HF propagation involves several interrelated phenomena that results in a highly variable medium. This variability is a challenge to anyone that needs to design and implement effective high speed digital communications systems for HF. The ability to quantitatively evaluate how successfull engineering designs carries though to real-world implementations, often makes the difference between success and failure. Experienced, well-equipped engineers use special tools such as channel simulators to shorten development cycles. These are invaluable for example, to verify dynamic range performance, acceptible signal to noise ratio performance, as well as a number of other factors such as adjacent channel interference and frequency/timing tolerances. These are very common real-world problems. Besides the evaluation of these basic factors, protocol performance is of equal importantance. This has to do with how efficient frame and character synchronization is, how effective error control works, and how successful protocol adaptation actually is.
Although some of these tests may be done by on the air tests, however, HF propagation conditions are almost impossible to repeat thus there is not really a chance for making comparitive tests this way. What really is needed is a means to create an artificial ionospheric test medium that can be reproduced at will. Only then is it possible to set up norms and milestones for performance evaluation.
Computer simulation is one way to obtain quantiative results. A simulation study based on theoretical concepts can provide the basis for establishing expected performance characteristics, also serve as a guide as to requirements for hardware and software expectations. It can provide an essential justification for continuing development work without the risk.
During test and development phases, real-time testing using an HF channel simulator is essential. The key to developing an effective waveform and protocol suitable for high speed HF digital communications, is in understanding the behavior of the ionosphere and how it will impact communications.
HF communication is typically characterized by multipath propagation and fading. Transmitted signals travels over several propagation modes to the receiver via single or multiple reflections from the E and F ionospheric layers. Because of different propagation times over different paths, signals arriving at the receiver may be spread in time by as much as a few milliseconds.
Ionospheric turbulence causes distortion in both signal amplitude and phase, in addition, different ionospheric layers move up or down, which leads to independent Doppler shift on each propagation mode. Ionospheric skywave HF, multipath arises from paths with different number of multiple reflections between earth and the ionosphere (multiple-hop paths) and from paths at multiple elevation angles connecting the same end points ("high" or "low" rays). Natural inhomogeneities of the ionospheric layers and polarization dependent paths because of magnetic-ionic effects also contribute to multipath.
The effect of these natural inhomogeneities in the ionosphere causes multipath spreads of 20 to 40 µs on each path or mode, and the high/low and ordinary/extraordinary rays results in a path spread of about 200 µs. For single hop links (800-2000 km), a maximum multipath spread of 100 µs is common. In this case, all paths are via the same reflection area and thus there is no significant difference in the Doppler spread on different modes. The channel is often a very slow fading channel, with time stabilities of 100 s or more, corresponding to a Doppler spread of 0.01 Hz. Multipath spread in the range of 1 to 2 ms for HF occur for short ranges (because of near vertical incidence) of under 800 km due to delayed energy arrival via repeated earth-ionosphere reflections or over long paths (2000 to 10000 km) that require two or more hops. On these long skywaves, different spread, controlled by the Doppler shift differences can range up to 1 to 2 Hz (fades per second.)
Short-term distortion on the HF channel can therefore be described in terms of the parameters that specify the time-spread and frequency-spread characteristics, i.e., differential propagation delay between modes, and the strengths, Doppler spread on each mode.
Shown below is an actual example of these different mechanisms in action. (This illustration provided by courtesy of J.P. Martinez (ref 1). Martinez experimentally recorded an event on November 9, 1994 that consisted of saving a digitized audio tone of a remote broadcast station's carrier. The broadcast station's carrier was located on 7.7 MHz and arrived via the ionosphere; the broadcast station being located on the island of Gibraltar and the receiver located on the South coast of England. Subsequent processing of the recorded digital data revealed frequency-domain behavior over time. For this, the results of 256-point FFTs are presented as pixel intensity values on the Y-axis, with time plotted on the X-axis.
Martinez's Dopplergram illustrating several interesting ionospheric phenomena.
For the graph shown, each pixel point in time represents approximately 20 seconds of signal with UTC hour tic marks shown along the top. The Y-axis represents 0.025 Hz/pixel (256 pixels=6.25Hz). This representation effectively shows the history of a very slowly-changing process, with most of the finer, random events, filtered out to better illustrate the various propagation modes.
