5.2. End Node Installation
The end node was set up bearing in mind that in future stages of the project it is expected to be mounted on the cages anchoring buoys (i.e., seamark buoys). A buoy like these is constructed as a beam, that has a truncated cone situated at one fourth from its lower extremity that allows for the buoyancy property, whose upper end is at 3.5 m above sea level. Therefore, the transmitting antenna will be placed at such height by directly installing it on the beam. However, for these measurements campaign it will be temporarily mounted on a pole feigning the aforesaid altitude above sea level. Actually, as it will be seen later on, two different heights for the transmitting antenna were tested in order to check the feasibility of the link simulating a worst case scenario. Hence, the transmitter was firstly at an altitude
m and then at an altitude
m above sea level. Moreover, the end node with the pole supporting the antenna was set up on board of a boat.
Figure 5 displays the antenna installed on the pole lying on the floor of the boat and the view reaching the measurements spot.
So as to enhance the strength of the wireless link, a directional Yagi-Uda antenna having a gain dBi was selected. Another interesting characteristic is that it is pretty lightweight (i.e., 400 g) thus well suiting to its future employment on top of a seamark buoy. Indeed, its limited weight could prevent the buoy capsizing. The antenna was connected to a B-L072Z-LRWAN1 discovery kit board produced by STMicroelectronics via a 4 m long coaxial cable whose overall loss, dB, was measured by means of a vector network analyzer.
Eight different channels belonging to the 868 MHz ISM band were exploited to conduct the transmission tests so as to establish a frequency diversity scheme. Therefore, the B-L072Z-LRWAN1 discovery kit board performs a frequency hopping in a pseudo-random fashion for each transmission by switching amid the channels having the following carrier frequencies: 867.1 MHz, 867.3 MHz, 867.5 MHz, 867.7 MHz, 867.9 MHz, 868.1 MHz, 868.3 MHz and 868.5 MHz.
Finally, so as to speed up the tests, 6 discovery kit boards were sorted out. They transmitted the string that was introduced in
Section 5.1 respectively employing a different SF. All those boards were supplied by the boat cigarette lighter.
5.3. Gateways Installation
Earlier on, in
Section 3, it was claimed that two Gateways would be employed. Indeed, such a choice was made so as to improve the reliability of the link by establishing a space diversity scheme at the receiver side. In addition to it, LoRaWAN protocol itself provides for a time diversity scheme, that is ensured by the laws on temporal occupancy of the transmission band [
33], and for a frequency diversity scheme, that is implemented by switching the carrier frequency in a pseudo-random fashion for each transmission. In so doing, a more robust broadcast is achieved.
The Gateways along with their antennas were ashore installed on top of a knoll, whose altitude is 6 m above sea level, which belongs to the proprietary terrain of the aquaculture partner company. With a view of better exploiting the first Fresnel zone, the receiver antennas were placed on a 7.2 m pole raised on top of the knoll. In particular, they were installed on the pole upper extreme by mounting them on a 2 m bar fastened to the pole forming a cross. Hence, by summing up the altitude of the knoll and the length of the mounting pole, the receiver antennas were located at a height of m above sea level. As it will be seen later on, such a set up was far from being sufficient to allow for the clearance of the first Fresnel zone, but, in order to keep the installation procedures as simple as possible, the antennas could not be placed any higher since only a temporary installation was planned for these early days of the validating experimentation phases of the project.
The Gateways were implemented by means of a RAK831 produced by the RAKWireless, which is a multi-channel LoRaWAN concentrator, driven by a Raspberry Pi 3 model B.
Figure 6 shows the Gateways setup on top of the knoll.
Two directional helical antennas having a gain dBi were selected for the Gateways. They were in turn connected to the Gateways via two 15 m long coaxial cables having a total loss of dB each, that was measured by exploiting a vector network analyzer. Finally, Internet connectivity was guaranteed by a 4G LTE router since there were no other opportunities in loco. This slightly slowed down the communication between the Gateways and the network server introducing a minimal amount of latency: nevertheless it was not an issue at all.
5.4. Measurements
The tests were performed from offshore, by reaching the closeness of the breeding cages, towards land covering the distance
km.
Figure 7 reports the position of the end node (i.e., point A), the position of the Gateways (i.e., point B) and the length of the link (i.e., the red line), whilst
Table 1 lists the coordinates of points A and B.
The tests were accomplished with smooth sea conditions having a maximum wave height of 50 cm. It was a sunny day with a mean temperature of 18 °C and a gentle breeze whose average speed was of 4 m/s. Testing the system with more severe marine and weather conditions would have been extremely interesting. Unfortunately, though, the boat we had available for the fullfilment of the tests could not cast off in such cases. However, as it will be explained later on (see
Section 6.2), sub-GHz wireless links (e.g., the ones exploiting LoRa modulation) are robust to these sorts of effects. Moreover, 600 packets were transmitted by respecting the laws on temporal occupation of the transmitting band [
33] in an overall amount of time of approximately 14 h.
The measurements campaign was sorted out in 2 groups, namely
and
, by covering the same link from point A to point B (see
Figure 7) and by testing all the SFs by transmitting an amount of 300 packets per test collection (i.e., 50 packets per SF per collection). The test series differed from each other for the heights of the transmitting antenna above sea level (see
Section 5.2).
