Navtex/DSC decoder performance
Hi!
I’m very curious to know how well the internal Navtex/DSC decoders provided in the KiwiSDR compares with external specialist navtex/DSC decoders. Has anyone done comparison tests or have any other means of comparison? I suspect performance differences would be most apparent under weak signal or high static conditions.
Can anyone point me in the right direction.
Thanks a lot
Steve
Editor Medium Wave News
Comments
I can't speak to relative Navtex/DSC performance directly. But I can tell you about the FSK decoder on which they are based.
The current Kiwi FSK decoder is based on some old open source code I found a long time ago. Same for the CW decoder. But they both are known to have only modest performance.
I've been looking at porting the open source Fldigi code which includes FSK and CW decoders and many other ham digital modes (MFSK etc). But this will be a big task and I don't know when it might occur. Or how much Navtex/DSC decoding might improve. I guess one way you could estimate that is to run Fldigi on a second Kiwi channel via a virtual audio cable and compare the two.
I now have some overnight tests comparing the KiwiSDr using its internal JNX decoder versus an RSPdx+YaDD external decoder.
The latter generated 30-50% more decodes in an overnight run from the same off-air RF signal.
Interestingly it is possible to see a fair number DSC transmissions in the waterfall that generate no internal decode.
It has been suggested to me that the performance is the result of an overworked Beaglebone, rather than differences in the RF signal chain.
In my humble opinion upgrading the internal decoder is probably not necessary because good external decoders exist which will be used by serious listeners of Navtex & DSC in part because of their integration with online databases.
Yes, this doesn't surprise me. The intent of the Kiwi extensions was always to get something working quickly that had reasonable performance, starting with whatever open source code I could find. It's the novelty of finding everything in one place, ready to go, with no (or minimal) configuration.
Interestingly, the FSK (hence NAVTEX, DSC and Selcall) decoders actually run in Javascript on the browser. So the Beagle limitations don't play that big a factor. I notice the NAVTEX decoder does pretty well on weak MF channels (e.g. 490, 518 kHz). So perhaps I'm being too aggressive in rejecting messages if there are errors (you can see the detected errors by setting the display menu to "+errs").
If/when I port Fldigi a new FSK decoder won't be that much work. And the high-level NAVTEX, DSC and Selcall code will pretty much stay as is.
Hi Steve,
Interesting. I understand it was a quick test, but I don't think your comparison is a level playing field, as you are using a separate receiver, as the RF characteristics, AGC, RX bandwidth etc. may be different.
Apples vs. Oranges.
As a result, upgrading the KiWi decoder may not bring about as significant an improvement as you wish, if it is also an RF limitation.
I'd suggest, as John originally commented, that you pipe the audio from the KiWi to your external decoder and then make the comparison.
Apples vs. Apples.
Just a passing observation, apologies for the multiple analogies. I've not used my quota and have a backlog to get through :-)
Regards,
Martin
Hi Martin
Agreed about your apples & oranges comment. However I've just got results from a test last night that used one KiwiSDr set to decode on 2187.5kHz. Simultaneously the audio output was fed to YaDD.
That meant the antenna, the RF chain and A-D convertor were common to both decoders. The only difference would be in the digital domain. I think both paths would be constrained by the same passbandwidth of 270Hz in the Kiwi. (Isn't that maybe a bit narrow?)
The external YaDD decoder decoded 186 messages from coast stations versus 176 decoded by JNX internally in the Kiwi (7 hour overnight run).
My preliminary conclusion: “YaDD external to the Kiwi is the superior decoder for DSC signals but the internal decoder in the Kiwi isn’t bad”
(One cannot compare this test run with my earlier test - different days, different frequencies, different transmitters; almost certainly different levels of noises, multipath distortion or selective fading).
I've tried searching to see if anyone has done robust side by side tests of different digital decoders for DSC & Navtex and I cannot find any. For example does a commercial decoder have any advantage over an opensource piece of software?
regards
Steve
Hi Steve,
OK, thanks for doing that test.
I think it provides a bit more confidence that the KiWi isn't doing too badly.
You could always drag out the KiWi receiver passband if you felt it was a bit too narrow.
I guess the commercial decoders would be more resilient, as everything can be designed and optimised for the one single purpose.
Many years ago, a radio friend helped design the original ICS Navtext receiver. But I see that their product range has moved on since then.
But at more than £350 GBP I'd certainly hope it outperformed a KiWi :-)
Regards,
Martin