What can we expect from a HF propagation model ?

Requisit and specification of amateur programs (III)

What can be expect now from amateur programs ? Light propagation prediction programs like Ham CAP, G4ILO's HFProp and alike have been designed to make the use of VOACAP easy for hams. They should not be compared bit-to-bit with a complex engine like VOACAP, Proppy or the old WinCAP Wizard 3.

The first applications will never ask you to enter "Man-made noise in -dBW at 3 MHz in the 1 Hz bandwidth" or "the multiplier to adjust the predicted critical frequency of the Es layer", nor will they suggest that you choose between the CCIR.013 and SAMPLE.022 antennas. These small applications present prediction parameters using the terms accepted in the Amateur Radio, and assign correct defaults to the ones that should not be adjusted in ham applications. The main purpose of these light ionospheric models is to let the operator see at a glance the trends of propagation while he/she is on the air - with a minimal distraction from the transceiver. They are useful, and many amateurs keep these program running on their computer screen whenever they do DX'ing, and they help them a lot.

Globally, all these programs, from the simplest to the most complex can be grouped in two categories :

- VOACAP-based programs showing a prediction accuracy potentially equivalent, and respecting "VOACAP rules"

- Non-VOACAP based programs that provide an overview of propagation conditions. They are based on the simplified models and their accuracy is inherently lower than that of the programs in the first group, ProLab Pro being probably an exception.

The most important in all these applications are not the number of parameters that they claim to use but the accuracy of their predictions. Idem for their user interface : some are on-the-air operations oriented while others are research-oriented programs.

Not all analysis and prediction programs are thus the same. Without knowing anything in this matter, we can already foresee several features, source of potential inaccuracies, weaknesses, gaps or bugs that will be disappointing, annoying or irritating, and will lead you at the end to select another product better designed, more accurate, complete or user-friendly.

At least five major features belong to the requisit, what other call the specification of any modeling program, and especially an amateur propagation prediction program :

- the reference model

- the graphic user interface or GUI

- the outputs

- the exchange of data

- the available resources

The reference model

If it is not itself the reference model, but you probably don't run the master application on the supercomputer, Hi!, like any modeling program, a propagation analysis and prediction program is based on a reference ionospheric model. As we just told these down-sized versions of the reference model can be more of less complete, flexible, and thus provide a more of less accuracy.

Models used are the key arguments to take into account when selecting a product. At the limit, hundreds of parameters injected in approximate algorithms will never give you an accurate prediction (hopefully they are tricks to improve results). If a publisher uses a well-known model, performing and recent, it will not hesitate to speak about it to attract customers. If nowhere on his website or in the product specification there is mention of a model or a method based on scientific studies, you can already conclude that the product uses probably approximations and will never give you accurate predictions. At best such a product uses median values, old but efficient algorithms known to provide general forecasts, overviews, but they are unable to give you an accurate forecast for a specific day, and even less at short-term. Its forecasts show 50% of confidence, in other words a 50-50 reliability; they can be exact today and false tomorrow, all the art to interpret statistical results.

If you wish to get accurate forecasts, a flexible and powerful product, do always select either a VOACAP-based application or an advanced model using real-time data and a down-sized version of IRI-2001 and ,when they will be available, a down-sized version of the International Geomagnetic Reference Field (IGRF) without to forget to take into account weather conditions (for the top band propagation).

The GUI

Propagation prediction programs differ also from each others by the quality of the graphic user interface, what you see on the main screen, in its message boxes, its scrolling menus or when displaying a forecast.

The GUI depends on each publisher and sometimes on the operating system. At first sight it is a minor detail without relation with the reliability and accuracy of the forecast. However this interface is the only way that you have to communicate with the model; this is through its screens that you feed it and receive outputs. So, when you request the program to display a coverage prediction around a target location or at the earth scale for example, the screen resolution comes into play. When you click somewhere in the chart to display an additional chart or figures, subroutines must be able to locate your cursor with accuracy on the chart or on the map and then provide an answer with all required accuracy.

The GUI must also response quickly to your requests and without bug. Here all is a matter of design and development, for short of know-how. Some applications are written in elementary languages, sequential or procedurals (GWBasic, Fortran, etc) while others use object-oriented language (3GL, etc). Some use much resources, others are supported by a technical team contrary to many freeware, but not all, that are provided as one states "as it, without warranty, expressed or implied, made by the publisher as to the accuracy, suitability and functioning of the program and related material".

At last, be aware that no product is never perfect. In the worst case, here is a real-life example. A well-know publisher has quoted in his product specification "supported by Mac OS, Windows 98/NT/2000/XP", except that he visibly not tested the basic functions of the program under Windows XP. Indeed, what to conclude when bugs occur at first run on the main propagation map and when the download function is almost inop... This is unacceptable and such advertising are misleading.

