tions for the four 5%-mile paths is illustrated by the 
plots of hourly mean levels for 5 days in August 1944 
shown in Figure 3. The range of variation in level 
increases with the excess of path length over optical 
range, and there is an obvious similarity between the 
three nonoptical paths as regards the larger changes 
in level. This similarity does not extend to the optical 
path AC, which shows signs of an inverse correlation 
with the major variations of the nonoptical paths. 
For a standard atmosphere (4% earth radius) the 
receiver on the AC path is near the record maximum 
of the interference field. For fairly small departures 
from standard (curvature corresponding to, say, 1.0 
to 1.7% times earth radius) the range of variation 
caused by interference is quite small, about +4 to 
—1 db relative to free space field; the smallness of 
the variation is due to the appreciable effect of diverg- 
ence of the reflected ray in this case. 
While other factors beside interference with the 
reflected ray are almost certainly operative in produc- 
» ing variations over the optical path, slow fading with 
a range of the order of 10 db is quite common; and 
a slightly substandard index gradient, which would 
produce a marked decrease in level for the nonoptical 
paths, would leave the AC path in the neighborhood 
of the first interference maximum. 
The free space levels (48 db above 1 py from A, 
44 db from B; difference due to different radiated 
powers at this period; 144 db estimated atmospheric 
absorption allowed for) are marked on the plots of 
Figure 3. For all four paths the highest levels reached 
are in the neighborhood of the free space value; levels 
several decibels above free space are occasionally 
reached during good periods, but they are only rarely 
maintained for as much as a few hours (e.g., path 
BD, May 12, 13 and August 5, 6). The long-time 
average level for path AC should be close to free space 
level, probably about 2 db above. It is actually about 
5 db below for the first half of August 1944 and 
about 3 db below over a period of several months. 
While some of this discrepancy may be due to residual 
experimental error, the fact that it appears to be 
least during periods of poor transmission between the 
low stations (e.g., August 10 in Figure 3) may be 
significant. 
For the nonoptical paths BC, AD, and BD the 
standard level is 20, 37, and 79 db below free space 
level, respectively, for S band. The corresponding 
figures for X band are 28, 53, and 113 db. The’stand- 
ard level is shown in Figure 2, and in Figure 3 for 
paths BC and AD; for path BD it is over 30 db below 
the receiver threshold. While it is rare, during the 
summer months, for the level to remain near stand- 
ard for a large fraction of the time (in February, 
however, the AD level was within 2 db of standard 
for about 25 per cent of the‘time), the minima of 
the major signal variations usually lie within about 
=£5 db of the standard level except (1) during runs 
of particularly good weather and (2) during fogs 
conditions which are likely to be associated with a 
substandard index gradient. Striking examples of the 
latter occurred on June 4, 5 (Figure 2) and August 
10, 11 (Figure 3). On the last occasion the BC level 
went about 20 db below standard. 
The X-band results for the 57-mile paths are gen- 
erally similar to those for S band as regards the major 
variations, but the range of variation is noticeably 
larger, particularly on paths BC and AD, which are 
not mueh Jonger than optical range; the increase in 
range of variation for these pathis is of the same order 
as the difference in standard level for the two wave- 
lengths. In general, short-period variations are larger 
and more rapid for the shorter wavelength. 
Figure 4 shows the results for both S and X bands 
over the 200-mile paths AE (high stations) and BF 
(low stations) for part of the same period as shown 
in Figure 3. Atter allowing for the estimated atmos- 
pherie absorption (6 db for S, 16 db for X) the free 
space levels are similar for the two wavelengths, ex- 
perimental uncertainty being appreciably greater at 
APPENDIX 
“general” level, which is in fact that obtained under 
“well-mixed” meteorological conditions. 
Further details of the path and a discussion of the 
results in relation to general meteorological conditions 
over the path have been given in two National Physical 
Laboratory reports,*°* which cover the first year’s 
operation. A further report is in preparation. The aim 
hete is limited to a general description of the type 
of results obtained, with examples of some character- 
istic signal records. 
Figure 5 gives a plot of hourly mean level for March 
1944 which clearly shows the two main characteristics 
of the signal: the reasonably constant general level 
and the regular diurnal cycle which occurs with radia- 
tion nights. The period March 21 to 26 is typical of 
an undisturbed run of clear nights; note the period 
of marked substandard signal in the early morning of 
March 27%, indicating that condensation near the 
ground has reduced the water vapor content there suffi- 
ciently to make the lapse rate negative. Intermittent 
rain in bad weather periods usually gives a more vari- 
able level than cloudy weather with no precipitation ; 
a small rise in level is often observed with continuous 
rain (direct effects of rain on the equipment have been 
carefully guarded against and may be assumed negli- 
gible) and a more marked rise with clear skies in 
daylight following rain. These effects are readily ex- 
plicable in terms of changes in water vapor distribu- 
tion. 
The work of the past 6 months (Summer 1944) has 
shown a definite correlation of high level at night with 
temperature inversion whether with clear or with 
variable skies; on the other hand, clear or variable 
skies with no temperature inversion (e.g., with incom- 
ing cold air) show no night peak of signal. In general 
the increase of level on an initially clear night is 
arrested by the development of low cloud or of fog. 
Double maxima are often observed in the night peaks 
(e.g., March 15 in Figure 5). 
