1、 Rec. ITU-R F.1487 1 RECOMMENDATION ITU-R F.1487*TESTING OF HF MODEMS WITH BANDWIDTHS OF UP TO ABOUT 12 kHz USING IONOSPHERIC CHANNEL SIMULATORS (Question ITU-R 213/9) (2000) Rec. ITU-R F.1487 The ITU Radiocommunication Assembly, considering a) that ionospheric radiocommunication in the HF bands is
2、an economically effective transmission medium for many services requiring beyond line-of-sight operation; b) that simulation of the ionospheric propagation channel could reduce the time and expense of studying and testing the performance of such service systems; c) that some administrations have rep
3、orted good correlation between the results of laboratory tests conducted on simulators and the results of tests of data modems in operation, recommends 1 that for the simulation of HF ionospheric transmission for systems up to about 12 kHz, the methods described in Annex 1 are preferred; 2 that the
4、method described in Annex 2 should be referred to for comparative testing of HF modems; 3 that when simulators are used to predict, in a quantitative sense, how well a particular modem may be expected to perform on HF circuits, the representative channel parameters listed in Annex 3 be considered on
5、 a provisional basis. ANNEX 1 HF ionospheric channel simulations 1 Introduction HF ionospheric radiocommunication is typically characterized by multipath propagation and fading. The transmitted signal usually travels over several paths or modes to the receiver via single and multiple reflections fro
6、m the E and F layers of the ionosphere. Since the propagation times over the paths are different, the signal at the receiving antenna may consist of several multipath components spread in time over an interval of up to several milliseconds. The average heights of the ionospheric layers may increase
7、or decrease with time, which introduces different frequency (Doppler) shifts on each of the multipath components. The ionosphere is also turbulent which causes Doppler spread (fading) of each component and a resultant fading of the composite received signal. All of these effects produce multiplicati
8、ve signal distortion and degradation of the performance of communication systems. _ *This Recommendation should be brought to the attention of Radiocommunication Study Groups 3 (Working Party (WP) 3L) and 8 (WP 8B). 2 Rec. ITU-R F.1487 If a continuous wave (CW) signal is transmitted over an HF link
9、the spectra of the received multipath components can appear as shown in Fig. 1. Four paths are present: one-hop E mode (1E), one-hop F mode (1F), two-hop F mode (2F), and a mixed mode (e.g. 1E + 1F). While the two magneto-ionic components (labelled a and b) in the 1E mode have about the same frequen
10、cy spreads (fading rates), their frequency shifts are significantly different, allowing them to be resolved in frequency. On each of the other three modes, both the spreads and shifts of the magneto-ionic components are essentially the same and they appear as one. The short-term multiplicative disto
11、rtion characteristics of the HF channel can thus be described in terms of the parameters that specify the signal losses, time-spread and frequency spread characteristics; i.e. the differential propagation times on the several paths, and the signal strengths, frequency shifts, and frequency spreads o
12、n each path. These parameters are subject to change on a diurnal and seasonal basis, as well as generally being different on different geographic circuits. 1487-01abBDCAFIGURE 1Example power spectra for the multipath components of a CW signalFrequency, (Hz)Tap-gainspectra,fi() (dB)A: Path 1 (1E mode
13、, a and b represent magneto-ionic splitting)B: Path 2 (1F mode)C: Path 3 (Mixed mode)D: Path 4 (2F mode)FIGURE 1 / F.1487-01 = 13 CM To compare the performance of two or more systems over real HF links, they must be run simultaneously because propagation or channel conditions vary uncontrollably and
14、 cannot be repeated at other times or over other links. The use of a channel simulator has the advantages of accuracy, regularity of performance, repeatability, availability, a large range of channel conditions, and lower cost. However, these advantages are limited if the channel model on which the
15、simulator design is based is not valid. This Recommendation describes a stationary Gaussian scatter HF channel model. It is valid for use in 3 kHz channels, and may be applicable to bandwidths up to 12 kHz wide. A practical implementation of this model may operate at baseband (audio) frequencies and
16、 thereby act directly on the output of a transmitting HF modem and provide signals directly to a receiving HF modem. However, when considering the performance of HF systems the effect of other system components should always be taken into account (e.g. transmit and receive filters and level or gain
17、control). Rec. ITU-R F.1487 3 2 Gaussian scatter model A block diagram of the stationary Gaussian scatter HF ionospheric channel model is presented in Fig. 2. This is commonly known as Watterson model. The input (transmitted) signal is fed to an ideal delay line and delivered at several adjustable t
18、aps, numbered 1, 2, ., i, ., n, one for each ionospheric propagation mode or path. At each tap, the delayed signal is modulated in amplitude and phase by an appropriate complex random tap-gain function, Gi(t). The delayed and modulated signals are summed with additive noise (Gaussian, atmospheric, a
