatile interface
card which can also work with standard ISDN, this variant had an
oven controlled oscillator which can run standalone with the required
stability. But we don't have any of this BS-11 variant).
The phone itself synchronizes its clock from the network it receives.
If the phone searches for networks, the carrier search range is limited
to a certain window, it will only find a network if the carrier falls
inside this window.
Now the problem: If the phone is synchronized to an official network
and searches for other networks, it will only find them if they are
within the search range. No problem for an official network, they use
exact timing. However if the BS-11 network is not within the range,
its network is not found.
When the phone is powered on and not yet synchronized, the search range
window is larger, however the phone starts looking at the ARFCNs first
which it received the last time when it was registered to a network.
So it normally would find an official network first and miss the BS-11
if its is clock is off too far.
To confirm this, I run a test with an HP 8922 GSM tester. It can
simulate a network and also has a tuneable reference. If I start
with a phone registered to an official network, I can only see the
HP 8922 network if the refenrence is with about +/- 1 ppm. The same
is true the other way round, if I start beeing registered to the
HP 8922 network, I can only see the official networks if the reference
is within the above range. I used an older Nokia phone for my tests,
the range may depend on the phone.
If the BS-11 is locked to the E1 clock, its reference is controlled
by the oscillator of the E1 card. ISDN timing requirements are not
very strict, +/- 50 ppm is good enough. I don't have the datasheet
for the oscillator on the HFC-E1 card, probably the clock can
be in the range of +/- 20 ppm.
Two ideas to solve this: Start at a place were no other network
than the BS-11 network can be received (OK, could be difficult ;-).
Or write the information of the BS-11 network to the SIM card
so that the phone starts receiving the BS-11 ARFCN when powerted
on. Basically this information is the cell allocation information
in SYSTEM INFORMATION 2 and have to be written to EF_BCCH on the
SIM card (if you want to change the location update information
on the SIM card so that the phone is already registered in advance,
EF_LOCI has to be modified).
One other thing: I did not notice those problems with the BS-11
yet. This ccould have two reasons: Differences in the oscillator
drift or because I usually turn off the HFC-E1 card if I don't
use OpenBSC. mISND on Linux is constantly running and the
E1 clock is on all the time, here in my configuration (Windows)
I turn off the HFC-E1 chip after running OpenBSC. According to
some documentation it can take up to 30 minutes until the
BS-11 synchronizes to the E1 clock. So maybe the BS-11 oscillator
is better than the oscillator of the E1 card and so the problem
only occurs if the BS-11 is locked to the E1 card ? Just a guess,
but I will try to see if I find an easy way to measure the BS-11
clock.
Best regards,
Dieter
--
Dieter Spaar, Germany spaar(a)mirider.augusta.de
