Besides the possibility to calibrate the receiving antenna factor and other electrical requirements such as shielding the setup and the design of our octocopter “PTBee” (cp. Figs. 1, 3) has to meet additional practical requirements. These include smooth operation, a quick exchange of batteries and general fulfillment of Electromagnetic Compatibility (EMC). On the one hand, motor controller switching, high motor line and battery currents could cause internal EMC problems such as distortions of the magnetic sensor to control the yaw angle, but will also interfere with the fields to be measured, especially at frequencies up to a few hundreds of MHz. On the other hand, we plan to fly in electromagnetic harsh environments. Therefore, shielding against high power RF external sources is necessary to protect the UAS itself. Of course, shielding against external RF radiation will also reduce the emissions from the UAS itself. Since payload is always limited on flying platforms, the shielding has to be of light weight. To reach the defined safety limits, all electronic instrumentation need to be encapsulated by the shielding, including motor controllers, flight controller, navigation controller, data sampling and storage unit, etc. To prevent overheating of the electronics inside, a mesh shielding has been designed that allows for some air flow through it. Another advantage of the mesh shielding is that the internal barometric pressure sensor of the UAS contributing to the height information relative to the ground level is not affected.

As the battery is covered by the shielding, the design has to allow a quick exchange of it. A low-weight aluminum frame with top and bottom plates and with mesh inserts forms the outer body of the shielding that takes all of these conditions into account. The top plate also serves as mounting pad for the different sensing heads and the GPS antenna. The inner part of the bottom plate can be unlocked and released, thereby giving access to the battery compartment. The outer ring of the lower plate serves as support for the antenna of the remote-control receiver, the downlink for the flight status to be received by a smartphone application, and the uplink for the RTK correction signal. All input and output signal paths are filtered using narrowband bandpass filters.

For simplicity reasons, the sensing head is typically mounted on top of the UAS, which moves the center of gravity of the UAS. Test flights with simple ballast instead of the electronics have shown, that the UAS can handle its payload even in windy conditions. The octocopter setup seems to be more stable compared to UAS configured with a smaller number of motors/propellers, especially in case of motor/controller failure. For some applications, e.g. when the measurement signal directly comes from “down under” the sensor is mounted underneath the UAS, using an updated calibration factor.

One of the advantages of UAS is their ability to approach predefined waypoints (WP) (cp. Fig. 4), a locus in space (WGS84 format and height above ground – not absolute height), and to trigger some action when the WP has been reached. In our case, these WP are defined using the software MK-tool provided by HiSystems GmbH (Fig. 2). The tool uses georeferenced satellite images as maps. Once defined, the WP data are transmitted to the UAS with all necessary parameters such as horizontal and vertical (ascending/descending) speed, waypoint radius (WPR), dwell time at the waypoint, steering of the UAS towards a fixed cardinal direction or towards one or a set of predefined points of interest (POI). After the automatic start procedure, the microcopter floats to the first WP. The predefined WPR between 1 and 10 m determines the region in space around a WP, where the navigation controller assumes that the given waypoint has been reached. The manufacturer recommends a value of around 10 m for this parameter. Thus, this point reached by the microcopter may be apart from the desired point in space by a few meters, depending on the wind conditions, the actual payload and some parameters of the flight controller itself, and the actual satellite constellation. The typical accuracy of the GPS and GLONASS signals without differential GNSS improvement (d- GNSS) or real-time kinematic (RTK) will only enable accuracy of a few m in lateral position, but also for the altitude (Morgan-Owen and Johnston, 1995; Shao and Sui, 2015). We have tested the relative accuracy by repeated flight tests using the same flight pattern (cp. Fig. 5 flypath 1 and 2). The weather conditions were almost perfect – i.e. it was a sunny and not windy day (2–3 Bft). Flight tests were performed on our open area test site (OATS), which is a 50 m by 60 m size metal reflector on the ground with at least 150 m clearance around it. Thus, a maximum number of satellites can be received. For simplicity, we have chosen a circular track with 100 m diameter in 20 m height. The WPR was set to 10 m (cp. the red circles in Fig. 5). When the UAS realized that the WP has been reached, it hovered there for 10 s and then continued its flight to the next WP. The time needed to catch the WP was almost negligible. Next, we have decreased the WP to 3 m. It took the UAS quite a while (about 15 s) to recognise the actual position being located inside the waypoint radius though it did not change its position. Further reduction of the WP radius caused the octocopter to stay at its position trying to reach the first WP. Especially under windy conditions and with high payload this localisation was not acceptable.

