1 The equivalent noise power Pe expressed in terms of a given bandwidth Be is given by
where
Pm is the power measured at bandwidth Bm
DISCUSSION OF MEASUREMENT RESULTS
When plotted against the log of the distance, most of the DGPS signal data appear relatively linear with a roll-off of 20 dB per decade. Deviations from this linearity appear in several cases. Aransas Pass and Mobile Point data both show higher than expected signal levels between 300 and 500 km (Figure 4 and 7). The cause of this is unknown but is probably due to excess noise. Both the Pigeon Point and Point Blunt daytime data show patterns of deviation that are similar (Figures 8 and 10). There are two regions of signal attenuation both of which are located physically in areas of rough terrain. The first dip coincides with crossing the coastal mountain range between Santa Cruz and San Jose on Highway 17 (92 km from Point Blunt and 33 km from Pigeon Point). The second dip coincides with crossing the Sierra Nevada maintain range east on Interstate 80 (200 km from Point Blunt and 250 km from Pigeon Point). It is interesting to note that the Sierra Nevadas behave much the same as an attenuator. As the rough terrain is traversed, the signal strength drops at a much faster rate, but once the relatively flat terrain of Nevada is approached, the signal drops off at approximately the same slope as before (20 dB per decade when compared against the yellow reference line). What appears to be a very strong signal at approximately 400 km for both transmitters is, in actuality, noise from a relatively severe but short-lived spring storm while crossing the plains of Nevada. This elevated environmental noise was prolonged (lasting several minutes) and as strong or stronger than the DGPS signals at any time during the measurement, demonstrating the need to use the Type 9 message for broadcasting corrections. Cape Mendocino data (Figure 12) show a wide scattering of signal levels but when examined carefully appear to be bimodal. About half the data track the red reference line represented by the equation S = 100 - 20 * log10 (d) where S is the signal strength in dB µV / m and d is the distance in km. The other half track a line with the same slope but 20 dB lower (represented by the yellow reference line). This is particularly evident in the histogram of deviation from the least squares fit (Figure 29). It is believed that the reason for the bimodal distribution is that data were acquired in two directions from the transmitter. South of the transmitter, the terrain is characterized by rolling hills separated by relatively flat regions. North of the transmitter, there are cliffs, heavy forests, deep ravines, and mountainous areas. Fort Stevens data (Figure 13) show a relatively wide deviation from a 20 dB per decade roll-off. Much of these data were acquired in areas of rough terrain, rolling hills, and the gorge of the Columbia River. The FAA beacon data going north to Cheyenne, Wyoming have a narrow deviation from the least squares fit (Figures 32 and 33) and track very closely to a roll-off of 20 dB per decade (Figure 14). Data collected on the same signal going west across the Rocky Mountains, however, look quite different (Figure 15). In this case, the signal shows considerable attenuation when crossing the mountainous terrain. It is interesting to note that when acquiring data through the Denver area (between 20 and 70 km from the transmitter) the environmental noise was considerably higher. All four of the Gulf Coast beacons show higher signal strengths, but this may be due in part to the upward bias caused by picking the peak power across a 500 Hz span (previously mentioned).
Cumulative distributions of the deviation from the least squares fit are relatively consistent among the four different transmitter sites in the Gulf States (Figures 16 - 23). There is a probability in this case, of approximately 0.98 that the DGPS signal will be greater than 10 dB below the least squares fit. There is a slightly greater spread for nighttime data acquired on Pigeon Point and Point Blunt, and daytime data for Fort Stevens (Figures 24 - 27 and 30 - 31). Cape Mendocino data, as previously mentioned, show wide variations with a bimodal distribution (Figure 28 and 29).
COMPARISONS OF MEASUREMENTS WITH MODELS
One purpose of carrying out the field strength measurements described in this report is to compare the measurements with the predictions of propagation models. Agreement between the measurements and model predictions serves to validate the models and provides confidence in the use of the models to predict field strengths in regions where measurements have not been performed. On the other hand, disagreements between the measurements and model predictions can provide insight regarding limitations of the models or indicate a need to modify the models.
