5 ms·

2G is going away here in Australia. I wonder whether there's a 3G target in the far future.

seba_dos1 12y ago
Implementing 3G is more complicated than 2G by an order of magnitude (and OsmocommBB is still far from having their 2G support complete). Plus Calypso is one of the only baseband processors out there that accepts unsigned firmware (which is exactly why it's OsmocomBB's target) and 3G on Calypso is not possible.
- Zigurd 12y ago
  A major irony is that, in principle, a 4G radio could be made very simple, with all the protocol processing up in the application processor, and sharing code with SIP VoIP and UMA capability in the phone OS. But, as far as I know, no 4G radios are designed that way. Maybe when 3G starts to go away.
  - kristoffer 12y ago
    There are real time requirements on the protocol handling so you can't just send "raw" L1 transport blocks up to the AP.
    - Zigurd 12y ago
      I wasn't thinking of pulling the layer 1 up into the AP. But I don't know... do you have such requirements anymore? You have real time requirements for voice calls in 2G/3G hard handover. If you treat everything as if it were a vertical handover, I think you can even wash that requirement out. And you can fudge the perceived voice call continuity.
      - kristoffer 12y ago
        You have real time requirements in L2 (MAC) in LTE with e.g. HARQ with 4 ms to provide ack/nack. Might not be so hard to achieve but still real time.
        Zigurd 12y ago
        HARQ is a soft real time requirement, and 4ms is within what you can get in Linux interrupt response. That is, it's not going to force you to do hard real time hacks in your AP. Anyway, the real question is: Can you build a 4G radio you can trust because it does little enough that it can have open firmware and driver and protocol software that takes less than 10 years to write? I think the answer is "Yes. Maybe simpler than 2G." Or at least "Much more likely than with a 3G radio."

Implementing 3G is more complicated than 2G by an order of magnitude (and OsmocommBB is still far from having their 2G support complete). Plus Calypso is one of the only baseband processors out there that accepts unsigned firmware (which is exactly why it's OsmocomBB's target) and 3G on Calypso is not possible.

Zigurd 12y ago
A major irony is that, in principle, a 4G radio could be made very simple, with all the protocol processing up in the application processor, and sharing code with SIP VoIP and UMA capability in the phone OS. But, as far as I know, no 4G radios are designed that way. Maybe when 3G starts to go away.
- kristoffer 12y ago
  There are real time requirements on the protocol handling so you can't just send "raw" L1 transport blocks up to the AP.
  - Zigurd 12y ago
    I wasn't thinking of pulling the layer 1 up into the AP. But I don't know... do you have such requirements anymore? You have real time requirements for voice calls in 2G/3G hard handover. If you treat everything as if it were a vertical handover, I think you can even wash that requirement out. And you can fudge the perceived voice call continuity.
    - kristoffer 12y ago
      You have real time requirements in L2 (MAC) in LTE with e.g. HARQ with 4 ms to provide ack/nack. Might not be so hard to achieve but still real time.
      - Zigurd 12y ago
        HARQ is a soft real time requirement, and 4ms is within what you can get in Linux interrupt response. That is, it's not going to force you to do hard real time hacks in your AP. Anyway, the real question is: Can you build a 4G radio you can trust because it does little enough that it can have open firmware and driver and protocol software that takes less than 10 years to write? I think the answer is "Yes. Maybe simpler than 2G." Or at least "Much more likely than with a 3G radio."