In order to be able to reduce the waypoint radius without sacrificing the time to reach the desired WP, the single frequency GPS receiver u-blox LEA 6S was removed and replaced by a OEM615 (NovAtel, 2013), which is a state-of-the-art dual frequency airborne receiver. Its update rate for the actual position is improved to 20 Hz compared to 5 Hz before. Some additional hardware had to be designed in order to integrate the new receiver with the existing electronics (Fig. 6). Because u-blox and NovAtel use different protocols for transmitting the actual position information, the software of the navigation controller was also modified to accommodate a new protocol parser. Besides the signals from GPS and GLONASS satellites the NovAtel receiver uses additional differential correction data in RTCA or RTCM format from a NovAtel ground station (receiver model OEM628; NovAtel, 2013). The latter acquires samples from the GNSS satellite information for a long observation time, e.g. 24 h or even more. This significantly reduces the uncertainty of the position of the ground station receiver, until it has reached a lateral uncertainty of position of a few cm and a height accuracy of better than 10 cm (Fig. 7). From this known position and the actual GNSS information the ground station receiver calculates the offset and transmits the error correction using a radio link with an update rate of 1 Hz. The onboard receiver of the UAS takes this correction signal together with the actual GNSS data to calculate the updated position as input for the navigation controller of the UAS.

A test of the NovAtel ground station and a stationary operated UAS showed the drastic improvement in localisation close to the theoretical limit. Further tests with payload and under windy conditions will reveal the practical limit of the advanced positioning.

During its operation the UAS transmits the actual position of the UAS including the height, the time stamp, and the validity of the measurement data using another radio link. This flight log data can be displayed in real-time on an Android-based smartphone, for which an application has been written.

As already mentioned, the applications for the UAS-based platform are quite versatile including the measurement of radio frequency (RF) signals. For every application different RF frontends will be developed, which can be connected to the data processing unit using the same interface (Fig. 8). Until now, a superheterodyne receiver design is used (Pozar, 2005). All receivers provide a 50Ω input connector for the antenna. The RF signal is amplified, filtered and downconverted to an intermediate frequency (IF) of 70 MHz using a local oscillator signal from the data processing unit. The IF signal is then filtered by a Surface-Acoustic-Wave (SAW) filter, to suppress undesired mixing products and adjacent RF-channels. The IF-signal is amplified and can be measured at a 50Ω output port. The IF signal is sampled and stored by the data processing unit using a fast analog-to-digital converter (ADC) and a solid-state hard disc as mass storage device. To obtain a higher dynamic range a variable gain amplifier (VGA) is used as an IF amplifier (DVGA2-33+ from Mini-Circuits, http://www.minicircuits.com). The gain can be controlled by the data processing unit via a serial interface. To determine the correct gain, the IF power is measured after the IF filter using a logarithmic amplifier with demodulation (AD8307 from Analog Devices, http://www.analog.com). The device delivers a DC voltage that is proportional to the logarithm of the input voltage level, which then can be measured by the data processing unit using an ADC. Until now, highly integrated amplifiers are used for the receiver providing high gain, low noise figure, and high bandwidth (Fig. 9). Additionally, the receivers design can easily be adapted for new applications. Only a few frequency dependent components like the input and IF filters need to be replaced. However, the power consumption of approx. 300 mW for each receiver is relatively high.

As an example the electrical field strength in free space can be measured by calibrating the whole system, including the antenna, the RF frontend, and the data processing unit, in a known electrical field.

