Passive millimeter wave imagers for safety and security

Passive millimeter wave (PMMW) imaging technology offers great potential for modern sensing, scanning, and security needs. In security screening equipment, PMMW can see through a variety of clothing and baggage materials and is effective in detecting both metallic and non-metallic threats, such as plastic and ceramic weapons. PMMW can also provide all-weather imaging capabilities for air vehicles because the problems associated with atmospheric obscurants are negated at millimeter wavelengths.

The millimeter wave band extends from approximately 30 to 300GHz, and a wide range of prototype imagers have emerged in this band over the past 10 years. However, none have dominated the market due to high technology costs, poor sensitivity and image resolution, and the large physical volume of the necessary lenses, mirrors, and mechanical scanners. Furthermore, the small number of users of this spectral band means that a decrease in the cost of this technology is unlikely in the near future.

The implementation of novel aperture synthesis techniques¹ and the falling cost and improving performance of high-speed digital signal processors and microwave communication receivers offer an opportunity to break the deadlock in the development of passive imaging technologies operating below 40GHz.

In aperture synthesis, electric fields are sampled over an aperture and then processed electronically into an image. Eliminating focusing lenses, mirrors, and scanners can reduce the imager volume by orders of magnitude, and the PMMW imager is essentially reduced to having only planar (two) dimensions (see Figure 1).

Interest in aperture synthesis began in the 1940s, when large collection apertures were used to create images for radio astronomy at tens of megahertz, which required many hours of signal integration.² The technique has since dominated radio astronomy, and more powerful technology has enabled more channels, better image resolution, and higher frequencies to be used, as well as shorter operating times. Furthermore, digital technology can offer greater versatility to the technique, but until recently the technology to build a real-time PMMW imager using aperture synthesis was not available.

Figure 1. A mechanical scanning imager has a quasi-optical beam former before the receiver electronics (left), while the aperture synthesis imager has a digital beam former after the receiver electronics (right). As the antennas, receivers, and digital beam former can be packaged into a thin electronic substrate, the aperture synthesis imager offers a massive footprint reduction and can be integrated into walls for security screening and the skins of air vehicles for poor weather operations.

Aperture synthesis involves distributing an array of antennas and receivers over an aperture and correlating electric fields at each of these locations with the field at every other location. This is done at a rate of twice the radio frequency bandwidth (satisfying the Nyquist criterion) and the results are accumulated for an integration time. Therefore, for an array of n antennas distributed over an aperture, there are n(n-1)/2 correlations, forming a 2D spatial function.

In classical aperture synthesis, the far-field image is generated by taking the Fourier transform of the accumulated correlation function. In near-field PMMW imaging, the transform is more general and is described by a matrix multiplication, whereby a calibration matrix is evaluated by sweeping a noise source in front of the imager.

For security screening, the technique simultaneously creates a number of image planes, which means that all objects, regardless of their position in front of the imager, will be in focus. This is effectively voxel (analogous to pixel) imaging, which enables a 2D image of the subject to be created from any direction by collapsing one dimension of the voxel image. This approach offers a significant advantage over conventional quasi-optical imaging, where the depth of field (typically 5cm) requires subjects to remain motionless for several seconds during image acquisition. The closeness of the subjects to the imager also means they can be imaged with a spatial resolution sufficient for security screening, typically ∼1cm at low frequencies (<40GHz), significantly minimizing hardware costs. Furthermore, a moving subject can be imaged without blurring, as the electronic beam-forming process takes place on a time scale faster than that of physical movement, greatly improving the flow of people through the security screening portal.

The radiometric sensitivity of passive millimeter-wave imagers needs to be in the region of 0.1–1K to enable the detection of non-metallic threats for security screening and a range of obstacles for all-weather flying aids. The sensitivity of an aperture synthesis system is given by equation 1, where T_A is the scene radiation temperature (close to ambient at 300K), T_N is the receiver noise temperature (typically ∼200K below 40 GHz), B_RF is the radiation frequency bandwidth, t_F is the integration time, and F is the fraction of the aperture which is filled with receiving antennas. The fraction F is given by equation 2, where n is the number of antennas, A_ANT is the effective collection area of a single antenna, and A_SYN is the total aperture area over which the antennas are distributed.