19、nd/or man-made) and/or interference (unwanted signals) to form the output (received) signal. For the Gaussian scatter channel model each tap-gain function is defined by: )2exp()()2exp()()(tjtGtjtGtGibibiaiai+= (1) where, the a and b subscripts identify the two magneto-ionic components that can, in g
20、eneral, be present in each mode or path. The tildes indicate that )(tGiaand )(tGibare sample functions of two independent complex (bivariate) Gaussian ergodic random processes, each with zero mean values and independent real and imaginary components with equal r.m.s. values that produce Rayleigh fad
21、ing (i.e. that they are Gaussian scatter functions). The exponential functions in equation (1) are included to provide the required frequency (Doppler) shifts, iaand ib, for the magneto-ionic components in the tap-gain spectrum. 1487-02XXX nG1(t) Gi(t) Gn(t)FIGURE 2Block diagram of HF ionospheric ch
22、annel modelInterferenceNoiseOutput signalTapped delay lineInput signalFIGURE 2 / F.1487-02 = 8 CM Each tap-gain function has a power spectrum, fi(), that in general consists of the sum of two magneto-ionic components, each of which is a Gaussian function of frequency, , as specified by: GfaGfaGfbGf9
23、GeaGeaGebGe9+GfaGfaGfbGf9GeaGeaGebGe9=22222)(exp212)(exp21)(ibibibibiaiaiaiaiAAf (2) where, iaAand ibAare the component attenuations, and the frequency spread on each component is usually determined by 2iaand 2ib. Equation (2) is illustrated in Fig. 3a). Six independent parameters specify a tap-gain
24、 function and its spectrum: the two attenuations, iaAand ,ibA the two frequency shifts, iaand ib, and the two frequency spreads, by 2iaand 2ib. 4 Rec. ITU-R F.1487 The tap-gain function described by equations (1) and (2) is general in that it applies when the spectra of the two magneto-ionic compone
25、nts are significantly different and the difference in their delays is negligible. Only one of the two terms in equations (1) and (2) is required in the following cases: when the frequency shifts and frequency spreads of the two magneto-ionic components are nearly equal, their spectra nearly match, a
26、nd a single term can be used with the tap-gain spectrum in Fig. 3b); when the two magneto-ionic components have a significantly large difference in delay. In this case, separate delay line taps with appropriate spacing should be used, with each of the two corresponding tap-gain functions and spectra
27、 consisting of a single term, again as illustrated in Fig. 3b). 1487-032ib2iaiaib2iiba) Two Gaussian scatter power spectraFrequencyFrequencyb) One Gaussian scatter power spectrumPowerPowerFIGURE 3Tap-gain power spectra in Gaussian scatter modelFIGURE 3 / F.1487-03 = 12 CM Rec. ITU-R F.1487 5 3 Specu
28、lar modes The Gaussian scatter model can accurately represent the majority of typical HF ionospheric links. A specular component in a skywave mode can easily be simulated by adding a non-fading delay tap with the same frequency offset as the corresponding mode spectrum. Thus, in Fig. 3 specular comp
29、onents would appear as Dirac-delta functions at ia, iband i, as applicable. The ground wave present on a short link is essentially non-fading and can also be represented by a non-fading tap at the appropriate group time delay. 4 Models for HF channels with more than 12 kHz bandwidth While the aforem
30、entioned Gaussian scatter model has been validated for use with 3 kHz HF channels, and may be suitable for channels up to 12 kHz wide, its suitability for even wider bandwidths (viz., extended bandwidths between 12 kHz and about 1 MHz) is problematic for at least two reasons. First, the specificatio
31、n of a model to characterize such extended bandwidth channels has proven to be difficult, owing to the lack of channel information for that regime. Secondly, there is paucity of extended bandwidth modems which are most appropriate for validation of the model. Despite difficulties in generalization,
32、an extension of the 3 kHz channel model is contemplated, based upon a belief that such a model may be required in the near future. Accordingly, studies are being undertaken to accommodate channel bandwidths in excess of about 12 kHz using logical extensions of the present Gaussian model. While the l
33、ogic is generally defensible, more development and validation is needed before a general extended bandwidth channel model can be recommended. For very wide bandwidth simulations this extended approach may not be tenable because the channel conditions may vary within the transmission bandwidth. ANNEX
34、 2 Comparative testing of HF modems 1 Introduction The generality of the simulator architecture described in Annex 1 allows the simulation of an extremely wide range of HF ionospheric channel conditions. This can be problematic when a general comparison between different HF modems is required. This
35、Annex describes a technique that employs a simplified HF channel simulation to provide a compre-hensive performance characterization of HF modems in a graphical form. 2 Modem characterization technique The characterization technique utilizes a simulation of two independently fading HF skywave modes.
