Committee-Led Measurements of Advanced Imaging Technology Millimeter Wave Scanners at U.S. Airports
The committee authoring this report also conducted an independent study to measure the power density of millimeter wave radiation emissions inside and outside of operational advanced imaging technology (AIT) millimeter wave scanners at four airports. This study, discussed in this chapter, greatly expands upon the previous laboratory and simulation studies discussed in Chapter 4 by performing (at airports) radio frequency (RF) time-domain power-density emission measurements of the L3 (formerly, L-3 Communications Holdings) ProVision ATD (where ATD stands for automatic target detection) and ProVision 2 airport scanner systems currently deployed at airports by the Department of Homeland Security (DHS), Transportation Security Administration (TSA). As discussed in Chapter 2, these AIT units function by actively emitting electromagnetic waves at frequencies in the tens of gigahertz to construct images used to identify concealed objects. Thus the primary focus of this study was to collect realistic emission data within the primary emission operating band of 24 to 30 GHz over a wider area than is typically occupied by a human during ProVision scanner operations. These measurements were compared to the published maximum permissible exposure (MPE) guidelines of several agencies (see Chapter 3). To give confidence in the accuracy of the “infield” measurements, this study performed a more thorough analysis, including modeling and predictions based on system specifications, in support of the airport measurements.
After releasing a request for proposals, the committee selected BerrieHill Re-
TABLE 6.1 Airport Category, Total Advanced Imaging Technology (AIT) Scanners, and Typical Throughput for the Year 2015
|Airport Security Category||Total AITs in USA||Total AIT Throughput in USA||Average Throughput per AIT per Year||Average Throughput per AIT per Month||Average Throughput per AIT per Day|
TABLE 6.2 Location, Type, and Serial Number with Total Advanced Imaging Technology (AIT) Scanners Throughput of ProVision Machines Tested
|Airport Location for the Measured AIT||Date Measured||L3 Technologies ProVision Type||ProVision Serial Number||Total AIT Throughput|
|Dayton, OH (DAY) —Category I||2/1/2016||ProVision 1 (AIT-3)||L100104200824||522,275|
|Cincinnati, KY (CVG) — Category X||3/8/2016||ProVision 3 (L3 AIT FIS)||L100120301084||218,404|
|Tyson, TN (TYS) — Category II||3/9/2016||ProVision 1 (AIT)||L100113200889||1,785,344|
|Memphis, TN (MEM) — Category I||3/11/2016||ProVision 2 (FIS-AIT)||PV-20275||403|
search Corporation (BRC), with subcontractor Applied Research Associates (ARA), to provide technical design advice, test planning, test execution, data reduction, and technical reporting for the ProVision scanner measurements performed at four airports selected by the committee from a longer list of TSA-chosen airports—Dayton International Airport (DAY) in Ohio, Cincinnati/Northern Kentucky International Airport (CVG) in Kentucky, and McGhee Tyson Airport (TYS) and Memphis International Airport (MEM), both located in Tennessee. Table 6.1 identifies airport security category,1 total AITs, and typical throughput for the year 2015, while Table 6.2 identifies the model and serial number of ProVision scanner systems measured and their total AIT throughput.
While the TSA and L3, the company that produces the scanners, provided
1 A security category is a designation given to an airport to mark the amount of traffic flow, security strategic importance, and personnel need. Airports with the “X” designator have achieved 100 percent AIT millimeter wave scanners for screening at all lanes at all check points; category “I” is typically reserved for an airport located in a small community near a large city, while airports located in smaller communities not near a major city are given a “II” designation. Airports located in even smaller communities with a low amount of traffic are usually given a “III” designation.
crucial physical configuration and proprietary RF waveform data in order for the committee and contractors (BRC/ARA) to properly design and prepare for the field measurements at the airports, they were not involved with any of the data acquisition, data processing, or data reduction activities. The remainder of this chapter summarizes the test plan and technical details of the RF measurement approach, measurement results, measurement uncertainty, and prediction results.
