Phil,
[…] but even then the transmitter would only have to move one or
two SSB channels, and then only on each transmission. I don't see this as a
problem for any modern radio.
+ Indeed, at present the transmission frequency programmed into the
transceiver often is offset from the receive frequency programmed into the
transceiver, in order to keep the transmitted signal within the transmit
bandpass (in an ICOM 7610, 1500-2000 Hz, but YMMV). Moving the Tx frequency
independently from the Rx frequency in every cycle is not a new thing; a wider
FT8 allocation would only increase the maximum offset, from ~2 kHz to something
larger, perhaps 8-10 kHz.
If you need 5 transmissions to get through, and then you do with
a high SNR, the cause could be either a series of collisions or a very rapid
propagation improvement. Note that you don't need to look only at the
transmissions addressed to you; you can look at the SNRs of the same stations
talking to someone else (or calling CQ). And you can look at those sent after
your own QSO.
+ If I call a DX station with a consistently strong SNR, and he consistently
works others, after perhaps five such QSOs I QSY to a different df, as I
suspect that my frequency is not clear at the DX receiver. (I may choose, for
example, and not necessarily in this order, the df of the station he just
worked, since he can certainly hear at least something there; a particularly
high or particularly low frequency, since these regions of the spectrum are
typically less densely occupied; a mid-band frequency that seems clear at my
receiver, in case the DX station has a somewhat narrow bandpass at his
receiver; a frequency occupied by a strong local or domestic station, on the
same cycle phase as I am, successfully working the same area of the world as
the DX station -- although I am careful to decode the local from time to time
to watch for signs that I am interfering with him; a frequency adjacent to the
DX transmit frequency, as I have found several DX stations that apparently
operate with extremely narrow audio filters; and, if all else fails, directly
on the DX Tx frequency, as there seems to be an unexpectedly large number of
people in the world who will only reply to callers on their own df.) A QSY is
not performed unless the transmission immediately prior is decoded and, of
course, is not calling me; I typically time my QSYs to occur after the first DX
transmission to a new caller. To one listening, this sequence of Tx channels
may certainly seem random and in my darker, more frustrated moments I have
suspected that a random sequence may, in fact, be superior.
+ If the DX station is weak, at the threshold of detection, so that occasional
messages are missed, I don’t know that my operating is any different. I still
require a decode of the first DX transmission to a new caller before I QSY;
since I may miss a few such messages the average QSY rate is somewhat slower,
but as far as I can remember nothing is different.
+ It is also true that there is high-level timing associated with FT8 channel
collisions: Since a completed QSO takes four to six cycle durations (15s each)
to complete, interference typically ends after 4n to 6n cycle durations (60n
seconds to 90n seconds), where n is a low integer equal to the number of QSOs
made by the interfering station. (Here we ignore CQs and resent QSO messages.)
That is, the duration of FT8 QSOs is fixed (discounting resent messages); if
your signal is not clear at the DX receiver, it may be clear after a QSO
duration, when the interferer completes his QSO.
+ The reciprocal case, interference to the DX station at my receiver, in my
experience is nearly always observable in the Rx Frequency window, where one
will observe other stations, often with SNRs 20 dB or more higher than the
desired DX. (Two-pass decoding isn’t perfect.) When the interfering station
stops, successful decoding of the DX station returns.
+ I’ve run across QSB and other rapidly-varying periodic channel impairments
(e.g., Doppler spreading on polar paths), but outside of 6m and 160m, and
possibly 80m, they have had periods much longer than a typical FT8 QSO. (I am
omitting short-term, aperiodic transient phenomena like meteors and lightning
QRN.) For this reason I believe most temporary loss-of-signal conditions on
80m-10m are due to collisions, not the vagaries of HF propagation which, I
believe, have longer time scales. But I agree, it would be nice if some
measured data existed somewhere and am quite willing to change my opinion in
the face of data. 😉
I point out that there may actually be a more serious "collision"
problem at a higher protocol level. If you call CQ, and then five stations
answer you, you can pick only one to answer. The other stations may keep
calling you, especially if they can't hear the station you're actually working,
and they may collide and/or cause QRM to other users.
+ I’m not sure I follow your explanation of this “piling-on” problem. If I
reply to Station A, other stations may continue to call me, but it’s not
because they can or cannot hear the station I’m actually working. After all,
if they’re calling me they’re transmitting at the same time as Station A, so
they’ll never hear him anyway. In my experience, stations that are decoding me
well will wait until I send a message containing “73” (either “RR73” or “73”),
and then make a tail-end call, since they know that that is their signal that
I’ve finished the present QSO – and that it’s pointless to call me until the
present QSO is finished, anyway. Stations not decoding me well often continue
to call through the present QSO (since due to missed message decodes they may
be unsure of where I am in the QSO), which I actually don’t mind, since they
frequently have a poor SNR at my receiver, too, and having multiple
opportunities to decode a weak DX station calling me I view as a Good Thing™.
I agree that it’s an opportunity for generating QRM to other users, but so is
the traditional analog DX pileup. Besides, if a DX station is generating that
much activity I would submit that it’s better for all concerned for the DX
station to move to his own RF frequency, away from the FT8 “watering holes,”
and use the Fox & Hound DX mode. Until that point is reached, however, I’ve
never found a shortage of stations to work in the scenario you describe and so
I wonder if it’s truly a problem to solve – or if I’ve misunderstood you.
