I. INTRODUCTION
There are 10 Amateur Bands between 50 MHz and 10 GHz inclusive, with widely differing propagation characteristics. There are currently (as of 2-20-18) 5 slow JT modes and 3 fast JT modes, the latter allowing 15 second or shorter receive cycles. Although the program developers categorize FT8 as a slow mode and I have therefore included it in that category, it also uses 15 second receive cycles. Many of the modes have multiple sub-modes and if you add up all of those, there are currently 32 modes and sub-modes to choose from.
Furthermore, on a given band propagation might be via tropospheric scatter, via meteor scatter, via rainscatter, via aircraft scatter, via Aurora, or via EME, for example (the possibilities depend on the band). Each of these different propagation modes will have its own path attenuation and frequency dispersion/frequency shift characteristics.
Also, station equipment characteristics need to be considered when deciding on the best mode/sub-mode to use. The characteristics that are important to consider include the frequency stability of the sending and receiving stations, ERP, and the noise figure (sensitivity) of each station’s receive system.
Although the choice of mode/sub-mode for a given QSO attempt at first glance seems complicated, in fact in order to determine the optimal digital mode to use for a given QSO attempt one just needs to compare a few characteristics of each digital mode/sub-mode to the appropriate corresponding characteristics for the path/band/propagation mode to be used, and the best match will thereby be easily selected.
The mode/sub-mode characteristics that need to be considered in this process are:
1. Sensitivity of the mode (i.e., how weak a signal can it copy. Conventionally expressed as dB above the noise, or in the case of the WSJTX modes, in “JT dB”)
2. Tone spacing (separation between two adjacent signal tones, expressed in Hz)
3. Receive cycle time (total time used for signal reception, sandwiched between two transmit sequences)
4. Time required for a single complete message to be received. This is NOT necessarily the same as #3 above, because several modes include multiple sequential transmissions of the complete message in a single receive cycle. For example, for JT9H the time required to transmit the complete message is 0.425 seconds, even though the receive cycle time for this mode can be set to approximately 5, 10, 15, or 30 seconds.
The band/propagation mode characteristics that need to be considered are:
1. Signal-to-noise-ratio at the receiver
2 Frequency dispersion and/or Doppler shift caused by the scattering objects in the case of rainscatter, aircraft scatter, Aurora, EME
3. Time available to complete the QSO. This might be an hour or more for a non-contest experimental session, or as short as a couple of minutes for an aircraft scatter contact with an airplane flying perpendicular to the inter-station path
4. The time available to complete reception of a single complete message element. This ranges from a tiny fraction of a second for a meteor scatter path to the full receive cycle time for a stable tropo path.
The main station characteristics affecting mode/sub-mode selection are:
1. Effective radiated power
2. Receive system sensitivity
3. Frequency stability.
How does all of this come together?
1. The poorer the signal-to-noise ratio of the received signal at the receiving station, the more sensitive must be the chosen mode/sub-mode.
2. The greater the path frequency dispersion/Doppler shift/frequency instability of the transmit and receive stations, the greater the tone spacing will need to be.
3. Short receive cycle times permit more rapid QSOs, but mode sensitivity is reduced for shorter cycle times, all other things being equal.
4. With meteor scatter under usual non-shower conditions with only very short meteor pings available, only modes such as MSK144 with very short (0.020-0.072) message completion times are useful. Aircraft scatter has less stringent requirements in this regard, but relatively short message completion times are often needed for successful aircraft scatter contacts on the microwaves, where ISCAT-A, ISCAT-B, and JT9F, JT9G, and JT9H may therefore be needed.
The slow QSO modes as defined by the WSJT/WSJTX developers are FT8 (which has 15 second receive cycles), JT4, JT9A-H Slow, JT65, and QRA64. Except for FT8, each of these slow modes have one-minute-long (actually ~47-49 second) receive cycle times.
The fast QSO modes as defined by the WSJT/WSJTX developers are ISCAT-A, ISCAT-B, JT9E-Fast, JT9F-Fast, JT9G-Fast, JT9H-Fast, MSK144, and MSK144-sh. The JT9 fast sub-modes suffer significantly in terms of sensitivity when compared with the corresponding JT9 slow sub-modes, as you will see below. MSK-144 and MSK144-sh have a niche role for meteor scatter and will not be discussed further here.
II. Specific Mode/Sub-mode Characteristics
Now we are going to review the specific characteristics of the modes, limiting the discussion to what is important for operating in the frequency range of 50 MHz through 10 GHz, inclusive, and dropping MSK144 and MSK144-sh from the discussion. If you want more extensive information on the modes/sub-modes, you can start here:
First we will consider mode/sub-mode sensitivity. Note that within each of the modes mentioned above, as sub-mode tone spacing is increased the sub-mode sensitivity will decrease, all other things being equal. Also, reducing receive cycle time will reduce sensitivity by about 1.5 dB each time the receive cycle time is halved (10*log10(sqrt(2))).
