What can we expect from an HF propagation model ?

In HF, reliability represents the average monthly time availability for each hourly computation. This is a statistical factor. Usually amateurs use a value of 50% reliability, even less in CW. 50% means that the mean monthly level for the specified circuit would be available during at least 15 days per month. Of course the higher the best.

This factor adding a limitation in the computation, the higher the reliability the more conservative, pessimistic is the estimation. Commercial circuits like the ones of VoA, BBC or RFI for example use a 90% reliability in order that all listeners receive their signals in excellent conditions. It means that the circuit would be available as predicted at least during 27 days per month. In a world coverage map, the more conservative coverage area is thus also smaller.

As with any statistics, data interpretation is the harder part of the simulation. Amateurs don't like to set a too high reliability because the estimation is too pessimistic; we all know that even when a simulation states that bands are closed from 15 to 10m, we have still to interpret properly this estimation. Concerning DX contacts, does it mean that stations located 2000 km away are also concerned by this "blackout" or only those located over 5000 km away ?

At left a forecast calculated by DX Toolbox for a 100 W PEP transmission from ON in summertime by 2200 UTC on 20 meters. The audio reports doesn't correctly match the prediction : signals from K are readable even if they are weak (they will be much stronger 1 to 2 hour later), FY is as strong as expected but worst, stations from LA or UT are predicted unreadable although they are strong, and VK8 stations (at Darwin, North of VK) using the short path, so-called unreadable, show a strong audio. DX ToolBox displays the signal strength but not the S/N required reliability, the power at receive station, etc, because the program makes rough assumptions and uses always the highest reliability. At right the point-to-point prediction established by VOACAP for the circuit Brussels (ON) to Darwin (VK8) at the same time. The SNR of 50 dB shows indeed that the 20-m band in open to VK near 2200 UTC with a S/N of 10 dB only, very weak. Approximatively the same SNR chart but weaker (SNR of 5 dB) was predicted between ON and FY. In the field the S-meter was strong for FY (S-6) and very weak for VK (S-1) but with a strong audio (equivalent to S-7 to S-9) in both cases. Both applications forecasted well some openings but not all, and both are wrong about some details (Signal strength to LA and UT for the first, and SNR to FY for the second model although its prediction for VK was correct). Of course one sample does not represent the overall performance of these programs.

In fact the simulation doesn't work like this. Only the propagation chart calculated for a band, a given date and time at the required reliability give you a global estimation of the propagation conditions at the earth scale. When you ask for a prediction for a specified path (e.g. ON to LA as displayed in the below charts) only the propagation conditions along this point-to-point circuit are taken into account, most of the time using median or highly reliable values. But sometimes, even using statistical values, conditions in the field can be quite different. The reason is related to the accuracy of models and the update of not of "smoothed" values of the solar and geomagnetic data.

Let's take an example, or rather seven examples. I worked Bodo, in Norway (67°N, 14°E) on August 8, 2004 at 12:00 UTC (SSN 29, SFI 86) using a G5RV dipole tight in the E-W direction. Here is its radiation pattern calculated with MultiProp (settings are 100 W PEP, beam heading 0°, 20-m band, QRM "rural", chart calculated for August 2004 with CCIR/Oslo Coefficients.

Most applications predicted that there was no opening to expect to LA above 14 or 15 MHz as show graphs displayed below. In retrospect, their accuracy was more than approximative. On the other side, most programs using the VOACAP engine, sometimes "updated" with real-time ionosonde data say all the contrary, and that was indeed confirmed in the field...

Here are forecasts calculated by seven applications recently published using, as far as possible, the same input parameters. Honor to the elder, W6ELPro which chart is displayed at right, predicts a MUF at 14.1 MHz in Norway.