The data processing unit is capable to sample up to 3 received signals simultaneously. Therefore, the electrical field vector can be calculated measuring the field strength of 3 probes where each probe is sensitive for one field component in an orthogonal system. Of course, scalar or 2-dimensional measurements are also possible. A variety of linear antennas has been developed for different frequencies and applications. Measurements have been performed with passive dipoles, active dipoles, passive bowtie antennas and passive monopoles. Because of the limited space on the UAS, electric small antennas are preferred to detect electromagnetic waves up to 1 GHz. For simplicity, the radiation pattern of each antenna should be as close as possible to an ideal, electrical short, linear antenna (Balanis, 2005). Since the radiation pattern is not ideal, a model for real antennas has been developed. The signal of the real antenna is modelled as a superposition of the signals of an ideal isotropic and an ideal linear antenna. The parameters for this model are again determined through calibration in a known electrical field.

It is the task of the data acquisition board to control the receiver frontend and to pre-process and save the RF data on a suitable storage medium. Figure 10 gives an overview of the components.

An “Arria V” FPGA is the central design unit that controls all components and employs an embedded 32bit processor (Altera, www.altera.com). A coherent clock distribution fed by a stable temperature-controlled crystal oscillator (TCXO) drives all frequency-dependent components as local oscillators (LO) and A/D converters. Three receivers can be connected that allow to measure the field polarisation with an orthogonal antenna-triple. The 16 bit-wide A/D converters have sample rates up to 160 MHz. That ensures to sample wideband signals of radar applications within WERAN.

LO frequencies are directly generated on the board using integrated PLL/VCO chips. As stated above, their signals are fully coherent with AD sampling. That simplifies the RF frontends which hence do not need own LOs. Actually, the RF can be provided with two LO frequencies. This enables RF concepts with two intermediate frequencies (IF) which is useful at carrier frequencies of several GHz. The control board of each RF frontend also includes an automatic gain control (AGC) to adapt the RF signal strength to the ADC input. Variable gain amplifiers (VGA) are connected via a serial peripheral interface (SPI).

RF data acquisition is performed by direct IF sampling, i.e. a band pass signal (but not a baseband signal) is obtained. Any necessary signal pre-processing is carried out within the FPGA. However, it is the concept of the WERAN data processing that a nearly unaltered signal is stored on the octocopter platform to be analyzed by subsequent post processing. A high-bandwidth data recording is realized by implementing a SATA-IP (Serial ATA disk) core in the FPGA. It operates at SATA 3-speed (6 GBps) and allows the direct connection of a micro SATA solid-state disk (SSD). A LAN interface allows for the download of the measurement data after landing of the octocopter without removing the SSD from the UAS. The processor interfaces with the main octocopter controller via an I/O interface. Optionally, it can use its on-board GNSS module to obtain time and position information.

A common misconception of the Nyquist criteria is that the sampling frequency must be twice the highest frequency present in the signal s(n). In fact, the sampling frequency needs to be only twice the signal bandwidth to prevent aliasing: fS>2Δf

Undersampling is corresponding with the use of a sampling frequency which is less than the highest frequency present in the signal. In Fig. 11, from (Zumbahlen, 2008) this principle is illustrated in frequency domain. Signals below 0.5fS are located in the so-called first Nyquist Zone (NZ), see panel a. Sampling these signals preserves their original carrier frequency. The number of NZ increments by every 0.5fS. Frequencies higher than half the sampling rate are folded back into the 1st NZ. Hence, a signal located in any higher NZ will give an image in the first NZ (panel b, c). No signal information is lost except for the value of the original carrier frequency fC. An additional frequency reversal occurs, if signals are located in even Nyquist zones (panel b).

If equation fS=4fC2NZ-1 applies, then the image of fC is safely placed in the center of the first NZ (=0.25fS; Zumbahlen, 2008).

Then the primary function of an anti-aliasing filter is to ensure that the band of sampled signals must not overlap any multiple of fS/2, i.e. it is limited to a unique Nyquist zone. In the WERAN RF frontend designs a steep band pass filter is placed in the IF section. The VOR RF frontend offers a 70MHz IF. After further amplification on the digitizer board, this signal is then initially sampled at 105 MHz. This converts the IF into the second NZ, resulting in an image in first NZ at 35 MHz. A chain of FIR (Finite Impulse Response) decimation band pass filters inside the FPGA then reduces the sampling rate dramatically, but finally it remains well above twice the channel bandwidth of 25 kHz when stored on the SSD.