RBN.mojo

From The DXSpider Documentation Wiki
Revision as of 18:24, 14 February 2023 by EA3CV (talk | contribs) (Created page with "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 wi...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

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 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 tried to do with DXSpider's RBN client.

The RBN has a number of problems (apart from the overwhelming quantity of data that it sends):

  • 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 frequency given is unreliable. I have seen differences as great as 600hz on CW spots.
  • There is far too much (in my view) useless information in each spot - even if one had time to read, decode and understand it before the spot has scrolled off the top of the screen.
  • 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. Especially if it mixed in with "normal" spots.

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@22:59:31 (chan) <- I SK0MMR DX de LZ3CB-#:   7018.20  RW1M           CW    10 dB  18 WPM  CQ      2259Z
05Jul2020@22:59:31 (chan) <- I SK0MMR DX de W9XG-#:    14057.6  K7GT           CW     7 dB  21 WPM  CQ      2259Z
05Jul2020@22:59:31 (chan) <- I SK0MMR DX de G0LUJ-#:   14100.1  CS3B           CW    18 dB  20 WPM  NCDXF B 2259Z
05Jul2020@22:59:32 (chan) <- I SK0MMR DX de LZ4UX-#:    7018.3  RW1M           CW    13 dB  18 WPM  CQ      2259Z
05Jul2020@22:59:32 (chan) <- I SK0MMR DX de LZ4AE-#:    7018.3  RW1M           CW    28 dB  18 WPM  CQ      2259Z
05Jul2020@22:59:32 (chan) <- I SK0MMR DX de W1NT-6-#:  28222.9  N1NSP/B        CW     5 dB  15 WPM  BEACON  2259Z
05Jul2020@22:59:32 (chan) <- I SK0MMR DX de W1NT-6-#:  28297.0  NS9RC          CW     4 dB  13 WPM  BEACON  2259Z
05Jul2020@22:59:32 (chan) <- I SK0MMR DX de F8DGY-#:    7018.2  RW1M           CW    23 dB  18 WPM  CQ      2259Z
05Jul2020@22:59:33 (chan) <- I SK0MMR DX de 9A1CIG-#:  7018.30  RW1M           CW    20 dB  18 WPM  CQ      2259Z
05Jul2020@22:59:33 (chan) <- I SK0MMR DX de LZ7AA-#:    7018.3  RW1M           CW    16 dB  18 WPM  CQ      2259Z
05Jul2020@22:59:33 (chan) <- I SK0MMR DX de DK9IP-#:    7018.2  RW1M           CW    21 dB  18 WPM  CQ      2259Z
05Jul2020@22:59:33 (chan) <- I SK0MMR DX de WE9V-#:    10118.0  N5JCB          CW    15 dB  10 WPM  CQ      2259Z
05Jul2020@22:59:34 (chan) <- I SK0MMR DX de DJ9IE-#:    7028.0  PT7KM          CW    15 dB  10 WPM  CQ      2259Z
05Jul2020@22:59:34 (chan) <- I SK0MMR DX de DJ9IE-#:    7018.3  RW1M           CW    31 dB  18 WPM  CQ      2259Z
05Jul2020@22:59:34 (chan) <- I SK0MMR DX de DD5XX-#:    7018.3  RW1M           CW    21 dB  18 WPM  CQ      2259Z
05Jul2020@22:59:34 (chan) <- I SK0MMR DX de DE1LON-#:  14025.5  EI5JF          CW    13 dB  19 WPM  CQ      2259Z
05Jul2020@22:59:34 (chan) <- I SK0MMR DX de DE1LON-#:   7018.3  RW1M           CW    24 dB  18 WPM  CQ      2259Z
05Jul2020@22:59:34 (chan) <- I SK0MMR DX de ON6ZQ-#:    7018.3  RW1M           CW    22 dB  18 WPM  CQ      2259Z
05Jul2020@22:59:34 (chan) <- I SK0MMR DX de OH6BG-#:    3516.9  RA1AFT         CW    15 dB  25 WPM  CQ      2259Z
05Jul2020@22:59:35 (chan) <- I SK0MMR DX de HA1VHF-#:   7018.3  RW1M           CW    30 dB  18 WPM  CQ      2259Z
05Jul2020@22:59:35 (chan) <- I SK0MMR DX de F6IIT-#:    7018.4  RW1M           CW    32 dB  18 WPM  CQ      2259Z
05Jul2020@22:59:36 (chan) <- I SK0MMR DX de HB9BXE-#:   7018.3  RW1M           CW    23 dB  18 WPM  CQ      2259Z
05Jul2020@22:59:37 (chan) <- I SK0MMR DX de SM0IHR-#:   7018.3  RW1M           CW    21 dB  18 WPM  CQ      2259Z
05Jul2020@22:59:37 (chan) <- I SK0MMR DX de DK0TE-#:    7018.3  RW1M           CW    26 dB  18 WPM  CQ      2259Z
05Jul2020@22:59:37 (chan) <- I SK0MMR DX de OE9GHV-#:   7018.3  RW1M           CW    40 dB  19 WPM  CQ      2259Z
05Jul2020@22:59:37 (chan) <- I SK0MMR DX de CX6VM-#:   10118.0  N5JCB          CW    20 dB  10 WPM  CQ      2259Z
05Jul2020@22:59:37 (chan) -> D G1TST DX de F8DGY-#:     7018.3 RW1M         CW  23dB Q:9* Z:20           16 2259Z 14
05Jul2020@22:59:38 (chan) <- I SK0MMR DX de HB9JCB-#:   7018.3  RW1M           CW    16 dB  18 WPM  CQ      2259Z
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
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 (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 "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 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 RBN feed with normal spots.

b) Each spot sent out to a node user has a "Qwalitee" marker, In this case Q:9*. The '9' means that I have received 9 copies of this spot 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 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...], 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 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: 

23:22:45 (*) RBN:STATS hour SK0MMR raw: 5826 sent: 555 delivered: 70 users: 1

For this hour, I received 5826 raw spots from the CW etc RBN, which produced 555 possible spots, which my filter reduced to 70 that were actually delivered to G1TST. For the FTx RBN, I don't have a filter active and so I got all the possibles: 

23:22:45 (*) RBN:STATS hour SK1MMR raw: 13354 sent: 1745 delivered: 1745 users: 1

So how do you go about using this:

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. 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

Now create connect scripts in /spider/connect/sk0mmr (and similarly sk1mmr). They look like this:

/spider/connect/sk0mmr: 

connect telnet telnet.reversebeacon.net 7000
'call:' '<node callsign here'

/spider/connect/sk1mmr: 

connect telnet telnet.reversebeacon.net 7001
'call:' '<node callsign here'

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 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
    GB7DJK DXSP  5-Jul-2020 1722Z     7h 6m 8s   0.02   2    300    89               Y    163.172.11.79
    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 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.

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 (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 minutes is a reasonable compromise. The delay is configurable, 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 than the CW etc RBN connection.

If a connection drops and reconnects. There is no delay or extra 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:

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, from:

Kin EA3CV Andy G4PIQ Mike G8TIC Lee VE7CC

But if you have comments, suggestions or even brickbats, please email me or the support list.

Dirk G1TLH

Retrieved from ‘https://wiki.dxcluster.org/index.php?title=RBN.mojo&oldid=102’