-6th July 2020
+9th July 2020
+-------------
+
+IF YOU ARE ON THE MASTER BRANCH, STOP (I REPEAT) STOP READING THIS
+DOCUMENT AND READ UPGRADE.mojo FIRST.
+
+
+If you are not on an existing 'mojo' branch, or you have a build less
+than 260, please stop reading as well and read UPGRADE.mojo (in the
+section about installing packages), as you need to install some extra
+packages before restarting the new version of the code. If in doubt
+try installing the packages again as this will either confirm that all
+is already done or will install what's missing.
+
+Assuming you have done the above:
The latest release of the Mojo branch of DXSpider contains a client
for the Reverse Beacon Network (RBN). This is not a simple client, it
attempts to make some sense of the 10s of 1000s of "spots" that the
RBN can send PER HOUR. At busy times, actually nearly all the time, the
-spots from the RBN come in too fast for anybody to get anything more
+spots from the RBN come in too quickly for anybody to get anything more
than a fleeting impression of what's coming in.
Something has to try to make this manageable - which is what I have
The RBN has a number of problems (apart from the overwhelming quantity
of data that it sends):
-* Spotted callsigns, especially on CW, and not reliably
+* Spotted callsigns, especially on CW, are not reliably
decoded. Estimates vary as to how bad it is but, as far as I can
tell, even these estimates are unreliable!
* The format of the comment is not regular. If one has both FTx and
"all the other" spots (CW, PSK et al) enabled at the same time,
one's eye is constantly having to re-adjust. Again, very difficult
- to deal with on contest days.
-
+ to deal with on contest days. Especially if it mixed in with
+ "normal" spots.
-So what have I done about this:
-
-* As you can see, there are frequently more than one spotter for a
-callsign:
+So what have I done about this? Look at the sample of input traffic
+below:
05Jul2020@22:59:31 (chan) <- I SK0MMR DX de KM3T-2-#: 14100.0 CS3B CW 24 dB 22 WPM NCDXF B 2259Z
05Jul2020@22:59:31 (chan) <- I SK0MMR DX de KM3T-2-#: 28263.9 AB8Z/B CW 15 dB 18 WPM BEACON 2259Z
05Jul2020@22:59:40 (chan) <- I SK0MMR DX de HB9DCO-#: 7018.2 RW1M CW 25 dB 18 WPM CQ 2259Z
05Jul2020@22:59:40 (chan) <- I SK0MMR DX de EA5WU-#: 7018.3 RW1M CW 19 dB 18 WPM CQ 2259Z
+* As you can see, there are frequently more than one spotter for a
+ callsign:
+
* I normalise the frequency and cache up to 9 copies from different
-spots. In order to do this I have to wait a few (comfigurable) seconds
-for the client to collect a reasonable number of copies. More copies
-may come in after 9 copies have been. Once I have enough copies to be
-sure that the callsign is at least agreeed upon by more than one
-skimmer, or the wait timer goes off, I emit a spot. An example of
-which is shown above in the spot sent to G1TST. By this means I can
-reduce the number of spots sent to a node user by up to a factor of 10
-for CW etc spots and about 8 for FTx spots.
+ spots. In order to do this I have to wait a few (configurable) seconds
+ for the client to collect a reasonable number of copies. More copies
+ may come in after 9 copies have been received. Once I have enough
+ copies to be sure that the callsign is at least agreeed upon by more
+ than one skimmer, or the wait timer goes off, I emit a spot. By this
+ means I can reduce the number of spots sent to a node user by up to a
+ factor of around 10 for CW etc spots and about 8 for FTx spots.
+
+ For example, from the trace above, all the RW1M RBN spots become just
+ one line:
+
+DX de F8DGY-#: 7018.3 RW1M CW 23dB Q:9* Z:20 16 2259Z 14
* No RBN spots can leak out of the node to the general cluster. Each
node that wants to use the RBN *must* establish their own
connections to the RBN.
* Currently no RBN spots are stored. This may well change but how and
- where these spots are stored is not yet decided. Only "new" spots
- will be stored (if they are).
+ where these spots are stored is not yet decided. Only "DXSpider
+ curated" spots (like the example above) will be stored (if/when they
+ are). Sh/dx will be suitably modified if storage happens.
-* There are some things that need to be said:
+* There are some things that need to be explained:
a) The input format from the RBN is not the same as format emitted by
the cluster node. This is part of the unhelpfulness to mixing a raw
from different skimmers and, in this case, they did not agree on the
frequency (7018.2 - 7018.4) which is indicated by a '*'. The frequency
shown is the majority decision. If this station has been active for
-some time and he is still calling CQ after some time (configurable,
-but currently 60 minutes) and gaps for QSOs or tea breaks are ignored,
-then a '+' character will be added.