In other charts (signal level prediction) W6ELPro predicts a field strenght in LA of 44 dB over 0.5 μV, thus S9+10. According to G4ILO's HFProp (below left), DX openings are not numerous at that time and mainly located to northern latitudes. The MUF is at 15.1 MHz at 12:00 UTC and the signal to LA is strong (meter below right in the green). However, it does not permit to get more information. For DX ToolBox (below center), the target location is in the silent zone with a MUF near 13 MHz and the signal estimated to 30 dB (cyan). Note that outside the skip distance, the field strengh reach about 25 dB. However its silent zone it at least 30% too wide.

Now let's see what predict the applications using the VOACAP engine. For WinCAP Wizard 3 (below right) using an SNRxx of 50%, thus not too optimistic and not too pessimistic either, the SNR is over 50 dB at 12 UTC and qualified as "fair" (good) in SSB, in spite of a strong field strength estimated at S7.5. The MUF is predicted at 13.7 MHz. This is the first application to provide a posteriori a valuable estimation.

For GeoAlert-Extreme Wizard (below left) the MUF ≤ 20 m or 14 MHz over Scandinavia. According to DXAtlas (below center), that uses real-time ionosonde data, the MUF is at 17.300 MHz over Bodo (mid of Norway). At last, for its companion, Ham CAP (below right) the SNR in LA is 42 dB, strong too.

Checking these predictions with real QSOs, I noticed that contacts were possible with all european countries from EA to LA with signals up to S-9, even on 28.5 MHz. On 20 m my reports from LA were indeed S9+, some S-units better than predicted.


From left to right, screen dumps from GeoAlert-Extreme Wizard, DXAtlas (with Ham CAP et IonoProbe), and Ham CAP (stand-alone) for a short circuit of 2000 km between Belgium and Norway, on August 8, 2004 at 12:00 UTC.

Generally speaking we can say that upper bands were close for DXing. But not entirely. Circuits to Spain or Norway for example shown a good S/N ratio over 50 dB. But predictions estimated the BUF limited to 10 MHz at noon and the MUF oscillating between 13 and... 17 MHz.

If I had to trust these predictions, theoretically I might work those countries up to 30, maybe 20m, but not higher. However, in the field there was some activity on 17, 15 and still more on the 10-m band where they were amateurs from EA and LA. Of course the 10-m band was not crowded and only a handful amateurs were on the air. And for DX stations upper bands were indeed closed excepting a good opening to South Africa and Madagascar forecasted by some applications.

Reliability must thus be used with care, and preferably in addition to other estimations of the ionospheric status and field strength (MUF, BUF, SNR, SNRxx, dBW, dB>mV, etc).

This analysis is thus a perfect example of the art to not always trust in charts or, if you like, to interpret them correctly when they use statistical data instead of real-time data...

As we told previously, amateur transmitters show an output power in a ±10 to 20 dB range (10-100-1000 W PEP). A good propagation program should be able to show the effect of power gain on the area coverage, and hopefully most recent programs take it into account. Chart and maps show those sensitivities over a ±10 dB range by using either a false-color pattern that highlights the increasing in coverage or simply put in light a larger area on the worldmap as shown below.

Like on the air, the increasing (or decreasing) in signal strength is more perceptible on the weakest signals. The WinCAP simulation shown below displays very well this effect, increasing the weakest signals (early morning and late in the evening) of 1.5 S-unit or 10 dBW.

Such simulations can thus also be used to show the effect of potential changes in the station's hardware (transmission line of lower loss, receiver of higher sensitivity, use of a linear amplifier, antenna of higher gain, etc, as many parameters took into account by VOACAP).

The effect of a power gain of 10 dB (100 W to 1 kW) on the propagation. The MUF/LUF is given for a path from ON to ZS. Above the results displayed in DX ToolBox, below in WinCAP Wizard 3. In the best conditions, the VOACAP model shows that the power gain does not increases the S/N ratio of strong signals (S/N > 10 dB or > 8 dBW) but well the weakest that jump from 8 to 18 dBW or 1.5 S-unit (S-5 to S-6.5). A higher solar flux (e.g. SFI of 200 instead of 86 in this case) does not change this effect.