   (1)

     (2)

For indoor use, the scene contrast is typically 15K, which is the approximate radiation temperature difference between the building interior (set by air conditioning, for example) and the human body. For outdoor use, contrast is typically 150K in the lower frequency part of the millimeter wave band, which is the difference between ambient temperature and the cold sky.

Outdoor contrast often falls with increasing frequency as the atmospheric absorption rises. Understanding these sensitivities and contrasts for particular scenarios enables effective systems to be designed for specific applications.

The number of Nyquist-sampled pixels in an aperture synthesis image ranges from n² for a zero redundant array of n antennas to 4n for a fully filled aperture (F=1). A zero redundant array is a highly sparse array that uses the smallest number of receiver channels possible to reproduce, alias-free spatial information in the scene up to the diffracted limit allowed by the aperture size. The fully filled array uses the highest number of receiver channels that can be fitted into the aperture and offers the highest radiometric sensitivity. In practice, the choice of array geometry and its filling fraction will depend on the radiometric sensitivity and spatial resolution required for the particular application.

We developed a real-time aperture synthesis PMMW demonstrator to investigate the optimal calibration, operating trade-offs, and stability of this type of imager.³ Satisfying the Nyquist criterion on the temporal sampling in the receiver channels requires the rate of mathematical operations in the correlator to be 2n²B_RF. However, as radiometric emission is essentially broadband noise, digitization and correlation can be completed in single bits, and so these tasks can be performed using commercially available field-programmable gate arrays (FPGAs).

A diagram of the complete demonstrator system is illustrated in Figure 2. The sparse antenna array, designed from 32 individual 2×2 patch antennas on a hexagonal grid to minimize aliasing associated with sampling across aperture, is illustrated in Figure 3.

We also developed waveguide horn antenna arrays so that comparisons could be made between the two antenna types. The receivers are sensitive at 22.5GHz with a 300MHz radiation bandwidth and are packaged in dual-channel slimline modules, as illustrated in Figure 4. An antenna array with a filling factor of 34% delivers a 20×20 pixel image and has a 0.5K sensitivity at video frame rates of 25 frames per second. In-phase and quadrature digitization at several hundred megahertz is achieved using two stages of heterodyne down-shifting. The system has flexibility to investigate different array geometries, ranging from those with a sparsity of less than zero redundant arrays in two dimensions to fully filled arrays in one dimension.

Figure 2. Diagram of the 32-channel real-time aperture synthesis PMMW imager demonstrator system (top) and one of the single-channel front-end heterodyne receiver channels (bottom).

Figure 3. The sparse antenna array for the demonstrator made up of 32 individual 2×2 patch antennas. Each patch antenna is only 2mm thick.

Figure 4. Pairs of the 32-channel receivers are integrated into slim-line (5mm thick) dual receiver modules. A circuit board of one of these shown here is 1.5mm thick and 6cm long with two sockets for antennas as illustrated on the left.

The opportunity now exists to develop a PMMW imager architecture that is based on the new aperture synthesis technology and occupies a fraction of the volume of conventional quasi-optical imagers. Our demonstrator project represents a test-bed for the optimization of receivers, antennas, and their array geometries. Prototype imagers based on this architecture are likely to follow, typically with 300 receiver channels, replacing the FPGAs with application-specific integrated circuits and using monolithic microwave integrated circuits in the receivers to minimize volume and power consumption.

Furthermore, the antenna array, receivers, and digital electronics can be integrated into a substrate, a centimeter or so thick, using production techniques to minimize costs borrowed from the mobile communications and computer industries. Imagers can then be integrated into confined spaces to form security screening portals, developed into hand-held wands for proximity screening of people and baggage, or, as depicted in Figure 5, integrated into the skins of air vehicles for all-weather flight.

Figure 5. Examples of PMMW imaging systems using digital aperture synthesis technology.

The financial support for this work is gratefully received from the United Kingdom Technology Strategy Board.