Principal Investigators: Prof. Dr. Ilona Rolfes, Dr. Christian Schulz , RUB

Achieved results and methods

To answer the research questions of M02 in the 2^nd phase, we have conducted research in several areas. This includes research on optimized simulation approaches using Physical Optics (PO) and Complex Source Beams (CSB) as well as the investigation of environmental effects such as temperature and humidity. Furthermore, complex media such as flames and smoke are analysed, using highly stable phase measurements.

Physical Optics

An asymptotic simulation framework based on the application of the PO method was implemented, as shown in Fig. 1.  One of the goals of the framework is to simulate quasi-optical measurement setups in terms of their alignment and error tolerance. The spatial expansion in combination with high frequencies represents the performance limit for classical full-wave simulations. For this reason, the method of PO is used here. Typically, this works under the assumption of far field conditions and approximates Maxwell’s equations, but at the same time provides good quality results for increasing frequencies. Due to the various assumptions and approximations, the simulation can also quickly yield good results for medium-power systems. The framework thus offers a high degree of flexibility, which is complemented by the various options for setting up a simulation scenario. For example, it is possible to import and integrate complex geometric objects from computer aided design (CAD) tools as.stl-files. Alternatively, simple geometric structures can be described and integrated analytically. The same applies to the field sources used. The antennas can be pre-simulated with a commercial field solver such as CST Studio Suite and then be imported or the analytical description of a field source can be used. Furthermore, CSB can be applied, which represent compromise between pre-simulation and analytical formulation. In this way, the scene can be set up by the user and the simulation sequence is defined in the next step. This is necessary because the framework only simulates the path defined by a sequence during the simulation. This makes it possible to view different propagation paths separately. 2D field monitors or the scattering parameters of reflection and transmission can be used to evaluate the results. For quasi-optical measurement scenarios, the field monitor can be used to analyze the correct positioning of antennas and reflectors or whether the structure of the antenna should be further optimized. In this way, it is possible to make predictions about a possible re-adjustment of the setup, which can then be directly adapted and simulated again, so that the setup can be further optimized iteratively before it is build up in the laboratory. This saves time and is much more resource-efficient than moving and repositioning the equipment repeatedly during the adjustment process [1][2].

Complex Source Beams

Small-scale effects are of great importance in THz systems. Due to the small wavelength, almost any material exhibits small-scale features, e.g., surface roughness, which leads to small-scale scattering effects. These shortcomings can be addressed by dividing the simulation domain into a set of disjoint domains. Each domain describes a single scattering object such as a lens, a mirror or a material-under-test. For the simulation of complete transmission scenarios at THz frequencies, it is not efficient to perform full-wave simulations, such as FIT used by CST Studio Suite with respect to the required computing and processing capacity. In addition, the accuracy of far field based methods like ray tracing degrades significantly in near field scenarios. Thus, we investigated the CSB method to simulate electromagnetic wave propagation at millimeter wave frequencies, which is especially suited for simulation of short-range scenarios, where far field assumptions may not hold. With CSB, the near field of a scattering object is divided into several sources, which together represent the further wave propagation. In order to obtain convincing results with regard to an optimum calculation time, the electric field of a simulated scattering object is usually sampled with one tenth of the wavelength. In comparison to the classical PO method, the CSB incorporates directive source elements, which are chosen as Gaussian like beams, instead of spherical waves. Consequently, objects with a high directivity can be simulated more efficiently, leading to a lower sampling rate. The workflow of the CSB method can be seen in Fig. 2 for an exemplary measurement setup. A comparison with a full wave simulation is shown to validate the accuracy of CSB. In the example, the E-field simulated by the CSB method shows high agreement with a full wave simulation, especially in the region of the main lobe. We have seen an amplitude offset at the peak of maximum 0.3 dB. The results are valid in the near field and in the far field [3] [4].

To calculate the field distribution on a scattering body using the PO, N x M calculations are required, where N is the number of sampled near-field sources and M is the number of points to be calculated on the scattering body. The classical PO has a sampling rate of 0.1 of the wavelength, while the CSB method has a four times larger distance of sampling points, resulting in a lower sampling rate of 0.4 of the wavelength.The number of calculations for the CSB is thus reduced to roughly N/(4^2) x M, taking into account the sampling over a two dimensional enclosing surface wrapped around the antenna.

