Here's some data from to illustrate and expand on the "key point":
2018 WW RTTY 2018 FT8 Roundup
# Logs: 3,063 1,253
Dupes: 0.7% 3.0%
Busted Exchange: 0.7 1.0
Busted Call: 1.0 0.0
NIL: 1.8 5.1
Total Error Rate: 3.5 6.1
With the "perfect copy, or no copy" nature of the FT modes, one would expect
busted exchanges and calls to both be zero. The relatively high rate of
exchange busts in the FT8 Roundup seems due to serial numbers being off by
one between the sent and received logs. That is an aberration that needs to
be understood and addressed as a separate issue. This should not be the
case in WW Digi without serial numbers.
Note that the NIL rate went up significantly in the FT8 Roundup, largely for
the reason Rich describes below. This should come down with time and
experience as we all learn techniques and messaging to lower the ambiguity
of whether a QSO was completed or not.
The higher dupe rate in the FT8 Roundup is likely due to the lack of duping
capability in most of the software being used in that contest. Again, that
should improve with time. Both N1MM+ and WriteLog support WW Digi with the
full dupe-checking capability already built in. The challenge is for
participants to properly setup their logger with WSJT-X (N1MM+) or DigiRite
(WriteLog).
Ed W0YK
-----Original Message-----
From: CQ-Contest <cq-contest-bounces@contesting.com> On Behalf Of Richard
Ferch
Sent: 31 July, 2019 09:03
To: cq-contest@contesting.com
Subject: Re: [CQ-Contest] WWDIGI Contest Testing...
It seems to me that a lot of this discussion is missing a key point. The
problem is not what to do when you have 100% copy; the default protocol
takes care of that. It's what to do when copy is poor/intermittent/marginal:
CQ WZ1ZZZ FN31
WZ1ZZZ AA6AAA DM04
AA6AAA WZ1ZZZ R FN31
So far so good. At this point, AA6AAA sends either WZ1ZZZ AA6AAA RR73 if
his software copies the exchange, or WZ1ZZZ AA6AAA DM04 (repeats the
call message) if it does not copy the exchange from WZ1ZZZ. All of this is
pretty much automatic.
Suppose AA6AAA sends the RR73 message. His software auto-logs the contact
and as far as he is concerned the contact has been completed. But what if
WZ1ZZZ's software does not copy that RR73 message? He doesn't know which of
the two messages AA6AAA sent. His software automatically repeats the R FN31
message (the contest exchange), and is now waiting for the RR73 from AA6AAA
before it will finish the auto-sequence and log the contact. Meanwhile
AA6AAA has already logged the contact and may have gone on and started a
new contact with someone else.
If WZ1ZZZ copies AA6AAA sending to someone else in the next time slot, he
might conclude that he just missed the RR73 and log the contact manually,
but he might instead conclude that the contact was incomplete and not log
it (especially if he is depending on auto-logging). On the other hand, if
WZ1ZZZ does not copy anything from AA6AAA in the next time slot, he has no
information on which to base a decision whether to log the contact or not,
and possibly will not log it, resulting in a NIL for AA6AAA.
Is this different from analog modes? Qualitatively, no - you can come up
with similar scenarios. But in terms of the likelihood of the undesired
outcome (one side logs the contact and the other does not), I think the
probabilities are different. In analog modes, copy is rarely all or
nothing. WZ1ZZZ can usually copy enough of AA6AAA's second message to give
him an idea whether it was the exchange or a repeat request. In FT8/FT4,
copy is all or nothing - there is no partial copy to give the operator a
clue. In analog modes, if copy is too rough to base a decision, there is
the possibility of asking AGN? For practical purposes, this doesn't really
exist in the rigidly structured FT8/FT4 sequence. I believe the end result
is likely to be more unresolved ambiguous situations, leading to more
situations where one station logs the contact but the other station does
not.
My guess is that the error rates in WWDIGI are going to be quite different
from the contests we are used to. The rates for bad call signs and bad
exchanges should be virtually zero (the software doesn't generally miscopy
call signs or exchanges, it either gets them right or not at all), but I
believe the rates for NIL contacts will be quite a lot higher than in
traditional modes. There will be a temptation to blame this on contest
newbies showing up in the new mode, but IMO there are structural reasons
for expecting high rates of NILs.
73,
Rich VE3KI
_______________________________________________
CQ-Contest mailing list
CQ-Contest@contesting.com
http://lists.contesting.com/mailman/listinfo/cq-contest
_______________________________________________
CQ-Contest mailing list
CQ-Contest@contesting.com
http://lists.contesting.com/mailman/listinfo/cq-contest
|