Characterizing the RF emissions of ProVision machines is not trivial with signal strengths close to the noise levels. Even though they are active systems, the expected levels of RF emission are very low by design, so detailed measurement plans were validated to assure that this set of ProVision characterization measurements met the strictest criteria of data quality. While the physics and engineering of microwave and millimeter wave regions of the electromagnetic spectrum are quite mature technologically, there are not many widely used applications as sophisticated as the L3-Comm millimeter wave AIT scanner. For instance, while Wi-Fi commonly operates near 5 GHz, few widespread commercial products operate at higher frequencies. Because general-purpose test equipment required to detect emissions in the 24 to 30 GHz region is sophisticated and expensive but still not suitable for the unique test conditions with the portals, specialized equipment was designed, constructed, and tested for this study.
The ProVision scanner systems were designed for high throughput at airports, thus a passenger only has to stand in a stationary position during the short scan period of 1.3 s to provide the data needed to screen for concealed objects. However, during this 1.3 s period, a complex and technical sequence of events occurs. Figure 6.1 is a scaled photograph of the interior volume of a ProVision 2 unit from a top view perspective (i.e., looking down at the floor from the ceiling of the unit). In contrast to the ProVision ATD, this more recent model has a reduced physical footprint and lower overall weight to accommodate more airports. The RF and millimeter wave subsystems for data collection are identical, but the passenger “chamber” for the ProVision 2 is about 1 in. (i.e., 2.5 cm) smaller in overall diameter. The imaging subsystem is based on a scanning cylindrical geometry, and to properly characterize a passenger, two separate imaging arrays are installed on revolving masts in the ProVision, one facing the passenger’s front side and a second complementary system facing the passenger’s rear side (also see Figures 2.6 and 2.7). Each of the masts consists of two vertical linear arrays of ~200 receive and transmit horn notch antennas, approximately 1 cm × 1 cm each. One of the arrays sequentially transmits the “chirp” waveform from top to bottom, and the second ~200-element array listens for the reflected response from the passenger. The rear array executes the very same function but is offset in time by 1 microsecond (µs)
from the front array to prevent the two transmit arrays from interfering with each other. The linear arrays are on a mechanical rotation system that revolves the mast through a span of around 120 degrees (either right-to-left or left-to-right, depending on starting position) in 1.3 s. In each array, the transmit-receive pairs are sequentially activated down to capture a vertical data line of close to 400 spatial samples while the masts are in motion. The high number of samples is achieved by using each offset antenna element twice. Each vertical scan lasts ~3 milliseconds and is repeated ~200 times during the scanner rotation cycle. During each vertical scan, each element in the transmit array sequentially radiates a linear-frequency modulated (LFM) ramp or chirp waveform that begins at 24 GHz and ends at 30 GHz. The ramp waveform occurs over a period of ~6 µs of the ramp “on” time, followed by ~2 µs of “off” time when it is not transmitting.
The committee gathered the following information during its study:
- Power density emissions measurements inside the airports’ ProVision units where passengers would stand,
- Power density emissions measurements outside the airports’ ProVision units in the screening area at waiting passenger and TSA operator positions,
- Power density emissions calculations inside and outside the unit based on unit specifications and aided with computational electromagnetic analysis/software to validate the in-band RF power density measurements, and
- Power density emissions or field characterization measurements out of the operational frequency band (executed on a single ProVision scanner at DAY airport).
The challenge in measuring the RF power density in the ProVision systems is that each mast transmits a total ~70,000 frequency chirps/pulse, each lasting ~6 µs, each from a unique location in the cylindrical geometry during the 1.3 s scan time. In addition to this, based on L3 specifications and previous studies, the power density is expected to be on the order of microwatts per square centimeter (µW/cm2), which is very low and requires sensitive test equipment. These dynamic measurements present a challenge in that both the transmitter antennas on the masts and the detector antennas on the test equipment used to measure the power density are highly directional, as discussed in Appendix C (Figure C.4). As a consequence, the test instrumentation detector would only be sensitive to a very limited number of frequency chirps during each 1.3 s scan. Given the highly transient nature of the signals, it would be difficult to map out the power density inside the unit in a timely fashion. Therefore, to collect power density emissions measurements inside the units, the ProVision systems were placed in a state such that the masts were stationary at their midpoints (i.e., directly in front and behind a passenger) and not swept in position. The field probe antenna directly faced the masts, so that during a vertical scan, the transmitting antenna directly opposite the field probe would be very close in direct alignment, maximizing the coupling based on the directionality of the antennas. With the masts locked in this position, all of the transmit and receive antennas function identically to when they are swept under a normal 1.3 s scan. This set-up allowed the field probe to be repositioned to sample the power density within the chamber with the mast at fixed locations while still scanning vertically, as done under normal scanning modes.
