Announcement

Collapse
No announcement yet.

FR24 - Understanding submitted statistics (2nd try)

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • FR24 - Understanding submitted statistics (2nd try)

    Hi,

    Having had my last post fall foul of the anti spam filters and such, I'm trying a re-post.

    I've recently installed FR24 on a rpi, and am having difficulty understanding the statistics page.
    I've attached a snip from the 26th and 27th, on the 27th seems like the hits to positions was about 10%, but on the 26th was more like 50%, on both days there are positions shown on the polar plot in the 200NM and >200NM (max range show as 84nm on the 26th and 70nm on the 27th), but the histogram doesn't show these.
    On both days, dump1080 direct seemed to show lots of planes (well relative I guess) and no difference, but the FR stats show a very different story between the 2 days - Is this normal ?

    (I won't be editing any typo's this time round btw)

    Snip26.PNG
    snip27.PNG

    Thanks in advance,

    Mack-Jsy

  • #2
    If you go back to the thread it still exists. I approved both but no-one has replied

    Sent from my XT1092 using Tapatalk
    Posts not to be taken as official support representation - Just a helpful uploader who tinkers

    Comment


    • #3
      At least, I tried to. Every time it rescans it re flags.

      General rule..don't use @s. For email addresses only
      Posts not to be taken as official support representation - Just a helpful uploader who tinkers

      Comment


      • #4
        Cheers Oblivian, the at symbol is well and truly off my keyboard for fr24 stuff...

        You may as well delete that msg tbh, as it's doomed I say..... doomed...

        TIA

        Comment


        • #5
          Originally posted by Mack-Jsy View Post
          Hi,

          I've recently installed FR24 on a rpi, and am having difficulty understanding the statistics page.
          I've attached a snip from the 26th and 27th, on the 27th seems like the hits to positions was about 10%, but on the 26th was more like 50%, on both days there are positions shown on the polar plot in the 200NM and >200NM (max range show as 84nm on the 26th and 70nm on the 27th), but the histogram doesn't show these.
          I collect statistics directly from my dump1090 instance using Munin, but I haven't bothered to sum the 5 minute reports to see whether plane count and position reports tally with the statistics page.

          I have noticed, however, that my max range reported by Munin is consistently and substantially longer than that reported by the statistics page. I haven't seen anybody confirm this nor does fr24feed say what data it reports exactly, but I surmise that FR24 does some sort of filtering to discard poor quality data eg single-point tracks so that chance distance messages don't distort the rankings or give dodgy feeders an unfair advantage over those with better quality radars, and perhaps to encourage the former to improve their station by whatever means they can.

          Comment


          • #6
            Originally posted by Strix technica View Post
            I collect statistics directly from my dump1090 instance using Munin, but I haven't bothered to sum the 5 minute reports to see whether plane count and position reports tally with the statistics page.
            I'm guessing I need to look at installing dump1090-mutability rather than the out of the box FR24 image (and then using your plugins that I've found on github

            That's todays project (last weeks project was a cantenna 1090 !)

            Comment


            • #7
              Originally posted by Strix technica View Post
              but I surmise that FR24 does some sort of filtering to discard poor quality data eg single-point tracks so that chance distance messages don't distort the rankings or give dodgy feeders an unfair advantage over those with better quality radars, and perhaps to encourage the former to improve their station by whatever means they can.
              Believe so. The sampling and upload is done in packets, so one can only surmise it is analysed as part of this.