Thus, take care when buying a new propagation program putting in light a new GUI or features. Download always first the demo when available and check on the FAQ or on the Internet (eHam, forum, etc) if other users are satisfied of this product because, unfortunately, it seems that no publisher will never tell you all the true for marketing reasons.

Outputs

Depending on the reference model, the program must be able to output a result, a forecast in form of tables, report, charts or maps, static, dynamic or animated.

Too many programs are still using outmoded menuing or provide practically the same interface as the engine that drive their forecast, and designed more then ten years ago when the graphic interface was in infancy and still seldom used. To believe that imagination and originality are no more the interest of publishers...

At last, outputs should be flexible, using various type charts or reports, custom layout, etc. The program must also include a printing option, a small function that some applications forgot. Hopefully at the time of printscreens, OLE and other DDE this problem can sometimes be bypassed but its omission it is definitely not a clear and complete design. 

The exchange of data

A complete application is flexible what means that it should be able to fit also to external needs and change for example its layout or data format on request. But first af all a well programmed application and complete must be able to export its data or outputs in other formats. In addition it should be able to receive external inputs through batch user files, macros and other shell mode.

You imagine well that very few programs support all these functions. In fact I don't know any supporting simultaneously more than three of these features, excepting those running of supercomputers. We start thus from a wrong side, but don't go way, the situation is not as desperated as it looks to.

The available resources

In addition, the design and the GUI are strongly depending on the available resources, I mean the power of the computer running the application. To get quickly a forecast that means to use one or more fast CPUs able to process many operations in a single second, have at disposal a large amount of memory to handle many processes, functions and variables simultaneoulsy, and extended and fast mass storage disk space (hard disk of high capacity and fast access) to install the application and exchange exeternal data. Hopefully I can already to tell you that any Pentium or faster computer is today capable to support most of these functions. Some free and low cost elementary programs run even on AT machines but you can probably no more find these ancestors, Hi !

A forecast for each variable

In buying a propagation program, we expect to get good results, a return on investment, Hi ! But after all what has been told about specifications of such programs, and after have used myself several of them intensively during months, you gave already understand that I have to refute or at least temper your good a priori about the performance of these applications because not all is perfect.

As we told, a least two constraints are inherent to ionospheric models to get an accurate forecast : the computer resources to handle many parameters and algorithms or functions used.

As we shortly explained what resources are require to use such programs, let's imagine that you own a powerful computer (at least Pentium II) or that you can use the one of your institution. Remain the question of the accuracy and the reliability of the prediction. An accurate program, able to give you a forecast with the highest reliability or a high degree of confidence means that the model don't only trust in statistic values but considers a great number of on-line variables, many functions and many sub-models processing specific varaibles, parameters and processes. At the limit, this program must be able to ouput a forecast for each variable and interactions between each of them as well. But hopefully amateurs don't request such a level of power and content also with a much lower reliability, often close to 50%

Knowing this, can a program running on a personal computer provides similar results than a forecast calculated on a supercomputer like the NEC multi-node SX-5 displayed at left ? Asking the question is already to answer it. Of course both systems are "hops" apart and any down-sized ionospheric model is by definition a lighter version of the master, less complex and thus providing more limited predictions or less accurate in both space and time. But sometimes small means also powerful, isn't it ?... Yes, it is, sometimes.

Let's talk amateur radio

Sometimes [amateur programs are powerful] means that in average this is... not the case. But before answering to this question and to explain why this is not always the case, we are going first to answer to another one to set the frame of our thoughts.

There are not many questions that an amateur radio asks about propagation. A propagation prediction program has to answer to two questions depending on whether the amateur is a DX chaser or a contester :

- For the DXer the question is straightforward : what will be best time and the best frequency to work such a DX country?

- For the contester, this is the contrary : he or she would like to know what band to use and at what time to get the largest coverage, worldwide. For some DX-peditions working with a large team the world map can be divided in several sectors (e.g. per continent). In this case the coverage will be more limited.

As we are not medium, simulation programs will provides us the answer to these questions. But as we introduced, many factors can affect the propagation conditions. What are they and how are they gathered in these programs ? We have explained that sun and geomagnetic conditions as well as the operators' working conditions represent the essential factors to take into account. Both groups are thus divided in several subsets, each being constituted of several data to set correctly if you want an accurate prediction.

The most used parameters

As we explained about ionospheric perturbations, HF propagation uses primarily sky waves that reflect on or in the ionosphere. Therefore ionospheric conditions are extremely important to get an accurate propagation estimation. The density and altitude of ionospheric layers are variable up to change from hour-to-hour and even in less than a quart of an hour. In addition, each band shows its proper behaviour and might be suddenly closed although it was open one hour ago or that the adjacent band is still open.