Given the extremely limited footprint in the ProVision units, the first challenge was for the BRC/ARA design team to create a specific field probe detection system that would (1) allow a probe antenna to cover the majority of the volume typically occupied by a human subject when the scanner is operated normally; (2) be precise enough in absolute positioning to produce highly repeatable results; and (3) measure with sufficient signal-to-noise ratio to determine the absolute incident power density with high fidelity.
In the interest of design, all measurements were made with a signal power detector instead of an array of detectors. In order to collect power density emissions measurements inside and outside of the airports’ ProVision units, the field probe (see Figures 6.2 and 6.3) was custom-constructed for use on either a ProVision ATD or ProVision 2 portal. A 6 ft linear bearing system supporting a movable mounting plate carries a WR-28 (24-30 GHz) RF waveguide-fed detector antenna. Within the ProVision units, three separate scan planes were identified as measurement planes within the unit volume, referred hereafter as Scan Plane 1, Scan Plane 2, and Scan Plane 3. Scan Plane 1 and 2 are “facing” the forward RF scanner bar, and Scan Plane 3 reverses the geometry of Scan Plane 1 to face the symmetric rear RF scanner bar. Scan Plane 1, 2, and 3 are all performed inside the ProVision machine, and the spatial measurement samples have a realistic coverage of the overall incident-radiated power for a typical passenger profile. Due to the extremely limited space within the ProVision units, a typical Cartesian X-Y field probe scanner was not employed because it simply would not fit within the chamber. Instead, the team designed a
simpler “radial” field probe device that samples along the linear slide access and whose angles are mechanically adjustable between ±30 degrees of vertical within the ProVision volume. As the project developed, extensive “off vertical” field probe measurements for ±15 degrees and ±30 degrees were collected, mostly in Dayton. Most of the measured power levels significantly decreased off the vertical line due to the mast being at a fixed location directly in front of the rig, and many measurements showed power levels below the measurement sensitivity of the equipment. The results reported here were for the vertical “on axis” orientations where the rotational angle (f) was zero degrees, and for which the radiated power is highest. Power measurements at other angles were lower (or undetectable) and consistent with predictions.
Figure 6.4 shows the mechanical scanner built in the BRC laboratory for this project prior to its deployment to the DAY, CVG, TYS, and MEM airports. The field probe device sits on a very thick and stiff steel plate in the shape of a baseball “home plate” so that the mechanical scanner is extremely stable no matter the angular orientation of the linear slide rail. The linear rail can be positioned continuously
from −30 to +30 degrees from vertical using precision off-set set screws, with the position confirmed by a digital level.
The 24 to 30 GHz probe antenna was attached to a precision optical rotating stage such that the ideal antenna polarization orientation was maintained precisely vertical (same at the transmitting polarization of the array on the mast) by counter-rotating the antenna the same angular movement as the overall probe f axis. A close-up of the actual probe antenna (Narda Model V637 standard gain horn) is shown in Figure 6.5.
