28 Dec 2015

More test results with WSJT-X V1.6

This came in earlier from Alan G8LCO. It seems that WSJT-X V1.6 is definitely the way to go.

Jan found out that his misleading results were due to having two programs on the same PC, now he sees the improvement on 1.60

David did a 8.5 Hour test and saw 23.7%  more spots on 1.60

I have posted a note of my test results on the LF group so the story is out now! The improvement is staggering!  And I do like the way that the facilities have been put together. The Hopping Scheduler is a very nice way of making clean band changes without the "Wrong Band" risk as well as being a great tool in it's own way. You also don't need to run a spec anl as well to see what is happening.
It seems to be a very nice package all around  with the JT modes as well. But it will take a bit longer to explore all of the possabilities.  Open Source software can really be fantastic when it is done well.

This was the very detailed analysis Alan G8LCO posted on the RSGB LF group:

WSPR is now included in the WSJT package as an alternate mode alongside JT65 and JT9. I loaded the free software just before Christmas running into some issues because I had not changed the audio settings. After changing to the settings in the instructons everything worked.

The WSPR mode use is a little different, band hopping is available so instead of manually having to change bands waiting for decoding to end etc etc you can now preset the next band(s) making the band change quick and certain. If every user picked up on this feature the "Wrong Band" issue which messes up the logs could be a past issue! that would be a very significant "WIN" for the WSPR database users!

The decoding now uses a two pass process, strong signals are processed then removed allowing weaker signals to be decoded. This makes a very substantial difference!

On Dec 27 I ran a comparison, I used my normal antenna and Rx to produce audio which then went to two seperate Rx's, a laptop running WSPR 2.12 which was a known good performer and another PC running WSJT-X 1.60. I chose to listen to 160m as it was busier than 620m at that time. There is also the prospect of TA spots.  The test was run overnight.

It was quickly evident that the two pass decoder was producing more spots than the stand alone WSPR 2.12
In some time slots 1.60 produced 6 decodes to 2.12's  3 decodes !  The main reason was that when two signals were close the old decoder only  found the stronger signal wheras the two pass decoder seemed to decode all of the signals visible on the spectral display. I had decodes of signals 2 Hz apart with 20dB amplitude differences whereas 2.12 lost weaker signals 5Hz apart. Some -10dB signals decoded  1 in 6 times on 2.12 but every time on 1.60. SNR's were broadly the same however a few signals decoded 1,2 or 3 dB higher on 1.60.

One feature of WSPR 2.12 is that high level signals seem to top out around +13dB so very strong signals peg at +13dB or so. The 1.60 software has produced SNR numbers up to +20dB so we can better see crocs blasting away. Hitherto I have run Spectrum Lab alongside WSPR so I am well aware of the levels of some stations!

For those used to 2.12 WSPR the new version takes a bit of getting used to as there are significant differences  and many new facilities that extend the way we can use the software.

The spectral display has seperate gain and bright controls with the waterfall having it's own controls. There are several pallets and display widths to select and an early decode option that displays the decode earlier than 2.12 does. There is also a 2 min "thermometer" style display at the lower edge indicating the time in the slot cycle  and a box giving the  progression of the four time periods in the Hopping Shedule, night, dawn grey line, day and sunset grey line. That I found very useful.

For me the package seems to work outstandingly well, decoding is better than ever and quicker, the spectral display and waterfall are flexible without over complication and the very clumsy old style band changing is now very quick, risk free and certain. The development people and the testers have done a very good job.

However decoding weak signals is not simple, different people have different equipment, noise levels and operating practices. I would advise against running different decode software at the same time on one machine, there is a possibility of misleading results. I would like to thank Jan, David and Roger for helping clarify matters and their independent testing.


1 comment:

Anonymous said...

The thing I like about the new band-hopping is that, all things being equal, everyone will be on the same band on the same minute of every hour. If I spot K9AN at 12 past the hour on 17m, we both hop to 15m next and I can try my chances with them there. If I don't enable 15m then the program randomizes one of my selected bands for that slot instead. Stations who only TX on 10m should expect to see greater chances on these 10m slots, which are 18,38 and 58 past the hour. One other bonus, I selected MF as RX only and actually have spotted 3 different WSPR stations here in the states. Usually that band is extremely noisy for me though.

73 KC1AWS Gary