Having read about WSJT-X version 2.3.0 with the new FST4 protocols, curiosity drove me to perform some tests.
In the past, I wrote about FT8 and the fact that its sensitivity is considerably less than JT65 or JT9, leading to lost DX opportunities. Especially on 60 metres, the restricted power levels pose an additional challenge and in my case, reception is further impaired considerably by man made noise, estimated to be around 10 to 15 dB worse than in a quiet rural environment.
FST4-60 (1 minute periods) is said to be about 1.7 dB more sensitive than JT9 and a priori decoding can add a few dB [1]. In theory, a decoding threshold around -30 dB could be expected. In case FST4 would be adopted, the lost dB’s (FT8 versus JT65/JT9) could be reclaimed. With slower modes, even weaker signals might be decoded correctly.
The first test was set up to compare FST4 with JT9 and JT65. I ran JTDX and WSJT 2.3.0 in parallel, driving the IC7600 at low power and made sure that the two simultaneous transmitted carriers were equal, using a HP spectrum analyser to confirm the power levels.
I selected a frequency on 60 metres, just below 5357 and used web SDR’s for reception. The transmitted power level was adjusted to obtain reception near the decoding threshold. Test were done with the Twenthe Web SDR (grid JO32) and the Hackgreen Web SDR (grid IO83). I transmitted beacon messages.
FST4 looks indeed to be superior to JT9 and even JT65. Both JT9 and JT65 were getting unreliable below about -23 dB, whilst FST4 was still decoding around -25 dB.
It has to be said that the audio from the SDR’s is not perfect with some dropouts, so establishing the real world performance requires a real world approach.
Further testing
The next test will be to use a setup with a local receiver, connected to the transmitter via the transverter output, via a step attenuator to set the received power level. I will also ‘contaminate’ reception with local man made noise from the 60 m antenna.This setup was used in the past to confirm the reported SNR to be able to estimate the true received power from DX stations. .
Although the local transmission path is constant (no fading), I should be able to confirm the indicated SNR thresholds and compare performance with real world noise.
Results second test
Instead of hooking up the whole bunch, I chose to do a “dry” test, with 2 instances of WSJT-X, connected via a virtual audio cable. A software noise generator was used to insert white noise. After some fiddling, I got close to the thresholds and compared various modes. It has to be emphasised that the performance is less in the real world situation, because of fading and interference.
This is the list of thresholds obtained with the “dry” test:
SNR Mode -31 FST4-120 (120 sec periods) -28 FST4-60 (60 sec periods) -27 JT9 -26 JT65 with JTDX -25 JT65 with WSJT-X -24 FST4-30 (30 sec periods) -20 FT8 with JTDX -19 FT9 with WSJT-X -19 FST4-15 (15 sec periods)
The list shows that the slower FST4 modes are the better performers. The values mentioned in [1] could be confirmed, except the 15 second mode, which is worse than indicated. FT8 performs about a dB better with JTDX (I did not take the SNR reading from the software but derived the values from the software noise generator settings).
Real world testing with the slow FST4 mode is on the wish list now. It is also interesting to see the decoder performance with a bunch of signals in the selected passband.
With regard to FST4-15: if I were to choose, I would leave it out, because it performs worse than FT8 and occupies more bandwidth. I do not think that FST4-30 will be my favourite. Some decodes were missing and although it is about 4 dB better than FT8, the benefit is marginal.
If FST4-60 works well in the real world situation, it could claim its position in the list of options for real weak signals. Less than 0.1 mW RF power was more than adequate to decode my signal via the Maasbree Web SDR via skywave!
Will be continued!
[1] See: FST4 Quick start guide, page 1 bottom.