The ProVision RF power density emissions were collected by a NARDA V637 standard gain horn antenna. The antenna was mounted to a short, curved section of waveguide, then to a waveguide to coax adaptor. A short RF cable connected the feed system to the input side of a Ka-band RF solid-state amplifier, which is independently powered by a ±15 V direct current (dc) power supply. The RF amplifier has a nominal 40 dB gain that feeds a 3 db RF coax power splitter/coupler. For the DAY airport measurement, one output of the coupler is connected to a Schottky
diode detector and the oscilloscope port of the Keysight 8990B time-domain power meter. The second port of the 3 dB coupler/power divider was attached via a longer RF cable directly to the Keysight 8990B power meter head. The Keysight had two power meter input ports and two standard oscilloscope input ports. The original DAY airport plan was to use the RF input port #1 on the Keysight 8990B power meter to measure the ProVision millimeter wave power emissions directly. A second 8990B oscilloscope input port #1 was used to attempt to measure the Schottky detector response, similar to the methods used in the previous Department of Homeland Security measurement report characterization. However, the Schottky detector did not have the necessary sensitivity to measure the RF emissions directly, and no usable data were retrieved with the Schottky diode detection methods in any of the DAY airport tests. The Schottky diode device is simply not sensitive enough for the power density emissions of the ProVision unit in the 24 to 30 GHz range, and therefore all DAY airport data using the Schottky diode detector was indistinguishable from random noise. For the subsequent three measurements (CVG, MEM, and TYS airports), the Schottky diode was not used at all. Fortunately, the first methods using the Keysight 8990B time-domain power meter yielded excellent data (includ-
ing DAY airport) and therefore was used for all airport measurements. It should be noted that the post-airport calibration data for the DAY airport measurements used a different calibration file than the other three airports because there were fewer RF components (and therefore lower losses) in the CVG, MEM, and TYS airport set-ups. A schematic view of the DAY airport field probe system that matches the physical layouts of Figures 6.2 and 6.5 is shown in Figure 6.6. For the CVG, MEM, and TYS airport measurements, only the single Keysight 8990B power meter port was used and the 3 dB power coupler and the additional RF cable were removed, which minimized losses and improved the system sensitivity by approximately 6 dB for the CVG, MEM, and TYS airport measurements.
One aspect of the ProVision experimental set-up required a significant change of the initial field design. Given the complex transient waveforms previously
discussed, the committee and BRC/ARA team originally assumed there was a “synchronizing” or “sync” pulse available that would trigger the Keysight 8990B instrument when to take data. While such a port exists on the ProVision ATD and ProVision 2 units, the sync pulse is only available by physically accessing the mast (moving measurement arm). The DHS/TSA project manager understandably did not want anyone to tie into or tamper with the airport ProVision machines in any way, so an alternative way was needed to synchronize to the ProVision waveforms without physically modifying the ProVision machine itself.
A solution was found by employing a second “fixed” 24 to 30 GHz antenna mounted staring directly at the rear ProVision array, pointed 180 degrees away from the primary probe antenna. The rear-facing sync pulse antenna would capture the initial pulses at the beginning of a linear scan. Since the ProVision scans from top to bottom, the antenna was positioned looking to the rear and pointed to the top of the rear array. This second fixed antenna would feed directly into the Keysight 8990B power meter port #2, and the instrument would sync the forward-looking probe measurements by looking for the first rear scan port. Because the two waveforms are only 1 µs apart in time, this was an elegant, non-invasive RF solution. Figure 6.7 shows how the two horns are interfaced with the data collection equipment. After the DAY airport measurements were completed and analyzed, the additional three airports (CVG, MEM, and TYS) acquired data using two different reference sync pulse methods. In the subsequent three airports, the team repeated the rear reference antenna case as used at DAY airport. The team also acquired data using a front-pointed reference antenna positioned several inches above the primary probe antenna, which allowed slightly higher time-domain sampling of the results (see Figure 6.8).
The projected planar position of the probe utilized a very simple polyvinyl
chloride (PVC) coupler and pipe so that the scan antenna could be projected forward of the scan plane by simply cutting a piece of longer PVC support pipe to the length of the desired extension, minimizing time-consuming field changes during the measurements. Ultimately, the team measured the incident power density at an initial Scan Plane 1 position and repeated the forward tests with the antenna extension removed, effectively moving the antenna 30.48 cm farther away from the front array. Because this induces more “spread loss” as the wave propagates away from the array, the power density emissions at the Scan Plane 2 center position were substantially weaker than the Scan Plane 1 emissions.
To assure complete traceability of the airport emission measurement system and that the system would operate correctly and during the limited time available for the field measurements at each airport, BRC/ARA fully tested the field probe in the laboratory first. Without the availability of a ProVision system for laboratory testing, a detailed RF waveform simulator was designed and built to emulate the ProVision RF emissions (see Figure 6.9). This first step in the signal simulation was a low-frequency, linear-frequency-modulated sweep lasting ~6 µs, followed by 2 µs of “off” time, and repeated the waveform for ~200 consecutive (~8 µs) periods. The baseband signal was swept from 1.0 to 3.875 GHz and fed through an RF up-converter that was mixed with a stable local oscillator (LO) at 11.125 GHz. The output of the mixer was then sent through a bandpass filter to produce a chirp ramp from 12.125 to 15.0 GHz. After a two-stage amplifier, the signal was then directed through a frequency doubler and a WR-28 waveguide bandpass filter, resulting in a linear FM output from 24 to 30 GHz, precisely the sweep time of the L3 ProVision system. This laboratory simulator was driven by an external sweep
trigger source that acts like a master clock. In the field measurements, the Keysight 8990B was triggered to receive by the sync pulse discussed in the previous section.
