Figure  shows the frequency-dependent complex ARC, which is also denoted as ΓH/V(f) in the following, versus frequency for the three AUTs in magnitude. Shown are the measurement results under dry conditions alongside the changed curves as they appeared directly after wetting with water. As expected, significant differences can be observed between the dry and wet states for both polarizations. Narrow band regions are recognizable where the ARC magnitude decreases after wetting and also such in which the ARC magnitude increases. These two types of regions are highlighted by two colours in Fig. , allowing a comparison with previous observations where the effect of a wetness-caused change in the ARC was exploited to estimate WAA and reproduce local rainfall events qualitatively . In the frequency band from 38.425 to 38.575 GHz and for the horizontal polarization which has been considered there, a decrease of the ARC under wet conditions was observed. This is consistent with the present analysis using direct VNA measurements, see Fig. .

Since drop patterns form randomly during the process of moistening the radome, the impact of radome wetness on the ARC was tested with repeated experiments in order to ensure qualitative reproducibility. For this purpose, the spraying procedure was deliberately changed between the individual experiments to strive for maximum randomness in the drop formation processes. The changes in ARC occurring in the repeated experiments were qualitatively very similar, i.e., there were no significant differences in the frequency responses of the ARC in each state of maximum wetness. From this observation, we conclude that the actual drop positions have less impact to the ARC magnitude than the overall water content.

In order to enable a spatial assignment of the observed changes to individual components of the antennas and localization of changes, all ARC frequency domain data was transformed into time domain by inverse fast Fourier transformation (IFFT). Frequency windowing was performed using a 3-term Blackman window wb(f) in order to reduce sidelobes in the time domain. Mapping of the round-trip delay τ to the distance r was done by 1τ↦rτ=c0τ2, where c0 is the vacuum speed of light. Corresponding to the minimum and maximum frequencies for which the considered AUT is specified according to Table , range resolutions from Δr≈5–7.5 cm were achieved. The magnitudes of the resulting full unambiguous range profiles under dry conditions are shown in Fig. . At the very beginning, i.e., close to r=0 which corresponds to the calibration plane at the ends of the coaxial cables, there is a first apparent peak spanning over the immediate vicinity of the antenna. The second pronounced peak is significantly lower in magnitude compared to the response from the nearest range bins. Considering its spatial distance of 7.7 m from the calibration plane, it is assigned to residual reflection from the opposite wall of the chamber. Apart from these, no other relevant contributions are observed in the range profiles despite from the peaks at the ends of the individual unambiguous ranges, corresponding to the natural periodicity of IFFT results of band-limited signals.

The recorded datasets from a series of repeated experiments under the same conditions were analyzed together with the associated range profiles obtained by IFFTs of the frequency-domain ARCs. Exemplary drying processes can be observed over frequency and time in Fig.  and in even more detail in the video supplement . In the color coded polar plots provided here, the radial distance maps to frequency. Time in minutes is represented in azimuth. Color represents the magnitude of 2ΔH/V(f,t)=ΓH/V(f,t)-ΓH/V(f,t=0), which is the complex difference of the individual ARC frequency responses recorded over time t and the response from the very first measurement at t=0 under dry initial conditions. The index H or V corresponds to the measurement of horizontal or vertical polarization. The clearly noticeable discontinuities at some minutes after the respective starts of the measurements correspond to the time where the antenna was wetted manually, i.e., the start of the drying process. As time goes on and the radome dries off, the ARC returns back to its original dry state over the complete measured frequency range apart from some artifacts with negligible magnitudes of about or below -60 dB. The drying times are almost identical in both polarizations, which is to be expected. There are individual frequency ranges in which the ARC returns to its original state sooner than in other ranges.

In a way similar to Fig. , the time evolutions of the magnitude of the corresponding range profiles are shown in Fig. , where the radial distance now corresponds to the distance from the calibration plane, i.e., the coaxial connectors. Color represents the magnitude of 3Δ̃H/V(f,t)=F-1ΓH/V(w)(f,t)-F-1ΓH/V(w)(f,t=0) where the superscript (w) denotes the result after windowing the ARC according to 4ΓH/V(w)(f,t)=wb(f)ΓH/V(f,t), with wb(f) denoting the same Blackman window as it was used for Fig. . Again, the index H or V corresponds to the results for horizontal or vertical polarization. It was observed that the effects of water wetting and also the changes during the drying process for the most part occur within the first two meters, i.e., within close proximity of the antenna. This is acceptable and justified since the drops are obviously sitting directly on the radome. For this reason, only the first two meters of the distance profiles are shown in order not to clutter the graphical representation with irrelevant data. The wall reflection magnitude also shows slight variations over time which could potentially be considered to constitute WAA and path loss information. However, since the wall is not at far-field distance at all, it is not considered to be meaningful for actual WAA fits as they have been done in where a corner reflector at several hundred meters was used. Again, the magnitudes of the complex differences between range profiles during drying and the very first range profile, i.e., under dry conditions, are shown during the time of the drying process and the discontinuities shortly after the start of the respective measurements correspond to the time of wetting the radome. Once more, the drying process is observable from the smooth transition of the end of the drying process into the initial state of dry conditions.

Since it was not possible to keep all atmospheric parameters such as temperature, relative humidity and air convection constant over a longer period of time in the measurement chamber, the experiments were repeated several times in order to investigate how different environmental conditions and inevitably different drop distributions impact the drying process. While the complex ARC always returned to the initial response under dry condition for the three AUTs during drying, the individual drying processes did not always take the same amount of time. Significantly shorter drying times than in Fig.  while using the exact same setup for the individual antennas are demonstrated in Fig. . Beyond slight temperature and humidity variations, this is attributed to different particular outcomes of the drop pattern formation processes. Specifically, the effect of drops running off during or immediately after the spraying process can remove different amounts of water before observing the ARC during drying. The impact of this effect ultimately depends on the spraying procedure and the associated drop formation process which is difficult to predict.

The significant differences in the respective drying times in Figs.  and , however, clearly emphasize that the drying process of a wet CML antenna strongly depends on a manifold of parameters even under laboratory conditions. Operated under real conditions, a CML is exposed to many more and also larger variations of environmental parameters including temperature, humidity and wind speed which all have diverse influences on the drop formation as well as drying processes. It is obvious that predicting or estimating WAA in this situation will, most probably, remain a challenging, if not impossible task to master perfectly. However, with ARC information available, the overall wetness state of an antenna radome can be captured well for the majority of frequencies, even if this could bring the necessity of occasional ARC calibration to account for inevitable drifts of the entire antenna system, e.g., degradation of radome surface constitutions over time.