As antennas with a high directivity are mostly used in the THz frequency range, these can be represented with good agreement by a gaussian beam. Therefore, instead of describing the antenna with several Gaussian beams,  we have developed a new method to take only one CSB in consideration. For this purpose, the radiating field of an antenna is first simulated with a full wave simulation. Then, the field data is used in an optimization procedure to find the CSB that best matches the field data. A comparison between the CSB generated E-field and a full wave simulation shows that the results in the area of the main lobe show a high agreement, both in the near and far field range [5]. Furthermore, the phase progression also agrees well, whereby it was possible to show that the CSB generated E-field deviates from the full wave simulation by the size of the numerical dispersion and is thus, even more precise in the phase. With this analytical description, it is no longer necessary to sample a near field source, which consequently reduces the number of calculations to 1 x M for the calculation of the field distribution on a scattering body.

Environmental Effects

A key part of M02 dealt with various environmental factors that can influence the measurement signals and the measurement system itself. The focus was on two main areas: temperature and humidity on the one hand and flames and smoke on the other hand.  The procedure in the 2^nd phase consisted of a parallel workflow in which measurements were carried out in controlled, adjustable environments. At the same time dielectric models for our simulation environment were developed. In both cases, the impact of changes in the surrounding gas (MUT) is investigated. Since natural dielectric changes, such as those caused by humidity or temperature variations, are minimal, a very sensitive measuring methodology is required. Figure 3 shows an overview of the measurement procedure. Fixed target measurements are carried out with FMCW radar sensors. A reference measurement is taken in which the surrounding gas is in a defined initial state. The intermediate frequency signal of the radar sensor is evaluated via absolute values and phase values in order to enable a more sensitive measurement. In the next step, the gas in the measurement path is changed in a defined manner (e.g. change of temperature or fire) and the resulting change in dielectric properties can be observed and evaluated in the measured phase [6][7]. Furthermore, the compensation of sensor movements were investigated in frame of short-range FMCW imaging applications in cooperation with M04 [8]. Motion compensation in the context of radar imaging is usually related to the correction of deviations from an ideal trajectory. In contrast, in [8] we presented a method to take the sensor movement during a single FMCW ramp into account and therefore, addressed the effects caused by a continuous motion during the transmit/receive process. Hence, faster movement can be achieved during the scanning.

Temperature and Humidity

The measurement procedure explained above has been used to investigate changes in temperature and humidity. A climate chamber is used for this purpose, which can be accessed via opposite openings. One entrance was equipped with a dielectric window to enable the electromagnetic signals to be coupled into the chamber. The other opening was equipped with a heatable metal plate in order to have a fixed target. A W-band radar with a bandwidth of 20 GHz was used to measure two scenarios: First, the temperature was varied at a constant relative humidity and second, the relative humidity was varied and measured at a constant temperature. Figure 4 shows an example of each measurement scenario. In order to validate the measurements, various dielectric modeling approaches were investigated and compared at the same time. Using dielectric mixing theory, the gas was modeled as a mixture of air and water. As a second approach, the gas was modeled from the perspective of molecular spectroscopy, whereby each individual atom and molecular group are examined on the basis of their resonance behavior. Finally, an empirical formula from the International Telecommunication Union (ITU) was used. Figure 4 compares these three different models with our measurements. In general, it can be said that a linear increase in temperature causes an exponential behavior in the relative permittivity. A linear behavior can be observed for a linear change of the relative humidity.

Flames and Smoke

Together with M05, standardized test fires were investigated in the Heinz-Luck fire laboratory using W-band and D-band radar sensors [9][10]. At two different heights, the influence of flames and smoke could be measured separately. When the combustible is burned, various chemical components are released which rise due to the high temperature. This behavior can be observed very well in the measured phase of a highly stable radar signal. In Fig. 5, the measured phase is displayed over a period of approx. 700 seconds. The first measurement block in dark blue corresponds to the reference measurement and shows a stable measurement with little fluctuation. If the fire is now ignited, there is a large phase fluctuation and the phase increases on average. If the fire stops, the fluctuation decreases and approaches the behavior of the reference measurement. Depending on which fuel is burned, the strength of the phase fluctuation and the increase in the average phase change. So the different test fires can be distinguished by the measured phases. In the same step, dielectric models were investigated which describe the relative permittivity of the different gases. It became clear that a pure consideration of flames as ionized gas is not sufficient and that a complex model must be created that includes influences such as changing gas chemistry, temperature and ionization.