The L3 ProVision simulator mimicked the emissions of a ProVision system, and BRC executed a successful “over-the-air” test of the entire closed loop system, as shown in Figure 6.10. All waveforms, distances, baseline cable losses, amplifier gains, and equipment interfacing were simulated to assure that the entire field probe system and RF receiver subsystem would work as required in the field measurements at the selected airports. For the airport ProVision measurements, only the receiver equipment was taken into the field, not the emulator. The suspicion that the Schottky diode would not yield usable data in the field test was confirmed in the DAY airport measurements.
Upon testing the field probe system with the simulator, the TSA agreed that a ProVision machine in DAY airport could be used for the first measurement. The DAY airport field test measured the power density emissions for the three scan planes in Figure 6.2, as well as two additional scan planes, defined here as Scan Planes 4 and 5. Scan Plane 4 was envisioned to estimate the incident RF power ex-
iting the ProVision system at the TSA operator’s position, while Scan Plane 5 was envisioned to estimate the RF emissions at the closest distance the “next passenger in line” would see (see Figure 6.11). Figure 6.12 shows several of the scan planes as implemented in the field measurements.
The overall purpose of this measurement effort was to produce calibrated field power density measurements versus frequency in SI units that are widely recognized in RF and millimeter wave measurement standards worldwide. Normally, incident power density is either reported as E-field strength (volts per meter, or V/m), or incident power density, Pd typically reported as milliwatts per square centimeter (mW/cm2) or decibel-milliwatts per square centimeter (dBm/cm2). Because the overall purpose of this experiment was to measure the incident power density, this
value had to be “backed out” based on the overall RF (or millimeter wave) “trail of signal custody” from the field probe receive horn, through the various cables and modules, to the Keysight 8990B time-domain power meter. To calibrate what was measured, the Friis Transmission equation was used:
In Equation 6.1, Prcv is the received power at the field probe receive horn, Pt is the ProVision transmitted power, Gr and Gt are the receive and transmit antenna horn gains, and R is the distance between the antennas. The committee assumed no prior knowledge of the exact ProVision transmitted power (Pt) nor the ProVision transmit antenna gain (Gt), but the received power can be measured through the Keysight 8990B time-domain power meter as attached to the field probe RF network. Essentially, the linear mathematics of converting the instrumented received power detected, P8990B, to received power, Prcv, is straightforward.
|P8990B(f) = Prcv(f)· NGainrf#1(f) [W]||(6.2)|
|Prcv(f) = P8990B(f)/NGainrf#1(f) [W]||(6.3)|
where P8990B is the power measured (in watts) on the Keysight 8990B instrument, and NGainrf#1(f) is the net RF received power path gain, including all RF cables and the solid-state amplifier gains and losses of NGainrf#1(f). This term consists of the following:
|NGainrf#1(f) = Lrf1(f) · Lrf2(f) · Lrf3(f) · L3dB(f) · LWR28(f) ·Ga(f)||(6.4)|
Lrf1(f) = Loss of 1st semi-rigid RF cable,
Lrf2(f) = Loss of 2nd flexible RF cable,
Lrf3(f) = Loss of 3rd flexible RF cable (DAY only),
L3dB(f) = Loss of 3 db coupler (DAY only),
LWR28(f) = Loss of WR28 waveguide/coax adaptor, and
Ga(f) = Ka band solid state amplifier Gain.
The incident power density Pd, is found from the classical equation
where the effective aperture Ae(f) is defined as
and the dimensionless received probe antenna gain is Gr(f), and the wavelength (λ= c/f), where f is the frequency in hertz, and c is the speed of light in meters per second. Combining Equations 6.3 to 6.6 and solving for Pd, the incident field strength is calculated through the following relationship:
P8990B(f) is the measured power at the Keysight 8990B versus frequency,
NGainrf#1(f) is the measured net path gain through the Keysight 8990B RF cable path,
Gr(f) is the near field antenna gain of NARDA V637 receive horn, and
λ is the wavelength of the swept linear FM transmit frequency.