Most of the time this change in the propagation conditions are directed by the sun activity and in a lesser extend by the fluctuations of the geomagnetic field. Or said in other words, at some occasions, depending on the activity and the position of the sun (and thus the time of the day and the season) the propagation medium - the ionosphere - simply doesn't support any more transmissions. To avoid to be confronted to such problems, propagation program permit you to calculate in a few seconds the best time and frequency to increase your chances to make a good QSO at long distance.

Recall that the propagation question does not occurs at short distance where other factors intervene to permit of deny a communication circuit. First there is the gray line map and the sun position that you have to follow along the day to be able to work such or such countries located in the sunlight, then to this area you have to substract the skip distance that depends directly on the working frequency. Do not try for example to work a station located 500 km away on the 20 or 15m band. Even if that station works with 1 kW he has all chance to arrive on your antenna too weakly to be heard. That could however work if your signal is emitted vertically in using for example an antenna place close to the ground (2-5m high) that will display a main lobe close to 90° with the ground. But don't forget that in this case your sky waves have also more chances to cross the ionospheric layers without reflection... Add to this effect the Pseudo-Brewster Angle (PBA) that produce an attenuation of your signal over poor grounds. As you see many parameters can interfere with the theoretical propagation that forecasts an opening of a closed band to some countries.

This confirms already that an accurate propagation prediction program must incorporate the circuit specifications like your antenna specifications (power, gain, height, takeoff angle, etc), the expected signal-to-noise ratio and the ground properties (dielectric constant, conductivity) in addition to the latest solar and geomagnetic indices.

In this matter, what are parameters the most used in these programs ? In a model like VOACAP (signal model) or CCIR (atmosphere noise model the next input parameters are considered :

- Propagation environment (date, time of day, required S/N ratio and reliability, ...)

- Transmit terminal (mode, power, coupler loss, ...)

- Transmit antenna (frequency, antenna gain, main lobe heading, band range, takeoff angle, ...)

- Receive terminal (mode, signal sensitivity, atmospheric noise, man-made noise, ...)

- Receive antenna (frequency, bandwidth, gain, main lobe heading, band range, takeoff angle, ...)/font>

- Ground (conductivity, dielectric constant, Pseudo-Brewster Angle, ...)

- Solar conditions (smoothed sunspot number or average solar flux)

In addition some products add :

- Geomagnetic conditions (horizontal and vertical components, Ap, Kp, Q-days indices, ...).

If these parameters differ from the ones used by your current application, tell you that you have maybe forgotten something in selecting your program. As we have just seen, examining the propagation channel alone without considering terminals (TX and RX) and sometimes objects in-between will not reflect properly your working conditions and thus the prediction will be affected with a large biais factor. To get an accurate prediction suited to your working conditions, it is important to consider the entire end-to-end system, including not only the Sun and geomagnetic conditions but also your antenna specifications and the ground surrounding your transmitter (and if possible the receiver). This is using all these parameters that a model like VOACAP is able to simulate correctly the propagation between two stations or more exactly, to give very good results under "ordinary" working conditions : not too close to the top band, not too close to the polar caps and preferably using an antenna polarized the same way as magnetic field at the concerned latitude.... I know, that represents already many exceptions. But as we told earlier, nothing is never perfect...

The poor prediction capabilities of small programs is obvious in many amateurs products. The antenna properties for example are often ignored although they are emphasized in a model like VOACAP. Antenna modeling is not taken into account without reason. As you know its radiation pattern and pick up efficiency vary from location to location, from the heigth above ground and near obstructions. To get a model as close as possible to the reality it must also consider with the same importance parameters like the local ground conductivity, the dielectric constant, and optionally terrain variation factors.

These parameters are mandatory to establish a good propagation estimation chart, whether you work with only one circuit (from your station to a remote one) or operate a large network of circuits (this latter concerning first broadcasts).

Beside these parameters, the most important factor to consider is the status of the propagation medium itself that displays the most large variations. It must be know that HF propagation losses can reach 60 dB. In the field that represents the equivalent of changing your output power by a 6-factor ! In other words in the most complete programs, a bad setup will change your band from open to closed !

The other parameters have less impact on your signal. In increasing for example your ouput power from 100 W to 1 kW, you add only 10 dB more along your path. And quasi nonebody excepting a handful of professionals use highly directive HF arrays that exceed 20 dB.

Next chapter

Setting of a VOACAP model

Page 1 - 2 - 3 - 4 - 5 - 6 -