Hi Tom,
On Mon, Aug 01, 2016 at 03:57:36PM -0700, Tom Tsou wrote:
On Fri, Jul 29, 2016 at 5:18 PM, Alexander Chemeris
<alexander.chemeris(a)gmail.com> wrote:
Burst timing is also passed from osmo-trx to
osmo-bts, so should be trivial
to pass through.
Yes. I don't know where the TOA value ends up in osmo-bts, but the
value is definitely carried per-burst on the socket interface.
For C/I we need a comment from Thomas. Currently
we don't have it on the
osmo-trx interface and I'm not even sure we calculate it.
We don't have C/I directly. We do have a running average 'noise' level
(calculated channel power on non-active uplink slots) and the RSSI
value. The channel noise is queried with the NOISELEV command. RSSI is
carried per-burst on the socket message. C/I can be calculated from
those two values.
This long-time average noise level would make sens on a per-timeslot
basis, or evne on a per-lchan basis once osmo-bts started to send uplink
interference indications to the BSC (which currently it doesn't, on any
BTS model).
Do we need a periodically averaged C/I value or an
update every burst?
The background of this current question is GPRS, where an average over
multiple blocks inside a timeslot is not useful, as the resources are
scheduled fully dynamic.
So to me, averaging over more than the four bursts of one block seems to
make sense only in the context of circuit-switched channels, where the
dedicated channel of a single MS is averaged. But for GPRS, every block
might be a different MS, and the interference into the specific
dynamically allocated uplink blocks within the TBF might be completely
different than for other blocks scheduled to other MS.
I'm assuming that one of the most common interference patterns in
GSM/GPRS is from other networks/cells, which most commonly use frequency
hopping, and thus the uplink interference from one block to another (on
a non-hopping OsmoBTS) could be significantly different.
Literature describes GPRS link adation algorithms based on either CIR
or BLER, so we should have both available in the PCU to play with
different algorithms.
For EGPRS, the values we need to report into the PCU (specifically the
link adaption) are MEAN_BEP and CV_BEP.
Regards,
Harald
--
- Harald Welte <laforge(a)gnumonks.org>
http://laforge.gnumonks.org/
============================================================================
"Privacy in residential applications is a desirable marketing option."
(ETSI EN 300 175-7 Ch. A6)