36、 Magneto-ionic splitting is neglected i.e. only one term from equations (1) and (2) in Annex 1 is employed. Both modes have equal mean attenuation, equal Doppler shifts and spreads, and are separated by a multipath delay. The S/N is set by adding band limited Gaussian noise. 6 Rec. ITU-R F.1487 Whil
37、st Doppler shift is held fixed (usually zero), the performance of a modem is measured in accordance with a common set of test criteria, such as that specified in Table 1. BER is used as the measure of modem performance. At each Doppler spread/multipath delay combination the S/N which produces a BER
38、within the acceptable range is determined (the length of each test should be determined by the same algorithm as presented in Annex 3, 6; such test lengths have proved to be satisfactory in practice). The resultant data can be presented as a three-dimensional plot, whereby S/N (up to a high maximum
39、allowed value), Doppler spread and multipath delay are attributed to the three orthogonal axes. Within the three dimensional space, the locus of points of constant BER describes a surface which indicates the modem performance across the range of tested channel conditions. TABLE 1 Suggested criteria
40、for comparative modem tests In general, the modem performance surfaces produced by this technique have a low level, relatively flat working region, in which the modem achieves the required BER performance. This valley region is bounded by steeply rising sides that rapidly climb to the maximum S/N us
41、ed in the simulations, producing a plateau region in which the modem fails to reach the performance criteria. Indeed, the plateau region generally indicates that the Doppler and multipath spreads are so severe that increasing the S/N will not improve performance. Figures 4 to 6 show a series of exam
42、ple performance surfaces for a particular 8-PSK serial tone modem operating at user data rates of 2 400, 1 200, and 300 bit/s. The plots show that the modem can provide stable performance over a range of Doppler and multipath conditions, but that once certain limits are reached (e.g. about 4 Hz and
43、at about 10 ms for the 2 400 bit/s data rate) the modems performance is rapidly degraded. The expansion, and drop, in the valley floor as the data rate is decreased is a clear example of how the performance of HF modems may be compared using this technique. When reporting results of modem testing it
44、 is important to quote all necessary modem settings, as data rate, FEC and interleaving can have a significant effect on BER. Input parameter Range Increment Multipath differential time delay (ms) 0-4 4-12 12-20 0.5 1.0 2.0 Doppler spread (Hz) 0.1 0.5-4.0 4-20 20-40 Not applicable 0.5 2 4 S/N (dB) 1
45、0 to 50 1 Acceptable BER range(1)2 103to 0.5 103(1)The BER range may be varied to suit the requirements of the users application. NOTE 1 The length of individual tests should be set in accordance with Annex 3, 6. Rec. ITU-R F.1487 7 1487-04126 1014182432400123456789101112162010010203040500.10.51.5Do
46、ppler spread (Hz)Multipath (ms)S/N(dB)FIGURE 4Example performance surface for a 2 400 bit/s8-PSK serial tone modemFIGURE 4 / F.1487-04 = 12 CM 8 Rec. ITU-R F.1487 1487-05126 1014182432400123456789101112162010010203040500.10.51.5Doppler spread (Hz)Multipath (ms)S/N(dB)FIGURE 5Example performance surf
47、ace for a 1 200 bit/s8-PSK serial tone modemFIGURE 5 / F.1487-05 = 12 CM Rec. ITU-R F.1487 9 1487-06126 1014182432400123456789101112162010010203040500.10.51.5Doppler spread (Hz)Multipath (ms)S/N(dB)FIGURE 6Example performance surface for a 300 bit/s8-PSK serial tone modemFIGURE 6 / F.1487-06 = 12 CM
48、 ANNEX 3 Quantitative testing of HF modems 1 Representative channel parameter combinations Modem performance can be specified as the BER as a function of S/N for two independently fading paths with equal mean attenuation, equal frequency spreads and no frequency shifts. The differential time delay i
49、n the following sections is defined as the multipath delay between the two modes, while the frequency spread is the 2 value as used in equation (2) of Annex 1. When these representative parameter values are used for comparative testing it is recommended that the parameter values used should be quoted. For further information concerning latitude regions refer to the ITU-R Handbook on The ionosphere and its effects on radiowave propagation (1998). 10 Rec. ITU-R F.1487 2 Low latitudes 2.1 Quiet conditions
