Some L1 updates

Sylvain Munaut 246tnt at gmail.com
Thu Sep 16 11:13:38 UTC 2010


Hi,


In my sylvain/testing branch, there a a few interesting changes:

 * Priority support for items in a scheduler bucket

   Basically for each sched_item, there is a now a priority field and
they are executed in order.

   This allows for example to ensure that a scheduled freq change is
executed before accesing the L1 parameters, or to make sure that a
control channel RX is scheduled before a RACH TX ...

   For new items added into the bucket while executing itself (like
the abort), those will obviously not be able to obey the priority and
will always be executed afterwards.


 * Common DSP & TPU Scenario init & end

   Previously, each primitive _cmd initialized the TPU (with
network/offset) to the timeslot and also ended the tpu & dsp scenario.
   This scheme doesn't work when you have several primitives that have
a TPU window to open and/or a dsp scenario to run.
   (this happens when doing RACH in the same frame as reading a
CCCH/BCCH bursts, or in the case of SDCCH8[4:7]).

   Now, each item specifies via some flags if he has a need for TPU
and/or DSP. If one item of the bucket needs it, then the common code
will handle. :
    - Aligning the TPU (via offset/network) to the 'base' timeslot
(+3 TX offset still has to be added by the primitive)
    - Restoring base TPU alignement at the end.
    - Ending both TPU / DSP scenario when everything is done.


 * A5 support

   Andreas actually made L23 support and added a stub l1ctl_ command itself.
   Dieter and I actually made the L1 work in // and not surprisingly
came up with very similar results given it's mainly:
    . Load the key in DSP API NDB & set the mode.
    . Set a fn parameter properly at each frame
   The rest is handled inside the dsp.


I will now work on merging Dieter's work into the main branch above what I did.

 - TOA regulation loop to prevent drifting
 - TX power control
 - TCH support


Cheers,

    Sylvain Munaut




More information about the baseband-devel mailing list