Data augmentation

In [11] we have demonstrated how the integration of GANs within our simulation concept can further augment scenario diversity. As a qualitative evaluation reveals, various scenarios can be generated in a quick and simple, yet realistic manner. Consequently, the diversity of corresponding training data is also augmented indirectly.

A novel technique for condition monitoring with FMCW radar has been demonstrated in [12][13][14]. Here, knowledge-based and data-driven approaches are combined synergistically for distinction between different conditions. For this purpose, distinctive and highly interpretable features are manually extracted from range-Doppler radar data. Thereafter, a classification is performed applying state-of-art methods from machine learning, such as eXtreme Gradient Boosting (XGB). As experimentally evaluated with 68−92 GHz FMCW range-Doppler measurements, commonly encountered conditions can be reliably distinguished from each other despite a very small training dataset. Fully exploiting the 24 GHz bandwidth of the radar system, yields up to 97.69%.

 Selected project-related publications

1. Garten, C. Statz, S. Gerling, J. Jebramcik, J. Barowski, D. Plettemeier, I. Rolfes, "Considering Nonsurface Scattering in Physical Optics Approximations," IEEE Trans. on Antennas and Propagation, vol. 69, no. 8, pp. 4798-4807, Aug. 2021, doi: 10.1109/TAP.2021.3060043.
2. Sheikh, A. Prokscha, J. M. Eckhardt, T. Doeker, N. A. Abbasi, J. Gomez-Ponce, B. Sievert, J. T. Svejda, A. Rennings, J. Barowski, C. Schulz, I. Rolfes, D. Erni, A. F. Molisch, T. Kürner and T. Kaiser, "THz Measurements, Antennas, and Simulations: From the Past to the Future," IEEE Journal of Microwaves, vol. 3, no. 1, pp. 289-304, Jan. 2023, doi: 10.1109/JMW.2022.3216210.
3. F. Schenkel, C. Baer, I. Rolfes, C. Schulz, “Plasma state supervision utilizing millimeter wave radar systems,” Int. Journal Microwave Wireless Techn. pp. 1001-1011, 2023, doi:10.1017/S175907872200143X.
4. Schorlemer, C. Schulz, N. Pohl, I. Rolfes and J. Barowski, "Compensation of Sensor Movements in Short-Range FMCW Synthetic Aperture Radar Algorithms," IEEE Trans. Microwave Theory Tech., vol. 69, no. 11, pp. 5145-5159, Nov. 2021, doi: 10.1109/TMTT.2021.3108399.
5. Schenkel, T. Schultze, C. Baer, I. Rolfes, C. Schulz, “Radar-Enabled Millimeter-Wave Sensing of Flame Interactions”, IEEE Trans. Instrumentation and Measurement, 2024, accepted.
6. Schenkel, T. Schultze, C. Baer, I. Rolfes, C. Schulz, “Radar-Based Smoke Detection at Millimeter Wave Frequencies: Experimental Study”, IEEE MTT-S International Microwave Symposium, Washington DC, USA, 2024, accepted.
7. Schmitz, M. Vogt, M. Roitzheim, M. Hammes, J. Barowski, I. Rolfes and C. Schulz, ”Radar-Based Condition Monitoring for Enhanced Efficiency and Safety in Industrial Processes”, 11th International Radar Symposium (IRS), 2024, accepted.
8. Sheikh, Y. Zantah, I. Mabrouk, M. Alissa, J. Barowski, I. Rolfes, and T. Kaiser, "Scattering and Roughness Analysis of Indoor Materials at Frequencies from 750 GHz to 1.1 THz," IEEE Trans. on Antennas Propagation, vol. 69, no. 11, pp. 7820-7829, Nov. 2021, doi: 10.1109/TAP.2021.3076577.
9. Batra, J. Barowski, D. Damyanov, M. Wiemeler, I. Rolfes, T. Schultze, J. C. Balzer, D. Göhringer, T. Kaiser, "Short-Range SAR Imaging from GHz to THz Waves", IEEE Journal of Microwaves, Apr. 2021, doi: 10.1109/JMW.2021.3063343.
10. Starke, J. Bott, F. Vogelsang, B. Sievert, J. Barowski, C. Schulz, H. Rücker, A. Rennings, D. Erni, I.Rolfes, and N. Pohl, “A compact and fully integrated 0.48 THz FMCW radar transceiver combined with a dielectric lens,” International Journal Microwave Wireless Techn. pp. 1–12, Dec. 2023, doi:10.1017/S1759078723001368.