Is the communication circuit between New York and London "electrically" speaking identical to the circuit from London to New York ? At first sight it is, except that we can already note that, at daytime, the position of the sun affects the receive station of a strong QRN. Relevant observation.

For decades, engineers and physicists considered that the ionospheric signal propagation was essentially the same in both directions along a given path. Excluding short-term changes in the ionosphere (warm up, sporadic plasma clouds, storm, aurora, etc), this is true within a few decibels for the upper bands and in a lesser extend for the low bands as well.

However, when we take into account SNR predictions, thus the signal quality, circuits are no more reciprocal. These variations are primarily due to different atmospheric noise levels at the receive site.

Propagation charts calculated by DX ToolBox for the 20-m band on August 2, 2004 22:00 UTC for paths from respectively New York to London (left) and the opposite (right). In both cases of course the number of hops is 3 as listed in the blue upper bar. Both propagation estimation maps are different because at that time the position of the Sun is simply not the same over both QTH. In theory they are also different because the S/N ratio at both terminals is affected by noise generated not only by the sun but also along the path by thunderstorms, aurora, and sometimes QRM emitted by large industrial cities. The S/I ratio can help to predict these effects. At last, both circuits have also different losses. However all these variables, excepting the sun position, are not taken into account by this application that gives only a rought estimation of the ionosphere status, at earth scale (intensity of field strength, MUF, etc).

In the late '90s ACE-HF, an U.S. company managed by Richard P.Buckner, was commissionned by the U.S. Navy to study propagation circuits within the Atlantic area.Dick is also author of a propagation program of the same name using the VOACAP engine. His conclusions showed a directional difference of as much as 12 dB in received S/N ratios, what represents a power ratio that exceeds a 2-factor ! The source of this reciprocity failure was determined to be the different atmospheric noise levels at either end of the circuits. It appeared that lightning flashes from thunderstorms that concentrated in the summertime in the Caribbean area generated an increasing of atmospheric noise levels, similar to what we experiment on the low bands from 40 m and up when a frontal system moves close to our receiver.

However, if the problem is well-known, very few propagation programs take all these effects into account. A model like VOACAP or ICEPAC for example is ionospheric-effects oriented, this is a point-to-point signal model that only incorporates the SSN, the date, time and the circuit properties (working conditions at both ends of the circuit, including the ground properties and QRM at target location). The atmospheric conditions are simply bypassed.

IRI on the contrary or down-sized versions like DXAID or DXAtlas, tries to incorporate various processes like the electron temperature, the auroral precipitation or the conductivity but it doesnt' take into account the atmospheric conditions yet; maybe for after tomorrow, when a future release of IRI-200x will be interfaced with the best weather model from the World Meteorological Organization. That day maybe, we could say that we master almost the atmosphere. Remain to master the geomagnetic dynamo, the sun and the space weather models and to link all that stuff in our super meta big "sun-earth plasma model"... A fine project without any doubt !

Example of reciprocity predicted with ICEPAC for a circuit between Belgium and Australia on August 14, 2004 (SSN 85). The SNR is set to 50 dB, SNR required reliability SNRxx to 50% for a power of 100 W PEP in both antennas. Simulations are respectively viewed from the receive location in VK (left using a 3-element Yagi and from the receive location in ON using a 31m long dipole (right). Variations do not exceed 5 dBW or half a S-unit. At the time of the QSO (22 UTC, local midnight in ON, 7 am in VK) the position of the sun and thus the ionization level of the F-layer was very different over each location. Thanks to accurate algorithms, ICEPAC predicted that signal power at both receive locations 'd be identical (-146.44 dBW or S2) and the SNR over +16 dB, thus very weak. On the air VK signal arrived indeed very weak (S-1) but with a strong audio (equivalent to S-7 or S-9).

Now that you master or almost the main parameters used in the VOACAP model, I suggest you to read my review of VOACAP to understand how powerful and more flexible it is compared to simpler prediction programs.