Given the dimensions of the NARDA V637 horn antenna (2.69 cm × 2.08 cm), the far-field distance or range (R) where the “far-field” gain of the receive horn can be used is calculated by
Inserting the values for the NARDA V637 horn yields the far-field range curve versus wavelength, shown in Figure 6.13.
Based on Equations 6.8 and 6.10, as long as the NARDA receive probe antenna is greater than 11.5 cm from the ProVision transmit array, Equation 6.7 is valid to calculate incident power density.
Finally, the incident E-Field in free space is simply from Equation 6.7:
From a calibration standpoint, Equation 6.11 is only in volts per meter if Pd is in watts per square meter. Every single term, except the 4p constant in Equation 6.11, is dependent on frequency (f), so to execute this measurement with high fidelity, every gain and loss must be known at every measured frequency between 24 to 30 GHz. A cross-sectional drawing of the NARDA V637 antenna used to measure the received power is shown in Figure 6.14.
Because calibration was extremely important, the committee decided to fully validate the manufacturer’s boresight and off-boresight antenna gain. The NARDA manufacturer’s far-field receive horn gain, Gr(f), for the NARDA V637 is shown in Figure 6.15. Although the curve in Figure 6.15 is linear, experience with these horns indicates that there are low-level ripples in the amplitude gain versus frequency. Therefore, two different NARDA V637 antennas were measured in the laboratory, and an independent computational electromagnetic (CEM) solution was calculated for the same antenna, comparing it against the manufacturer’s published estimate for boresight gain. The manufacturer’s values (based on a simple electromagnetic gain model) are lower (and more linear) than the actual NARDA measured and calculated values shown in Figure 6.16. The committee also compared the BRC measured and predicted data to published data from the National Institute of Standards and Technology and the U.K.’s National Physical Laboratory.2
2 National Physical Laboratory, 2003, Measurement Techniques and Results of an Inter-comparison of Horn Antenna Gain in IEC-R-320 at Frequencies of 26.5, 33, and 40 GHz, NPL Report CETM 46, Centre for Electromagnetic and Time Metrology, Teddington, Middlesex, U.K., September.
The small variations between the CEM and measured antenna gains represent almost insignificant additional uncertainty in an overall error analysis. Nonetheless, the receive gain curve is critical to calibrate the power density data in the field.
The net RF gain of the principle RF paths from the antenna to the measurement ports should also be calibrated. As stated earlier, the RF path for DAY airport was slightly different from those for the other three airports, so this value must be established for each measurement configuration.
The combined losses of Equation 6.4 were measured on a calibrated network analyzer and carried throughout the analysis to produce the end-state incident field calculations. Prior to executing the first ProVision scanner measurement in DAY airport, the overall RF loop gain/losses were measured for the primary RF pathways outlined above. The DAY airport net gain versus frequency for the primary signal path for the Keysight 8990B is obtained from Equation 6.4, only expressed here in decibel (logarithm) format.
|NGainrf#1(f) = –Lrf1(f) –Lrf2(f) –Lrf3(f) –L3dB(f) + –LWR28(f) +Ga||(6.12)|
For the CVG, MEM, and TYS airports, the 3 dB coupler (Lpd3) and a cable (Lrf2) were eliminated, which effectively set these terms to zero. The NGainrf#1(f) for the CVG, MEM, and TYS airports is, therefore
|NGainrf#1(f) = Lrf1 –Lrf3 –LWR28 +Ga [dB]||(6.13)|
A plot of NGainrf#1(f) for DAY airport and CVG, MEM, and TYS airports (Equations 6.12 and 6.13) is shown in Figure 6.17. By removing a cable and the 3 dB coupler from the DAY airport RF set-up, the effective system gain increased approximately 6 dB across the entire frequency range.
The data shown in Figures 6.16 and 6.17 were used to convert the raw measured power levels, P8990B(f), measured at the field probe from the Keysight 8990B power analyzer to calibrated field power densities, Pd(f). The final data products report incident power densities at the scan area having used this calibration method.