+for a while and he is still calling CQ after some time (configurable,
+but currently 60 minutes), ignoring gaps for QSOs or tea breaks, then
+a '+' character will be added.
+
+If the "Qualitee" Q:1 is seen on a CW spot, then only one skimmer has
+seen that spot and the callsign *could* be wrong, but frequently, if
+it is wrong, it is more obvious than the example below. But if Q is
+Q:2 and above, then the callsign is much more likely to be correct.
+
+DX de DJ9IE-#: 14034.9 UN7BBD CW 4dB Q:5*+ 17 1444Z 14
+DX de OL7M-#: 14037.9 UA6LQ CW 13dB Q:7 16 1448Z 15
+DX de LZ3CB-#: 28050.2 DL4HRM CW 7dB Q:1 14 1448Z 20
+
+Having said that, I check every spots call with the standard DXSpider
+is_callsign routine, to check that the callign is a valid format. Then
+through the Prefix::search routine, to check that the prefix actually
+exists. If either of these checks fails then the spot is ignored.
c) I ditch the WPM and the 'CQ' as not being hugely relevant.
-d) If there is a Z:nn[,mm...] is there it means that this call was also heard
-in CQ Zone 20. There can a ',' separated list of as many zones as
-there the space available (and this spot call was heard by :-). You
-will notice the spot zone and skimmer call zone around the time. This
-can be activated with a 'set/dxcq' command. This is completely
-optional.
-
-e) I shorten the skimmer callsign to 6 characters, before (re-)adding
-'-#' on the end to minimise the movement rightwards as in the incoming
-spot from KO7SS-7-# just two lines below G1TST. There are some very
-strange skimmer callsigns.
-
-f) I have a filter set (accept/spot by_zone 14 and not zone 14 or zone
-14 and not by_zone 14) which will give me the first spot that either
-spot or skimmer is in zone 14 but the other isn't. For those of us
-that are bad at zones (like me) sh/dxcq is your friend. You can have
-separate filters just for RBN spots if you want something different to
-your spot filters. Use acc/rbn or rej/rbn. NB: these will completely
-override your spot filters for RBN spots. Obviously "real" spots will
-will continue to use the spot filter(s).
+d) If there is a Z:nn[,mm...], then this spot was also heard by
+skimmers in other zones. In this example, it means that this call was
+also heard in CQ Zone 20. This list does NOT include the cq zone of
+the skimmer nor the spot. If you would like to see these then do
+'set/dxcq'. This setting is active for all the examples in this
+document. This is completely optional.
+
+There can be a ',' separated list of as many zones where this spot was
+also heard by another skimmers, up to the space available in the
+comment area.
+
+DX de LZ4UX-#: 14015.5 ON7TQ CW 6dB Q:9 Z:5,14,15,40 14 0646Z 20
+DX de VE7CC-#: 3573.0 N8ADO FT8 -14dB Q:4 Z:4,5 4 0647Z 3
+DX de DM7EE-#: 14027.5 R1AC CW 9dB Q:9* Z:5,15,17,20 16 0643Z 14
+DX de WE9V-#: 7074.0 EA7ALL FT8 -9dB Q:2+ Z:5 14 0641Z 4
+
+e) I shorten the skimmer callsign to 6 characters - having first
+chopped off any SSIDs, spurious /xxx strings from the end, leaving
+just the base callsign, before (re-)adding '-#' on the end. This is
+done to minimise the misalignment of the spot rightwards, as in the
+incoming skimmer spot from KO7SS-7-# below. There are some very
+strange skimmer callsigns with all sorts of spurious endings, all of
+which I attempt to reduce to the base callsign. Some skimmer base
+callsigns still might be shortened for display purposes. Things like
+'3V/K5WEM' won't fit in six characters but the whole base callsign is
+used for zone info internally, but only the first 6 characters are
+displayed in any spot.
