Re: [44net] hardware vs. software

"Michael E. Fox - N6MEF" <n6mef@...>


56kbps (and above) AX.25 can operate without FEC.  

Not in the real world.  The required S/N and phase error levels required to achieve BER in the 10^-8 range at those speeds is simply not present at many/most sites. 


It is a matter of path loss and modem BER.  


Of course.  But what’s needed is not achievable to any practical degree at real world sites – at least not around here (Silicon Valley).  We did a lot of testing of AX.25 at 9600.   We tried several radios, several TNCs, measured BER, phase noise, deviation, etc.  In ideal locations, it worked fine.  But in real-world, everyday locations, we found that the increased number of retransmits caused by single bit errors in a packet size of 128 bytes to 256 bytes far, far outweighed the increase in baud rate from 1200 to 9600.  And yes, we set deviation carefully with calibrated service monitors.  Even the Kenwood D710 manual tells you that the received signal needs to be full scale on the meter or else 9600 isn’t going to work well.  And even with a strong enough signal, we have to worry about multi-path issues, particularly in urban environments. 


Our modems are designed to minimize BER and one may find that FEC is not required for a given network.

On the other hand, adding a protocol agnostic FEC in the modem is a possibility, exchanging raw bit rate for data correction.  We are interested in this capability and will be looking at use cases for it.


Well, if you consider the overall goal of getting as much throughput as possible between two sites, then even if the trade-off was 25%  reduction of the 56Kbps bandwidth in order to get 10^-8 BER, that’s still an enormous improvement over 9600 with no correction.  Seems like a no-brainer.   Add FEC:  I’ll take 8 please.  (Oh, right, pre-orders are closed.)  ;-)





Join to automatically receive all group messages.