For each of the RF airport measurements, the field probe data at each test point consisted of a series of linear FM pulses that was maximized when the field probe antenna was electrically aligned with the ProVision scanner element directly in line with the field probe. An actual raw time-domain linear FM pulse train measured in MEM is shown in Figure 6.18. Because this display of raw data is in decibels, and the dynamic range is −70 to −10 dB (6 orders of magnitude), it tends to slightly distort the “off period” of the linear chirp waveform. The fact that the chirp is actu-
ally not very linear on a decibel scale does not impact how the ProVision analyzes passengers, because these slight variations are removed in post data processing. The committee is concerned, however, only with the actual levels of the emissions at these frequencies.
ACQUISITION, EXTRACTION, AND CALIBRATION OF RAW DATA FROM THE KEYSIGHT 8990B TIME-DOMAIN POWER METER
Figure 6.19 shows the overall schematic (though not to scale) of the interior scan plane geometry of the committee’s test set-up.
The field probe device is placed inside the ProVision ATD scanner chamber, and the unit is positioned with the front antenna extension installed, which places the NARDA receive horn 28.7 cm from the transmit array.3 While the L3 ProVision system scans from top to bottom, the field probe is positioned from bottom to top. Because it takes ~3 µs for the ProVision unit to complete a single line scan, the time-domain instrument in DAY airport was adjusted to display 3.5 µs of time response. For each horn position, the ProVision scanner was strobed and data acquired. The receive antenna was then raised approximately 3 cm, and the process was repeated between 44 to 46 times until the probe antenna was at its highest height. For Scan Plane 2, the extension was removed from the probe assembly, and
the receive probe was then positioned 59.5 cm from the ProVision front transmit array. With this method, the entire field probe apparatus, which weighed over 150 lb and was not easily moved, did not have to be repositioned. However, for Scan Plane 3, the entire field probe was picked up and rotated 180 degrees, and the rear array was characterized at 29 cm from the rear ProVision array.
Figure 6.20 shows 13 data files (out of 45) spatially plotted relative to the Scan Plane 3 geometry. Each of the colors represents 10,000 data points, so the resolution of the printed data in Figure 6.20 is much higher than can be depicted in this figure. As the probe antenna is raised from floor to ceiling sequentially, the data file achieves a maximum value when the NARDA field probe antenna aligns with the ProVision transmit element closest to the probe.
For Figure 6.21, the data in Figure 6.20 were rotated by 90 degrees to more readily annotate the time and position information onto the plot for discussion. The time axis spans ~3 µs, which corresponds closely with the 3 µs total time it takes the ProVision system to scan from the top of the array to the bottom with the ~200 transmit antennas. The light blue values show the probe position from the floor of the ProVision. There are 13 probe positions shown in this compressed plot. Figure 6.22 results from zooming in by a factor of 100 on the “center” or data file “226,” which is 88 inches (225 cm) from the floor of the unit, and shows 1,000 data points that are close to the local maximum of the raw field strength data. Figure 6.23 shows the results of zooming in another 10 times.
Now it is easy to see the individual linear chirp waveforms for this case. However, the data are still not calibrated and Figure 6.23 represents P8990B(f). Before it can be calibrated in the frequency domain, the time-domain display of Figure 6.23 needs to be correlated with the actual frequency emissions of the ProVision. Knowing that the chirp is linear (in time) from the min to max frequency over the ~6 µs “on” period, the translation is fairly straightforward. This is shown in Figure 6.24 for a single waveform from the 226 case of Figure 6.23.4 An algorithm was created to search any vertical probe position for the five highest linear FM chirps in the vicinity of the boresighted transmit element. The beginning of the sweep was aligned to a reference point, and the five highest waveforms were used to create an “average” waveform, including the ~6 µs “on” period and the ~2 µs “off” period. A sample of MEM airport data, shown again in Figure 6.25, is an example of this five-waveform alignment. The time axis is converted to the frequency axis by the following linear conversion equation, assuming alignment of the relative time waveform at the beginning of the 24 GHz sweep.
4 This is illustrative only; the actual technique used to correlate time and frequency involved much more signal processing.
