[aprssig] time for APRS second generation network?

[Henk de Groot]

At 17:34 5-1-2005 -0500, Robert Bruninga wrote:
>No, because the ALOHA calculation already assumes PERFECT CSMA

No, ALOHA doesn't carriers sense, it is "transmit when you have something 
to transmit". The access mode with carrier prior to transmission is CSMA. 
If you are able to detect collisions during transmission than you have 
CSMACD. The latter is used for Ethernet, but is hard on RF unless you have 
a separate QRG to report back what happens on the TX channel.

There are 2 flavours of ALOHA: Pure ALOHA and Slotted ALOHA. In pure ALOHA 
users transmit whenever they have something to transmit. In Slotted ALOHA 
you do the same, but only start at a specific time, for example only start 
at every full second. This only works well with fixed length packets where 
the packet duration is (almost) equal to the slot time. It doubles the 
channel utilisation!

Optimal channel utilisation for:

Pure ALOHA: 18,4% (Abramson, 1970)
Slotted ALOHA: 36.8% (Roberts, 1972)
CSMA:
Depends on the persistency:
   1  persistent CSMA: ~55%
0.5  persistent CSMA: ~70%
0.1  persistent CSMA: ~90%

Now the latter looks good (90%) but it is at the expense of large delays 
(there are no free sigars)... With CSMA you can reach near 100% channel 
utilisation, but the persistency will be so low that the delay will near 
infinite...

CSMA breaks down when it cannot sense the other stations - than CSMA 
becomes pure ALOHA since you then start transmission without knowing who 
else is also transmitting. This is what happens to APRS a lot.

Better schemes are to use timeslots and have a system to reserve timeslots. 
This is how GSM and other TDMA networks work.

(With thanks to Andrew S Tanenbaum's book "Computer Networks").

Kind regards,

Henk.