A circumstance that has to be verified is to check whether or not points A and B were in LoS. Hence, by resorting to Equation (
6), by switching the transmitting antenna with the receiver one (see
Figure 3) and by plugging into the Equation
in place of
, it can be noticed that points A and B are always in LoS since the former is not behind the horizon because
is bigger than
D.
Table 2 summarises what has just been described and it also contains the results computed by applying Equation (
6).
At this stage, the concepts introduced in
Section 4 can be applied and evaluated to the wireless link that has just been illustrated. First of all, the maximum radius of the first Fresnel zone
can be calculated by applying Equation (
5). Since a frequency diversity scheme was implemented by means of frequency hopping among the channels listed in
Section 5.2,
is evaluated by exploiting as carrier frequency the mean value of the ones previously listed (i.e., 867.8 MHz). The same convention holds for all the following Equations in which the frequency of the carrier is involved. Concerning the maximum height of the Earth bulge
H amid the link endpoints, it can be derived by means of Equation (
7). The theoretical free space loss
can be computed according to Equation (
4), therefore an overestimation of the received signal power
can be obtained via Equation (
3). Indeed, the latter is estimated by plugging into its Equation the values cited in
Section 5.2 and
Section 5.3 (i.e.,
,
,
and
), by exploiting the maximum power output for the transmitter
according to the regional regulations [
33] (i.e., 14 dBm) and by considering a null value for miscellaneous losses
. Actually, there are several losses of that kind due to the fact that, as it will be seen in a while, the
of the first Fresnel zone is not free from obstacles. Unfortunately, though, the loss introduced by such a phenomenon is hardly numerically assessable. For the same reason, an overestimation of the link margin
may be computed by exploiting Equation (
2) bearing in mind that the Gateways shown in
Section 5.3 have a sensitivity which varies along with the SF and the bandwidth. Indeed, according to the LoRaWAN concentrators datasheet [
35] and considering that a bandwidth of 125 kHz was exploited throughout the tests, the Gateways sensitivity
spans from
dBm at
to
dBm at
. Another consequence stemming from the presence of miscellaneous losses is that a lower value of
, with respect to the one that would have been obtained from Equation (
3), was truly experienced on average during the measurements thus entailing a narrower
. Notice that all of these quantities are independent of the set of tests since they only depend on the length of the link
D. All the results stemmed out from the aforementioned Equations are reported in
Table 3.
The last condition to be assessed is the percentage of clearance the first Fresnel zone
. Indeed, despite the LoS status between the link endpoints,
is partially occupied by the sea because of the Earth bulge.
Figure 8 shows a qualitative report of the wireless link respectively when
is exploited (
a) and
is used (
b). Such images are obtained by making use of BotRf [
36] which is a Telegram Bot that aids in the planning phase of wireless links. The cyan line represents the direct LoS path between the link endpoints while the magenta one represents the
of
. Finally, the brown area indicates the Earth curvature while the green one indicates the terrain profile. The reason for which the green area is for the most part flat is that it is actually occupied by the sea. Either the brown and green areas take into account the factor
k (see Equation (
Figure 4)) for the bending effect on radio waves due to the declining of atmospheric pressure. As it can be noticed, LoS is effectively achieved and the first Fresnel zone is densely occupied.
At this stage one can evaluate the percentage of clearance of the first Fresnel zone. Due to the fact that the transmitting and receiving antennas were installed at different heights, let us suppose that they shared the same altitude so as to ease the computation. In so doing, an overestimate of the clearance will be obtained. However, this is not an issue at all since such assessment has the mere scope of giving a flavour of the toughness of this specie of transmission over the sea. The maximum extent of the occupation of the first Fresnel zone by the sea due to the Earth bulge is experienced at the midpoint of the link since in the same location occur both of the maxima of
radius as well as of the Earth lump. Such a percentage could be evaluated by firstly computing the distance between the sea level and the line connecting the link endpoints (i.e.,
D in
Figure 2) at half of the link length
D relying on the flat Earth model. This term can be computed as the sum of two factors: one is length of the minor cathetus belonging to the right triangle having half of the link length
D as major cathetus and the transmitting antenna as the corner of its acute angle, and the other one is the height of the transmitting antenna
. Secondly, the Earth curvature have to be taken into account by subtracting the maximum height
H of the Earth bulge. Finally, this difference is the portion of
free from obstacles. This procedure can be fulfilled via the Equation
The final step consists of plugging into Equation (
8) the altitudes of the antennas for each trial group (see
Table 2). The resulting values for
are listed in
Table 4.
It has to be underlined that the values within
Table 4 are so scant owing to the fact that the radius of the first Fresnel zone
is far longer than the antennas elevations. As a direct consequence, additional losses belonging to the miscellaneous class
to be accounted for within the link budget (see Equation (
3)) arise.
As it was previously said, figures within
Table 4 are overestimates of the clearances of the first Fresnel zone. Indeed, by exploiting BotRf, it can be observed that finer estimates, though still rough, of the actual clearances are
for
and
for
(see
Figure 9 bearing in mind that the same color convention of Figure
8 is observed).
Therefore, we deem we can label the whole tests setup as a worst case one (apart from the fact that a marine environment is itself a harsh environment) because of the following reasons:
Antennas on both the sides were not installed by making use of overhead spots in order to limit the complexity related to their installation which was thought to be only a temporary one;
As a direct consequence of the last point, only a limited clearance of the first Fresnel zone (far from being the ) was available;
For the transmission the bigger CR was adopted which might had caused the loss of some packets due to the inability of the Gateways to restore corrupted data.