The pileup problem could also be addressed at a higher level by
automatically answering each calling station in turn, or even better by
answering them all *at the same time*, i.e., with multiple transmit waveforms
on separate frequencies.
+ Multiple transmit waveforms on separate frequencies is actually already
employed in the FT8 Fox
<https://physics.princeton.edu/pulsar/k1jt/FT8_DXpedition_Mode.pdf> & Hound
(DXpedition) mode. It’s useful to a point, but its use can be controversial,
since transmit power is split among the transmit frequencies. The power drops
6 dB per signal split, so even going from one frequency to two can drop the SNR
on each below the detection threshold on the low bands. If used improperly its
effect is to enable the DX to work all the Big Guns, or all the locals,
quickly, while freezing out the Little Pistols and those on the other side of
the world, who were counting on FT8 and the brief sunrise opening to work a
really Rare One. Speaking from personal experience, it can be really, really
frustrating to copy a CQ from a rare DX station (because all transmit power is
on one FT8 frequency, and the SNR is above the decode threshold), reply to the
station, and then not hear anything else because the power split among the
separate frequencies means that none of the frequencies have an SNR above the
decode threshold.
I assume congestion is much lower at VHF […]
+ Indeed. It’s sometimes difficult to remember that FT8 was originally
designed for summer 6m multi-hop sporadic-E. It seems to be a physical law
that the most popular protocols see their greatest success outside their
intended use. IEEE Std 802.11 was for file transfer, not real-time video;
Bluetooth was to replace the PC-to-printer cable, not phone calls . . .
Ed N4II.
From: Phil Karn via wsjt-devel <wsjt-devel@lists.sourceforge.net>
Sent: Thursday, September 2, 2021 2:17 PM
To: wsjt-devel@lists.sourceforge.net
Cc: Phil Karn <k...@ka9q.net>
Subject: Re: [wsjt-devel] Idea for "frequency hopping" FT8 to reduce collisions
On 9/2/21 2:26 AM, alan2--- via wsjt-devel wrote:
Hi, I've been following this thread with interest and have a few
questions/comments please:
If the proposed protocol frequency changes are set to use the wider bandwidth
of a SDR receiver linked to a T/R switch with a standard voice rig as Tx as I
saw in one post, those frequency changes will presumably need to use CAT
control. Will that be reliably fast or stable enough with such relatively
frequent changes, for all rigs?
You would not need this to implement my original idea, which is to hop each
transmission *within* a single SSB bandwidth. The "hopping" would be done in
the transmit waveform generation software, and it is only within the range that
the receiver already demodulates everything anyway.
The only concern would be if the per-transmission hopping is performed over a
new, wider FT8 allocation that requires a SSB radio to be retuned beyond a
single SSB bandwidth. It would work best with a separate receive SDR able to
demodulate the entire thing at once, but even then the transmitter would only
have to move one or two SSB channels, and then only on each transmission. I
don't see this as a problem for any modern radio. Whether it's a problem for
very high-Q antennas I'll leave to the antenna experts. How many antennas, and
for what bands, have to be retuned for a change of 3 kHz?
Is there a way of getting some indication of how many collisions are currently
occurring in any user session, firstly to try and obtain some real world data
on how big and frequent the issue might be, and secondly if possible what the
decode conditions were? I base that on the vagaries of HF propagation that I
suspect might be the principal controlling factor and of course are entirely
unpredictable.
That's a *very* good question. We can infer some of this by looking at receive
SNRs when many transmissions were needed. If you need 5 transmissions to get
through, and then you do with a high SNR, the cause could be either a series of
collisions or a very rapid propagation improvement. Note that you don't need to
look only at the transmissions addressed to you; you can look at the SNRs of
the same stations talking to someone else (or calling CQ). And you can look at
those sent after your own QSO.
I point out that there may actually be a more serious "collision" problem at a
higher protocol level. If you call CQ, and then five stations answer you, you
can pick only one to answer. The other stations may keep calling you,
especially if they can't hear the station you're actually working, and they may
collide and/or cause QRM to other users. Because they get no response from you,
they may also think they're colliding when you're just ignoring them. They
might even increase power (a bad idea). Per-transmission hopping will
definitely reduce the pile-on effect that causes so much QRM when a
sought-after station calls CQ. Remember, he'll hear you no matter what
frequency you use within the SSB bandwidth. Think of my idea as something like
automated, random split frequency operation. (There's a lot of wisdom and
experience in conventional ham analog operation. We should use as much of it as
possible).
The pileup problem could also be addressed at a higher level by automatically
answering each calling station in turn, or even better by answering them all
*at the same time*, i.e., with multiple transmit waveforms on separate
frequencies. This latter approach would require a *great* deal of care to avoid
intermodulation distortion as the transmitter would no longer see a single tone
with a 1:1 peak-to-average power ratio. I would only do this in a SDR
transmitter carefully designed for the task. There's already too much
distortion in the common setup using analog audio between computer and
transmitter.
Are collisions an issue at VHF and above where propagation is different?
If this gets implemented it would be good to have it switchable in and out, so
users who are interested can compare what's happening.
I assume congestion is much lower at VHF, but repeated collisions on the same
frequency can still be a problem if you answer on the caller's frequency,
especially during a contest. So there's still a strong case to be made for
randomizing the frequency of each transmission. And of course it should be a
selectable option.
Phil
_______________________________________________
wsjt-devel mailing list
wsjt-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wsjt-devel