Because of the frequency in question (7.7 MHz), we are resonably sure that the propagation mode is most likely via the F-layer. Note that at about 06:00 UTC the signal penetrates and no signal propagation path to Earth results. Just before this happens, note the high F-layer ray (the so-called, Pedersen ray) appear lower in frequency than the main (low) ray. The high ray itself appears to be split in two parts each with distinct doppler shifts; the upper image being probably being the opto-ionic, or O-ray, and the lower image being produced by the extra-ordinary, or X-ray. The X-ray undergoes further retardation due to interaction with Earth's magnetic field. Shown is that the high and low rays of the O-trace penetrate first, followed by the X trace. This effect is distinct on this dopplergram, but only rarely is it identifiable by ear. If recognized, it appears as regular fading (QSB) that slows down to zero as the particular path fades out.
About 06:40 UTC the F-layer comes back in again and the process is seen in reverse, X-trace appearing first and splitting into high and low, followed by the O-ray. Further more diffuse propagation paths open up a few minutes later.
Watterson et al (ref. 2), using wide-band HF emissions over a path between Bolder,Co. and Washington, DC., proposed a model for narrow band HF channel. This model forms the basis for most modern HF channel simulation and used in both software and hardware channel simulation.
This model, known as the "Watterson Gaussian-scatter HF ionospheric channel model", assumes that the HF channel is non-stationary in both frequency and time, but considered over small bandwidths (<10 kHz) and sufficiently short times (<10 minutes), most channels can be considered representative by a stationary model.
The HF channel is modeled as a tapped delay line, with one tap for each resolvable mode (or path) in time. The delayed signal is modulated in amplitude, and phase, by a complex random tap-gain time-dependent function that is defined by:
Where a and b subscripts denote the i-th element in a time series representation for two magnetoionic path components. In this context, Gia(t) and Gib(t) represents two independent complex bivariate Gaussian ergodic random processes, each with zero mean and independent real and imaginary components with equal RMS values that produce Rayleigh fading. The exponentials provide frequency shifts fia and fib for the magnetoionic components in the tap-gain spectrum. Each tap gain has a spectrum that, in general, consists of the sum of two magnetoionic components, each of which is a Gaussian function of frequency, as specified by:
The Tap gain distributions for a two-ray model are shown in Figure 1 below.
Figure 1. The tap-gain function and it's spectrum: For a two component model
with two attenuation components, Aia and
Aib. The frequency spread on each
component is determined by 2
and 2
.
The frequency shift on the two components are given by
and
.
CCIR Recommendations for the Use of HF Ionosperic Channel Simulators. CIR Recommendation 520-12 gives guidelines for practical values for frequency spread and delay times between ray components:
| Condition | Freq. Spread (Hz) | Delay (ms) | |
| Flat Fading | 0.2 | 0.0 | |
| Flat Fading (Extreme) | 1.0 | 0.0 | |
| Good | 0.1 | 0.5 | |
| Moderate | 0.5 | 1.0 | |
| Poor | 1.0 | 2.0 |
It is proposed that these parameters be used to validate average and extreme conditions during simulation as well as during actual hardware testing.
There are several viable alternatives that may serve as a platform
for simulation work.
One possibility is Ptolemy, is a well known
system that is used in academic environments, however requires
moderately-sized UNIX-based workstations. There also are several commercially
available signal processing toolboxes available for use with scientific
development packages such as Matlab®, MathCad®, or Mathematica®
for example - all of which offers a flexible and powerful simulation
environment that often includes an interactive interpretor, code compiler,
and simulation module.
Presently, however, these solutions offer non real-time simulation. That has advantages but also disadvantages. The advantages being that systems can be modeled before any hardware is built. Testing existing hardware does however, require a real-time testing environment.
Several professional-grade channel simulators that allows real-time simulation of baseband HF channels are available on the market. These include units manufactured by Harris (RF-3460), Cossor (1250), Magnavox, and Signion Systems (http://www.signion.com/hfsim.htm).
Linux (GPL) version contributed by Tomi Manninen, OH2BNS (oh2bns@sral.fi)
© Johan Forrer, 1998-2008