              A lot of local-side applications don't 'drop' contact for a number of seconds/minutes after loss and can predict track (much the same way the FR24 website can between switching receivers and patchy areas). Likewise you could be getting 2 'blips' every 5-10 seconds (ADSB is generally 2tx packert/second from the ADSB Boxes) and still get a complete distance track and such plot. When infact your solid signal dropped outter edges some seconds earlier while the data was in between being bundled up and sent to FR

              So I suspect when you get to the next upload it goes.. 'well, I only saw it 3x in 4 seconds then. It must be out of range or bung'
              Posts not to be taken as official support representation - Just a helpful uploader who tinkers

              Comment


              • #8
                Originally posted by Mack-Jsy View Post
                I'm guessing I need to look at installing dump1090-mutability rather than the out of the box FR24 image (and then using your plugins that I've found on github

                That's todays project (last weeks project was a cantenna 1090 !)
                I certainly can't recommend Munin highly enough. As a sys-admin, it has saved my butt on more than one occasion, and I've even been able to diagnose hosting company hardware faults before they were even aware of them (despite being in a VM container).

                I've written Munin plugins for all sorts of things from system monitoring to intrusion attempt monitoring to traffic/vhost monitoring. So it was only natural that I should write some for dump1090, especially when -mutability so conveniently packages 5 minute data in JSON files

                Good luck, and have fun.

                Comment


                • #9
                  Originally posted by Oblivian View Post
                  The sampling and upload is done in packets, so one can only surmise it is analysed as part of this.
                  It's also possible that fr24feed does its own filtering before uploading data to fr24's infrastructure. It'd certainly make sense to use local processing power where possible.

                  Originally posted by Oblivian View Post
                  (ADSB is generally 2tx packert/second from the ADSB Boxes)
                  One positional and one velocity report, yes, but that doesn't account for additional messages that are responses to ground and TCAS interrogation messages transmitted on 1030 MHz.

                  Originally posted by Oblivian View Post
                  When infact your solid signal dropped outter edges some seconds earlier while the data was in between being bundled up and sent to FR
                  That's a distinct possibily I'd not considered!

                  Comment


                  • #10
                    Originally posted by Strix technica View Post
                    I certainly can't recommend Munin highly enough.

                    Good luck, and have fun.
                    Rebuilt from a stock image with mutability and have got Munin up and running with your plugins, so now I can start looking at where and what I need to do next - maybe consider an improved aerial (FlightAware 1090MHz ADS-B N-Type Antenna 26" - 5.5dBi) as although my cantenna is better than the stock aerial with the dvbt stick, not sure it's that great lol

                    Comment


                    • #11
                      Originally posted by Mack-Jsy View Post
                      Rebuilt from a stock image with mutability and have got Munin up and running with your plugins
                      Cool, sounds like it was painless enough.

                      If you haven't already, you might want to edit /etc/default/dump1090-mutability and set your lat/lon and also

                      Code:
                      EXTRA_ARGS="--phase-enhance --oversample --aggressive --mlat"
                      --mlat is required to pass message timestamps from dump1090 if your fr24feed has MLAT turned on. Without that flag, fr24feed won't complain, but it probably won't provide any useful MLAT data either. The other flags are extras peculiar to -mutability that may improve your message rates a little.

                      Originally posted by Mack-Jsy View Post
                      maybe consider an improved aerial (FlightAware 1090MHz ADS-B N-Type Antenna 26" - 5.5dBi) as although my cantenna is better than the stock aerial with the dvbt stick, not sure it's that great lol
                      Cantennas aren't bad and they're quick and easy, but there are definitely better options out there. I got pretty good results from building my own 8-segment colinear relative to a 1/4w whip or a 1/2w dipole (which is effectively what a cantenna is to a/c low to the horizon). They're a little tedious to build and require accuracy and careful attention to detail, but they're cheap (just a metre or so of good quality RG-6).

                      I varied the described design slightly; didn't bother with the weird 1/4w shorted last segment (didn't seem to improve things any), plus I added an inverted sleeve balun using F-connectors and F-couplers (so the first element is only 1/4w (adjusted for the coax's velocity factor) to the tin can, plus the can cut to 1/4w in free air. I have no idea how effective the balun is, but it doesn't seem to hurt things.

                      I haven't tried the FA antenna. The problem with whip antennas is that they require a mount with a decent RF ground to be properly effective, whereas the colinear doesn't.
                      Last edited by Strix technica; 2017-07-29, 16:11.

                      Comment

                      Working...
                      X