+
+05Jul2020@22:59:39 (chan) <- I SK0MMR DX de HB9JCB-#: 3516.9 RA1AFT CW 9 dB 26 WPM CQ 2259Z
+05Jul2020@22:59:39 (chan) <- I SK0MMR DX de KO7SS-7-#: 14057.6 K7GT CW 6 dB 21 WPM CQ 2259Z
+05Jul2020@22:59:39 (chan) <- I SK0MMR DX de K9LC-#: 28169.9 VA3XCD/B CW 9 dB 10 WPM BEACON 2259Z
+
+f) I happen to have a filter set (accept/spot by_zone 14 and not zone
+14 or zone 14 and not by_zone 14) which will give me the first spot
+that either spot or skimmer is in zone 14 but its companion isn't. For
+those of us that are bad at zones (like me) sh/dxcq is your
+friend. You can have separate filters just for RBN spots if you want
+something different to your spot filters. Use acc/rbn or rej/rbn. NB:
+these will completely override your spot filters for RBN
+spots. Obviously "real" spots will will continue to use the spot
+filter(s). If you use set/dxcq, then unset/dxitu and unset/dxgrid
+first. You only need to this once.
g) If there is NO filter in operation, then the skimmer spot with the
LOWEST signal strength will be shown. This implies that if any extra
-zone are shown then the signal will be higher.
+zones are shown, then the signal will be higher in those zones.
h) A filter can further drastically reduce the output sent to the
user. As this STATS line shows:
So how do you go about using this:
-First you need to create an RBN user. Now you can use any call you
+First you need to create one or two RBN user(s). Now you can use any call you
like and it won't be visible outside of the node. I call mine SK0MMR
-and SK1MMR.
+and SK1MMR. One of these connects to the "standard" RBN port that
+outputs CW, BEACON, DXF, PSK and RTTY spots, and the other connects to
+the RBN port that just outputs FT4 and FT8 spots.
set/rbn sk0mmr sk1mmr
connect telnet telnet.reversebeacon.net 7001
'call:' '<node callsign here'
-RBN port 7000 is the "traditional" port for anything except FT4 or FT8
-spots. They come from RBN port 7001.
-
Now put them in your local crontab in /spider/local_cmd/crontab:
* * * * * start_connect('sk0mmr') unless connected('sk0mmr')
* * * * * start_connect('sk1mmr') unless connected('sk1mmr')
This will check once every minute to see if each RBN connection is
-active, you can check with the 'links' command:
+active, you can check what is connected with the 'links' command:
Ave Obs Ping Next Filters
Callsign Type Started Uptime RTT Count Int. Ping Iso? In Out PC92? Address
SK0MMR RBN 5-Jul-2020 1722Z 7h 6m 8s 0 0 198.137.202.75
SK1MMR RBN 5-Jul-2020 1722Z 7h 6m 8s 0 0 198.137.202.75
-The connections are sometimes dropped or become stuck, I have a
-mechanism to detect this and it will disconnect that connection and
-the normal reconnection will happen just as any other (normal) node.
+The RBN connections are sometimes dropped or become stuck. I have a
+mechanism to detect this and it will disconnect that RBN connection
+and, by this means,it will be automatically reconnected by the
+crontab, just like any other (normal) node.
-It is put in the crontab, rather than started immediately, to prevent
-race conditions (or just slow them down to one disconnection a
-minute).
+I use the crontab, rather than restarting immediately after
+disconnection, to prevent race conditions (or just slow them down to
+one disconnection a minute).
-The first time a connection is made, after node startup, there is a 5
+The first time a connection is made, after node startup, there is a 5 (configurable)
minute pause before RBN spots come out for users. This is done to fill
up (or "train") the cache. Otherwise the users will be overwhelmed by
spots - it slows down reasonably quickly - but experiment shows that 5
globally, for all RBN connections, but in future is likely to be
configurable per connection. Basically, because the FTx RBN data is
much more bursty and there is more of it (except on CW contests), it
-could do with a somewhat longer training period.
+could do with a somewhat longer training period than the CW etc RBN
+connection.
If a connection drops and reconnects. There is no delay or extra
-training time.
+training time. If the node is stopped and restarted within a few
+minutes, then no training time will occur.
For users. At the moment. There is a single command that sets or
unsets ALL RBN spot sorts:
-set/wantrbn
-unset/wantrbn
-
-Very soon this will be replaced with a '(un)set/skimmer' command that
-allow the user to choose which categories they want. Filtering can be
-used in conjunction with this proposed command to further refine
-output.
+Users can select which type(s) of RBN/Skimmer spot they require using
+the 'set/skimmer' command. Users can enter the 'help set/skimmer'
+command for more information on this. They can also enter then 'help
+rbn' command for more information about how to interpret what DXSpider
+produces and some other useful background information.
This still very much "work in progress" and will be subject to
change. But I am grateful to the feedback I have received, so far,
Mike G8TIC
Lee VE7CC
-But if you have comments, suggestions and brickbats please email me or
-the support list.
+But if you have comments, suggestions or even brickbats, please email
+me or the support list.
Dirk G1TLH
projects / spider.git / blobdiff