To obtain the data in column one of the table below, I performed a MATLAB/Simulink simulation by adding incremental amounts of Gaussian white noise to WSJTX signals taken from the audio transmit output of WSJTX 1.9.0-rc1 for each of the modes/submodes analyzed here. Signal-to-noise ratios of ~5 to 50 dB were specified in the MATLAB code, with the added noise being gradually increased until the signal-to-noise ratio was so low that the signal could no longer be decoded. The signal-to-noise ratio specified in MATLAB below which decoding was no longer possible was recorded for each mode/submode, and the degradation of this value relative to the value of this parameter relative to JT9D for each mode/submode is given in column one of the table below. Positive numbers in column one indicate performance that is poorer than JT9D, and the larger the number is, the poorer the performance of that mode/sub-mode.
The WSJTX Manual given at the link shown above gives the sensitivity of JT9D in “JT dB” units as -24 dB. Using that value and adding the result given in column 1 for each mode/sub-mode, we can calculate a sensitivity in “JT dB” units for each mode and sub-mode. These values are listed in column 2 of the table below. The third column in the table below gives the tone spacing for each mode/sub-mode listed. The tone spacing values for each mode were taken from the Protocols section of the WSJTX Manual.
| Mode | Degradation from JT9D dB | Sensitivity JT dB | Tone Spacing Hz |
|---|---|---|---|
| JT9D | 0 | -24 | 13.89 |
| JT9E Slow | 0 | -24 | 27.78 |
| JT4C | 0 | -24 | 17.5 |
| FT8 | 3 | -21 | 6.25 |
| ISCAT-A 15 | 4.5 | -19.5 | 21.50 |
| JT9E-Fast 30 | 6.3 | -17.8 | 27.78 |
| ISCAT-B 15 | 6.5 | -17.5 | 43.10 |
| JT9E-Fast 15 | 8.5 | -15.5 | 27.70 |
Most of the time you will have only a qualitative sense of what the signal-to-noise ratio will likely be for a given path, unless you have worked that path before. If you have worked the path before, you will have a QUANTATIVE idea of what the signal-to-noise is likely to be, and what digital modes have worked for you on that path before. Acquiring this information is a GREAT reason to test things out with other stations “BEFORE the contest”, so that you can go right to the mode/sub-mode most likely to succeed for a given path and band when the contest comes. My program AircraftScatterSharp will also give estimates of the expected quantitative signal-to-noise values for both troposcatter and aircraft scatter propagation, expressed as dB above the noise. This is shown in the image below; the signal strength in dB above the noise is shown in the row labeled “Marg” ( short for “Margin”). You can read more about AircraftScatterSharp here:
http://www.nitehawk.com/w3sz/NEW_W3SZ_AircraftScatterSharp2017.pdf
I didn’t list the other slow JT modes/sub-modes in the table above, because (unlike the fast modes) their sensitivities are listed in the WSJTX manual and don’t need to be determined experimentally. The WSJTX manual specifies that the slow modes JT65A-C all have sensitivity of -25 dB. JT4A is specified to have sensitivity -23, JT4B -22, JT9A -27, JT9B -26, and JT9C -25 dB (JT units). I tested the JT4 submodes C, D, E, F, and G even though those sensitivites are given in the manual as a check on the correctness of my experimental methods. My experimental results for the JT4 submodes’ sensitivites are within 1 dB of the results listed in the WSJTX manual.
It is also important to choose the submode with the optimal tone spacing for the path. Picking a mode/sub-mode with tone spacing less than the frequency excursion/dispersion of the path will result in degraded decoding. So choosing a mode with tone spacing at least as great as the frequency dispersion/shift that will be seen during the receive period is important. On the other hand, choosing a mode/sub-mode with greater tone spacing than is necessary will result in reduced sensitivity, thus needlessly throwing away dBs of signal to noise ratio.
Tone spacings for JT4 vary from 4.375 Hz for JT4A to 315 Hz for JT4G. The tone spacing doubles with each increment, except that JT4C has 17.5 Hz spacing and JT4D has 39.375 Hz spacing.
Tone spacings for JT9 vary from 1.736 Hz for JT9A through 222.222 Hz for JT9H. The tone spacing doubles with each increment. For the FAST JT9 sub-modes (but NOT the Slow sub-modes) the data rate increases so that the time required for a single complete message to be received halves with each increment in tone spacing, going from a time of 3.4 seconds for JT9E-Fast to 0.425 seconds for JT9H-Fast. There are no differences in the tone spacing between the corresponding JT9 fast and slow sub modes.
So what you need to do to pick the best mode and sub-mode for a particular QSO is:
1. Decide whether you need the extra sensitivity of the slow modes like JT4, JT9, or even JT65. Pick (a) either one of those slow modes or (b) FT8 or one of the fast modes based on how much sensitivity you need. Refer to the table above for a hierarchy of mode/sub-mode sensitivity, and make your selection accordingly. For a path with stronger signals, you should be able to get by with any of the modes listed (considering only signal strength), and can use the modes nearer the bottom of the table to speed up the time to completion of the QSO. However, for weak signals, you will need to stick with the modes near the top of the table.