Predictions of field strength versus distance have been generated for the four DGPS beacons along the Gulf Coast and for the FAA beacon in Bennett, Colorado. The propagation models are described by Haakinson et al. [2] and references contained therein. The model inputs include transmitter location, frequency, measured field strength at a reference distance, time of day, and ground conductivity. Three types of field strength predictions are possible: smooth earth, which neglects terrain and uses a fixed value of conductivity along the path; smooth earth mixed path, which also neglects terrain but allows the user to construct a path along which the conductivity varies; and irregular terrain, which uses a mixed path and also takes into account terrain features from a terrain database.
Model predictions have not been developed for the beacons on the West Coast, because the complicated routes during the measurement campaign and irregularities in terrain and ground conductivity would make a comparison between predicted and measured field strengths quite tedious, although in principle such a comparison could be made. In the case of the Gulf Coast beacons, the relatively high values of ground conductivity and absence of terrain features result in field strengths (versus distance) that are nearly omnidirectional, and model predictions were made using the smooth earth model. In the case of the FAA beacon, there are dramatic variations in both terrain and ground conductivity, but the routes were quite simple (essentially due north and due west), so that mixed paths of ground conductivity could be developed for both routes using a conductivity database; model predictions were made using both the smooth earth mixed path and irregular terrain models.
The model predictions and measured values of field strengths versus distance for the four DGPS beacons along the Gulf Coast are shown in Figures 34 - 37. Values of the measured field strengths and corresponding reference distances, used as model inputs, are listed in Table 4. The values of conductivity that were used are the values at the transmitter location, obtained from a conductivity database. The model was run for daytime hours, since most of the data were obtained during the day. The model and measurements show good agreement at most distances, although the measured values are somewhat larger than the predicted values at large distances. This could be due to noise, which is expected to have a greater upward bias on the measured field strengths at lower values of the field strengths (greater distances), because the scale is logarithmic (dB). It could also be due to the fact that the field strength data for these beacons were acquired by recording the peak power over an entire span, as described in Section 4, resulting in an upward bias of the measured signal strengths.
The measured field strengths and model predictions for the FAA beacon are shown in Figures 38 - 41. Again, measured field strengths that were used as model inputs are listed in Table 4, and the models were run for daytime hours. Comparisons have been made for both routes that were driven (Bennett to Cheyenne and Bennett to Grand Junction) using both the smooth earth mixed path model and the irregular terrain model; thus, the effects of varying ground conductivity and irregular terrain can be separated. Not surprisingly, both models make similar predictions and are in good agreement with the data for the route from Bennett to Cheyenne, over which the terrain is relatively flat. However, the measured field strengths show considerable attenuation when crossing the Rocky Mountains between Bennett and Grand Junction. This is partly due to the relatively poor ground conductivity in the mountains (rock), as can be seen from the smooth earth mixed path model predictions, which neglect terrain. However, the deviations between this model and the measured field strengths, which are as large as 15 dB or more, indicate that terrain features also play a role in the large propagation losses that were observed in this region. Note that these propagation losses are fairly well described by the irregular terrain model. The magnitude of these deviations was somewhat unexpected, because the effects of terrain features on field strengths are not generally this large at these frequencies. For example, the report by DeMinco [3] contains numerous comparisons of predictions of the smooth earth and irregular terrain models for various path profiles and frequencies; these comparisons do not show deviations as large as 15 dB, although deviations as large as 10 dB do occasionally occur.
To further investigate the effects of irregular terrain, comparisons between the smooth earth and irregular terrain models were made for a path going west from Colorado Springs, Colorado, over the continental divide in the Rocky Mountains, and for a path going northeast from Sacramento, California, into the Sierra Nevadas. In both cases, the differences between the field strength predictions of the two models are not more than 0.1 dB. However, it has been shown by Furutsu et al. [4] that deviations as large as 15 dB or more can in principle occur in extremely irregular terrain for certain configurations of the transmitter and receiver.
We conclude that irregular terrain is unlikely to have a significant effect on field strength predictions at these frequencies; however, the effects need to be investigated on a case-by-case basis in extremely irregular terrain. The fact that the model predictions and measured field strengths are generally in good agreement provides confidence in the use of these models.