The magnitude of the peaks on radiation nights is 
usually 5 to 10 db; it can occasionally reach 15 to 
20 db particularly in summer. It seems very probable 
from the geometry of the path that earth-reflected rays 
play little part, at least for moderate degrees of bend- 
ing. It is therefore reasonable to seek to explain the 
larger variations as resulting from increasing ray 
curvature. In terms of the rough estimate mentioned 
above, a change from standard to “flat earth” condi- 
tions would give an increase in level of the order of 
10 db, which is a typical figure for the observed rise 
on an undisturbed radiation night. It is of interest to- 
note that free space level is never reached on this 
path; the highest instantaneous level reached is 10 to 
12 db below free space. In other words complete, or 
nearly complete, reflection regions do not exist at 
heights of the order of 1,000 ft or more (required to 
“clear” the barrier) over this path. This is in line with 
the observed lack of any effect of high inversion on the 
signal level. 
Overland Measurements: Whitwell 
Hatch to Wembley 
A single S-band link has been in continuous opera- 
tion oyer this 38-mile path since March 1943. Its 
terminals, with the transmitter in one of the Ad- 
miralty Signal Establishment buildings and the re- 
ceiver at General Electric Company Research Labora- 
tories, -were chosen for operating convenience rather 
than to meet any special requirements for the path, 
‘as an important subsidiary purpose was to provide for 
controlled long-period tests on equipment developed 
for use in the less readily accessible stations of the 
Trish Sea program. Apart from routine checks the 
equipment normally operates unattended; automatic 
frequency control at the reeeiver has been in operation 
since June 1943. But the receiver is provided with a 
relay-operated alarm which can be set to operate on au 
abnormal change of received level in either direction 
(normally downwards) and this has proved valuable 
491 
in calling attention to both faults and unusuai propa- 
gation conditions. 
The transmitter is an a hill 725 ft above sea level 
and the path runs northwards across the Thames 
valley and the western outskirts of London to the 
receiver, which is only 170 ft above sea level, in low, 
undulating, built-up country. For standard conditions 
the path is clear except for the last mile where trees 
and houses form a barrier elevated about 12 degree 
above the ray path. This introduces a local diffraction 
loss at the receiver which has been estimated roughly 
at about 30 db. This estimate is necessarily an un- 
certain one, both because of the complexity of the 
ceal barrier (which is approximated as one or more 
opaque straight edges) and because of possible sea- 
sonal variations. 
Seasonal variations in general signal level have been 
observed with a maximum in late summer of the order 
of 10 to 15 db higher than the single winter minimum 
recorded so far. An attempt to explain this variation 
in terms of changes in the horizontal plane diffraction 
pattern of part of the barrier with varying opacity of 
the tree background does not appear to be supported 
by the results of the past few months (Summer 1944). 
The mean level for the whole period is, however, close 
to 30 db below free space (52 db above 1 py receiver 
input) and is thus at least of the same order of mag- 
nitude as the estimated standard level. The unfor- 
tunate effect of this uncertainty regarding standard 
level is mitigated to a considerable extent by the fact 
that a land path of this kind gives an easily definable 
X band as regards absolute values. The standard 
levels are, of course, far below the receiver threshold, 
actually about 275 db for path AF on S and 490 db 
for path BF on X. 
The most striking characteristic of the results is 
perhaps the similarity in magnitude of the signals 
both for the two paths and for the two wavelengths. 
In general, signals are measurable for a greater frac- 
tion of the time on the longer wavelength and for 
‘the higher sites. (The difference of about 5 db in re- 
ceiver threshold sensitivity between S and X has only 
a slight effect on this.) At the peak of good periods 
the lower sites and shorter wavelengths sometimes 
reach rather higher levels, as in the case of the 57- 
mile paths the maximum signal level is frequently 
comparable with free space; rather rarely it exceeds 
free space level by something of the order of 10 db. 
For the 200-mile path the possible error in the esti- 
mate of atmospheric absorption is rather more serious, 
particularly for X band, but it seems improbable that 
this could alter the general character of the results. 
Comparison with the results for the 57-mile BD 
path in the bottom record of Figure 3 is interesting 
and is reasonably typical of the extent to which the 
performance of the long path can be predicted from 
the performance of the shorter one. It should be era- 
phasized that this was a period of good summer 
weather apart from the break on August 10-11. 
Figures 6, 7, and 8 show photographs of sections 
of the original signal records illustrating the main 
types of signal which are observed. The type of weather 
involved is shown on the record in each case, also the 
signal calibration. Figure 9 shows a good example of 
an effect which is quite often observed, particularly 
in the latter part uf radiation nights. It consists in a 
regular variation showing the characteristic rounded 
maxima and sharp minima of interference fading. This 
is in some cases superimposed on a nearly steady high 
signal (as in Figure 9). In other cases, as in some fog 
fades, it is superimposed on variations of a different 
type. It often starts and stops suddenly, completely 
changing the character of the record while it lasts, and 
that time ranges from one or two fading cycles to many 
cycles. This sort of effect has also been observed on 
other paths, over sea as well as land. : 
The striking thing about patterns of this sort is that 
they often correspond to reflection coefficients for the 
interfering ray which approach unity. In Figure 9 
the reflection coefficient calculated from the pattern 