2. Decide how wide your tone spacing needs to be based on system frequency instability, path frequency dispersion due to Aurora, rainscatter, etc., and Doppler shift if doing EME or Aircraft Sctatter. Match your expected frequency dispersion/Doppler shift/system frequency instability value with a mode/sub-mode that has compatible tone spacing. The tone spacing of the optimal mode/sub-mode needs to be at least as large as the total of your expected frequency shift/dispersion parameters.
That is all there is to it! Pick the mode and sub-mode based on the sensitivity and frequency tolerance that you think you need based on the band, propagation path/mode, and equipment characteristics as outlined above. If you have sufficient signal strengths to potentially use FT8 or one of the fast modes, you can reduce your QSO time to 25% of what it would be with the slow modes. You can always try a fast mode (or FT8) first and then drop back to a slow mode if you don’t have enough oomph to complete the QSO with one of the fast modes.
For example,if you are in a hurried contest situation and you want to do QSOs with 15 second receive cycle times, then you are limited to FT8, ISCAT-A, ISCAT-B, and JT9E-Fast through FJ9H-Fast. Of these choices, FT8 has the best sensitivity, so if the given path/mode frequency dispersion and transmit-receive instability of the system (including both the transmitting and receiving stations’ instabilities) is less than 6.25 Hz over the 12.6 second receive period, then FT8 is likely your best 15-second-T/R-cycle choice. If your system frequency instability is worse than that, then you should pick another mode instead of FT8, using the tables above as your guide. For the greatest chance of success, always choose the mode that has the greatest sensitivity that will also satisfy the receive cycle duration and frequency spread characteristics of your particular situation.
The best mode for a quiet-condition 10 GHz terrestrial QSO using troposcatter is NOT necessarily the best mode for a rainscatter or Aurora or EME 10 GHz QSO. The optimal terrestrial mode/sub-mode might also be the “right” mode/submode for aircraft scatter if the airplane is flying right down the direct path between two stations so that Doppler shift is negligible, but it will NOT be the right mode/sub-mode if the aircraft’s path is at a significant angle to the direct path so that there is significant Doppler shift of the received signal. You can play with various aircraft/path geometries for various frequencies using my program AircraftScatterSharp and see what Doppler shifts result from various aircraft scatter geometries at various frequencies, or review the table on page 23 of my recent paper on Aircraft Scatter, given at the NEWS Conference in 2017. These resources can be found at:
http://www.nitehawk.com/w3sz/AircraftScatter.htm
A pdf of the paper is here:
http://www.nitehawk.com/w3sz/NEW_W3SZ_AircraftScatterNEWS_2017_Paper.pdf
III. A Few Examples
There is of course much more frequency dispersion when rainscatter is the propagation mode than there is with a “quiet” tropo path. In January, 2018 while helping the NN3Q/R team get setup prior to that month’s contest, we happened to be out in the rover van during a rainstorm. Dave, K1RZ, was kind enough to get on the air so that we could try a rainscatter digital contact. I expected that there would be a lot of dispersion as there was no direct path to Dave, and the entire signal would be scattered. So I chose JT4G, as it has the largest available tone spacing, 315 Hz. It has a specified sensitivity of -17 dB, so by choosing the extremely wide tone spacing, we were giving up 4 dB relative to JT4C and 6 dB relative to JT4A. We were able to complete the contact, with Dave’s signal levels -16 in the NN3Q/R rover van and our signal report from Dave at -20 dB. The figure below shows the WSJT-X main window for this contact.
As you can see from the figure below, the frequency dispersion during this contact was on the order of 50 Hz or slightly less, and there was no appreciable drift or Doppler, so we could have used JT4E (78.5 Hz tone spacing, -19 dB sensitivity) or JT9F (55.56 Hz tone spacing, -22 dB sensitivity) if signals had been too weak for us to complete the QSO using JT4G (315 Hz tone spacing, -17 dB sensitivity), and we likely would have been successful due to the greater sensitivity of JT4E and JT4F compared with JT4G. Fortunately, we had enough signal strength even with JT4G and so this was not an issue.
Based on the fact that, as noted above, the fast JT9 modes have a 6-9 dB disadvantage relative to the slow JT4 and JT9 modes, we likely would not have been able to complete the QSO using one of the JT9 fast modes. Although FT8 and ISCAT-A have better sensitivity than the JT9 fast modes, the frequency dispersion on the 10 GHz rainscatter path was too wide for FT8 (6.25 Hz tone spacing) and ISCAT-A (21.5 Hz tone spacing). ISCAT-B had borderline tone spacing for this path (43.1 Hz) and in addition ISCAT-B has similar sensitivity problems as the JT9 fast modes, with a sensitivity of -17.5 dB.