OsmocomBB

[[PageOutline]]

= Minimalistic L1 interface =

This is the description of the minimal interface between L1 and higher layers (L23)

neccessary to establish a dedicated channel with the cell and perform procedures like

LOCATION UPDATE or AUTHENTICAITON.

== Message Types ==

=== SYNC_NEW_CCCH_REQ ===

Triggers the following actions in L1:

 * switch to new ARFCN

 * perform AGC + AFC

 * synchronize to FCCH

 * decode SCH and determine frame number

==== parameters ====

 * ARFCN + BAND

==== response ====

 * SYNC_NEW_CCCH_RESP

=== SYNC_NEW_CCCH_RESP ===

Sent from the L1 to the L23 after it has synchronized to a new cell (FCCH and SCH acquisition)

==== parameters ====

 * struct l1_info_dl

  * ARFCN + BAND

  * GSM frame number (t1/t2/t3)

  * Rx Level

  * SNR

 * BSIC

=== CCCH_INFO_IND ===

Sent by the L1 to the L23 whenever it has successfully decoded four consecutive bursts on

the CCCH.

It is up to the L23 to decide whether the burst belongs to BCCH, AGCH, PCH or SDCCH/4.

==== parameters ====

 * ARFCN + BAND

 * GSM frame number

 * Rx Level

 * SNR

 * 23 bytes L1 payload

=== RACH_REQ ===

Sent from the L23 to the L1 to instruct the L1 to send a RACH request to the current cell.

==== parameters ====

 * struct l1_info_ul

  * Tx power

  * TDMA frame number

  * channel type

  * channel number (timeslot, subslot)

 * 8 bit RA Reference

==== response ====

No response by L1.

The response by the serving cell will be an IMMEDIATE ASSIGNMENT in the AGCH on

the CCCH downlink.

=== DEDIC_MODE_EST_REQ ===

Sent from the L23 to the L1 to instruct switching from CCCH mode to a dedicated

channel.

Initially we only implement SDCCH/4 and SDCCH/8 on a non-hopping cell.

==== parameters ====

 * struct l1_info_ul

 * (arfcn|hsn,maio)

=== DEDIC_MODE_DATA_IND ===

Sent by the L1 to the L23 to convey the successful reception of four

consecutive bursts on the dedicated channel.

It is up to the L23 to do the time demultiplex of multiple SDCCHs in the same

on-air timeslot.

==== parameters ====

 * struct l1_info_dl

 * 23 bytes L1 payload

=== DEDIC_MODE_DATA_REQ ===

Sent by the L23 to the L1 to send payload data on a dedicated channel such as

SDCCH.

==== parameters ====

 * struct l1_info_ul

 * 23 byte L1 payload

=== LAYER1_RESET ===

Sent on (re)start of the Layer1. It indicates that all previous state was lost...

==== parameters ====

  * struct l1_info_dl

  * Only the msg_type has valid data.

== Implementation over RS232 ==

When we transport the messages of this interface over RS232, we have to consider 

 * multiplexing of L1A<->L23 messages with console/debug messages

 * bandwidth management

=== Multiplexing with console messages ===

Instead of a protocol with a fixed header (including length of payload), we chose a minimalistic

subset of HDLC.  This has the advantage that it provides multiplexing between up to 255 different

channels, and it is auto-synchronzing, i.e. when data is lost, the next frame start will still be

recognized by the other siede.

The minimal subset of HLDC consists of UI frames only, with no CRC.

=== Bandwidth management ===

The UART in the Calypso DBB can only run up to 115200bps with standards-compliant baud rates.

Higher baud rates are possible (up to 812500bps), but only as non-standard rates.  While there

are serial chips like FTDI's chip that can support this, regular serial ports and popular

PL2303 based cables don't support it.

Therfore, we have to assume that the RS232 link runs at 115200bps.

 * 115200bps means roughly 11,520 characters per second

 * at a TDMA frame interval of 4.615ms, this means 53 characters per TDMA frame

 * however, since four TDMA frames are merged into one L1 payload, we have about 212

   bytes of RS232 bandwidth for each L1 payload that we receive (260 bits == 32.5 bytes)

This means that regular L1A operation on a single dedicated timeslot is very well possible

even with lots of overhead for protocol layers, anciliary data as well as a bit of debug

output.

But it also means that we can not print much debug info from L1 during every TDMA irq.

Also, actual L1A messaging should have a higher priority than serial console messages.

It would be great to have a strict priority between those two in the serial driver layer

of the phone software.