Bill:

Thanks for the explanation of the technical backgrounds for FST4 and FST4W.

The transmitting station frequency stability is also critical as the net frequency stability between Tx and Rx station is the determinant of drift.

One thing I have to clarify is that my measurement was mostly about the transmitter side drift, not on the receiver side. I assumed Airspy HF+ had no drift at all, which might have been incorrect. Airspy HF+ is practically far more stable than FT-891, however, and the drifting pattern observed in my measurement did not correlate with the Airspy HF+ frequency drift, so I guess the 1.5Hz transmitting drift of 10.1355MHz for FT-891 was not that inaccurate.

73 de Kenji Rikitake, JJ1BDX

Alan and all:

I did a quick measurement of an unmodulated carrier frequency drift of FT-891 ("this transmitter"), with the following conditions:

• FT-891 uses RTTY mode without changing the keying signal (i.e., the output is a continuous carrier)
• FT-891 output: 5W
• Output carrier frequency: 10.1355MHz
• Receiving system for measuring the frequency: Airspy HF+ Dual Port, gqrx 2.11.5, and fldigi-4.1.13 frequency analysis mode
• Two cases were measured, ~5-minute transmission each, with ~5-minute waiting period

Summary of the result:

• The 5-minute frequency drift was ~1.5Hz (~0.148ppm), which explains why the transmitted signal of FST4W-300 by this transmitter on the 30m band was not able to decode.
• First 2-minute frequency drift was ~1.2Hz (in the 2nd transmission case), which explains why WSPR on the 30m band works OK on this transmitter.

Comments:

• I've tested FST4W-120 (2 minutes) on 10.1389MHz + 1536Hz with this transmitter and it worked OK.
• Extrapolating the result into the 80m and the 40m bands suggest that for the 80m band the estimated drift (~0.53Hz) is tolerable but for the 40m the estimated drift (~1.04Hz) is not tolerable, which was consistent with the result of my previous experiment.

The attached files are the PNG file to show the measurement data, the raw measurement data, R command to display the graph.

73 de Kenji Rikitake, JJ1BDX

ft891-10mhz-freq-error.png

Hi Kenji san,

thanks for your thorough receiver stability analysis. Some other points are relevant here:

• The transmitting station frequency stability is also critical as the net frequency stability between Tx and Rx station is the determinant of drift.
• The WSPR decoder has frequency drift compensation and can correct for linear drift, the FST4 and FST4W decoder does not have this ability. It was not developed because characterizing drift, and compensation of it, reduces the decoder sensitivity or increases the probability for false decodes.
• Doppler spreading on the propagation path(s) has a similar effect to frequency instability. The larger the Doppler spreading; the more signal energy spreads into adjacent tone frequencies. Doppler spreading increases with frequency so FST4 and FST4W will perform better, particularly for the longer T/R period variants, on MF and LF bands where we measure Doppler spreading as low as a 1/1000 Hertz on unimpeded Oceanic paths. On HF bands between 0.1 and 1  Hertz are more likely.
  

It would be reasonable to expect Polar paths or propagation during periods of raised A index to contribute to a reduction in sensitivity of these narrow tone spaced modes.

The FST4 and FST4W decoder has a tool to measure estimated Doppler spreading on successful decodes, to enable it start WSJT-X from a directory that contains a file called plotspec. Estimated Doppler spreading is appended to decoded messages.

Note this was a direct digital loopback connection and the spreading measured was probably due to reduced audio resolution as -45dB digital attenuation was applied.

Here is a more realistic decode from a simulated FST4W-120 signal:

which was generated as follows:

C:\WSJT\wsjtx\bin\fst4sim.exe "K1JT FN22 10" 120 1495 0.4 1 1.0 0.75 1 -25 T

Message: K1JT FN22 10                          W: T iwspr: 1
f0: 1495.000   DT:  0.40   hmod:     1   TxT: 109.3   SNR: -25.0

Message bits:
00001001101111111001101110010101000010100100001100000100100111000010000111000 000000000000000000000000

Channel symbols:
0132102300
3132223132
3111003310
0200103123
1032012003
0123131022
3113223331
2231210132
1023303022
1303322301
0333000112
3101231032
0111323332
0221232133
3003321223
2301321023

   1   0.40 1495.00  -25.0  000000_0001.wav

to simulate a possible somewhat disturbed HF channel.

73
Bill
G4WJS.

Alan and all:

I did a quick measurement of an unmodulated carrier frequency drift of FT-891 ("this transmitter"), with the following conditions:

• FT-891 uses RTTY mode without changing the keying signal (i.e., the output is a continuous carrier)
• FT-891 output: 5W
• Output carrier frequency: 10.1355MHz
• Receiving system for measuring the frequency: Airspy HF+ Dual Port, gqrx 2.11.5, and fldigi-4.1.13 frequency analysis mode
• Two cases were measured, ~5-minute transmission each, with ~5-minute waiting period

Summary of the result:

• The 5-minute frequency drift was ~1.5Hz (~0.148ppm), which explains why the transmitted signal of FST4W-300 by this transmitter on the 30m band was not able to decode.
• First 2-minute frequency drift was ~1.2Hz (in the 2nd transmission case), which explains why WSPR on the 30m band works OK on this transmitter.

Comments:

• I've tested FST4W-120 (2 minutes) on 10.1389MHz + 1536Hz with this transmitter and it worked OK.
• Extrapolating the result into the 80m and the 40m bands suggest that for the 80m band the estimated drift (~0.53Hz) is tolerable but for the 40m the estimated drift (~1.04Hz) is not tolerable, which was consistent with the result of my previous experiment.

The attached files are the PNG file to show the measurement data, the raw measurement data, R command to display the graph.

73 de Kenji Rikitake, JJ1BDX

Kenji

Testing conditions:
* The distance between TX and RX was about 10 meters

This does seem close, if both were connected to dummy loads there would still be ample signal pickup.

Result of FST4W-300:
* On 7.0388MHz + 1536Hz: undecodable
* On 10.1389MHz + 1536Hz: undecodable
Frequency drift of FT-891?

With a recommended maximum of 0.56Hz for -300 this is quite possible.

The HF+ is quite stable in the short term when warmed up. Using software like Spectrum Lab you can monitor frequency very precisely.

I am not convinced there is any need to use FST4W on the higher frequencies but there may be times that there is an advantage.

73 Alan G4ZFQ

Testing conditions:

• Transmitter: Yaesu FT-891 output 5W with SCU-17 and WSJT-X 2.3.0-rc1 on Ubuntu 20.04 with a mobile whip
• Receiver: Airspy HF+ Dual Port + gqrx 2.11 and WSJT-X 2.3.0-rc1 on macOS 10.14.6 with a YouLoop magnetic loop antenna
• The distance between TX and RX was about 10 meters

Result of FST4W-300:

• On 3.5688MHz + 1536Hz: OK
• On 7.0388MHz + 1536Hz: undecodable
• On 10.1389MHz + 1536Hz: undecodable

Any thoughts on why the signals of 40m and 30m were not decodable? Frequency drift of FT-891?

73 de Kenji Rikitake, JJ1BDX