[aprssig] Fillin digi experiment (2nd comment)
bruninga at usna.edu
Sun Jan 17 10:09:24 CST 2010
> The results so far are very encouraging.
By what measure? Only from 10,000 feet up can you hear the whole network in the area that has an impact on packets.
> [using 15-second buffer-listen-first]
> .. my station almost never digipeats,
> because the wide digi's hear almost every
> beacon, but occasionally it does digipeat
> when ... I fail to hear a digipeat from
A commmon logical error. "Not-hearing" something on the APRS channel reveals -nothing- about whether it was there. It -only- reveals what your station DID hear. Nothing else.
> (not much harm done in the latter case,
Not necessarily true. It is adding a dupe to the channel during a slot when other user inputs are also trying to propogate. And it is unneeded in *most* cases, because we all know that *most* not-heard packets are not-heard because they collided, so they DID exist somewhere...
Now the key is WHERE they collided... that is the $64,000 question. Because a "collision" is an RF geographical overlap. The same two packets that collide in one small geographic area at the same instance DID-NOT collide in the other 90% of the coverage area of the two sending stations.
So the packets DID GET OUT (elsewhere), just at that instant, not HERE. So adding another dupe HERE may have what appears to be a BENEFIT, but only at the EXPENSE of all surrounding areas that now will see it as an out-of-slot DUPE that collides with other user traffic.
> ..and certainly no worse
> than a standard fillin digi).
Not really true. A standard FILL-IN digi does not add out-of-slot dupes to the channel. It does one and only one thing. It responds to first-hop WIDE1-1 packets only. There are two cases:
1) Only the FILL-IN hears the packet. It digipeats it. Success as designed.
2) The FIll-in and one-or-more surrounding digis hear the W1-1 packet. They ALL digipeat it properly at the next instant. One slot time is used by all digis, and the packet further propogates from all digis. Success as designed.
In all cases, only the one slot-time is used by all digis so network efficiency remains high.
> The concept should be self-adjusting:
> stay mostly silent while the wide digi(s)
> are operating but take over the role if
> they go off air.
AH, now that is a different concept. Then I would make the DECISION time on the order of 30 minutes. (And not buffering any packets either!). If the failure of a major network asset has been detected, then I agree, activate the digi, but NOT on a packet-by-packet basis which can only add QRM and out-of-slot dupes...
> The 15 second time period has been chosen
> to allow a reasonable time for a digipeat
> from the "big four"...
Just remember two fundamental APRS principles:
1) It is *impossible* to know anything about network success, collisiosn, or channel loading by "detecting decoded packets"
2) Only CARRIER SQUELCH monitoring at the big digi's can tell you information about collisions, and loading.
3) What you do-not-hear says nothing about what the other digi's heard and digipeated...
4) all digipeats at each tier(hop) are supposed to occur at the same instant (DWAIT=0) so that packets do not take more than the minimum number of slot times.
The good news is that the effects of #1, 2 and 3 can be mitigated somewhat by owner design of his RF system with antenna placement, and shielding so that each MAJOR digi or packet source in his area is heard with 13 dB or more difference between all such sources.
Otherwise, I see only disadvantages to any hold&digi-if-not-heard algorithms...
More information about the aprssig