For instance, at t = 3 µs, the equivalent frequency is ~27 GHz. At ~6 µs, the frequency is ~30 GHz. In order to factor out frequency dependence (or dispersion) of the Keysight 8990B measurement RF network, the “raw” data acquired by the 8990B must be “calibrated” using Equation 6.7.
As discussed above, it was straightforward to convert the time domain power meter measurements to an equivalent frequency domain power meter measurement because the ProVision linear chirp waveform is very precise and repeatable by design. Going back to Equation 6.7, now restated as Equation 6.15,
The numerator is obtained on a frequency by frequency basis by averaging the five-waveform set to a single average waveform (Figure 6.25), then numerically aligning the time data to frequency using Equation 6.14. This data is then adjusted point by point by the average receive Gain [Gr(f)] from Figure 6.16 and adjusted by the NGainrf#1(f) point by point from Figure 6.17, using the correct curve for either the DAY airport or the CVG, MEM, and TYS airport data set. The end result of this calibration data is an individual Pd curve similar to Figure 6.26.
After examining a large amount of data from the DAY airport tests, the committee decided that executing measurements approximately every 3 cm (instead of every 0.5 cm) in height was sufficient to characterize the ProVision scanner chamber from bottom to top. Since a typical vertical scan employed 44 to 46 individual probe positions, for three scan planes and four airports, presenting the data in the format of Figure 6.26 would require 552 individual plots to display all the data from Scan Planes 1, 2, and 3 internal to the ProVision; therefore, each airport will show for each of the three scan planes all the height probe data on a single Pd plot, similar to Figure 6.27 in the next section.
As was discussed previously, in order to speed up the intensive data acquisition time demanded by the system, the ProVision scanner was operated with the scanner antenna mast fixed in its mid-scan position and repetitively swept in frequency. This was considered a service/engineering mode and allowed the efficient collection of tens of thousands of emission waveforms per hour on each machine. If the
normal operator “sweep mode” had been used, it would have been impossible to collect more than two scans a minute, or roughly 120 sweeps an hour. Having a stationary scanner mast at its mid-scan position did not compromise the technical measurement in any way, and in the Scan Plane 4 measurements, data that were identical in character were acquired in a transient and fixed mode. This confirmed that emission waveforms were identical in both the “service/engineering” and “normal sweep” modes.
Figure 6.27 is the DAY airport Scan Plane 1 multi-plot, with the average of all the waveforms for this scan plane indicated in heavy black. There are, in each of these plots, two additional lines plotted that were calculated by CEM levels based on the L3 antenna geometry and transmitter information provided to the BRC/ARA team. The curves represent the absolute maximum power (dash line) that the ProVision unit could theoretically radiate if all the internal connections experienced absolute minimum ohmic losses (not something that is physically possible, but a limit to use as a reference). This is labeled “Max Spec Power.” On the other hand, L3 tests the internal components, based on a minimum specification for radiated power during production testing. Using L3 estimates of minimum power, the committee calculated the “dash-dot-dot” line that should represent the lowest level on which a normal machine would operate. It is highly unlikely that L3 tests its machine emanations for minimum power at every frequency, so even if some points are below the minimum, that only implies that a machine is operating at slightly lower power than normal. The color sequence goes from the blue end for scans near the bottom of the ProVision scanner to the hotter reds for scans near the top of the chamber.
The spread of the data increases above 29 GHz. This occurs because the system signal-to-noise figure varies more in this region, inducing more calibration offset. The average value of the measured power density fell right between the expected radiated power limits of the DAY airport unit tested with an approximate upper limit of −50 dBm/cm2 (or 0.0001 W/m2), which is 100,000 times below the standard exposure limit if the signal had been continuous and not pulsed, which is the real case. The limits, if there were no losses, marked in the graph are the lines at approximately −43 and −54 dBm/cm2, representing 0.0005 and 0.00004 W/m2, respectively, which is equivalent to 20,000 and 250,000 times below the standard exposure limit, respectively, representing the unobtainable extreme limits.
Scan Plane 2 faces the front ProVision array but is placed 30.48 cm (12 in.) farther back than Scan Plane 1. One expects the measured power density in Scan Plane 2 to roughly equate to the additional range loss compared to Scan Plane 1. Because Scan Plane 1 was 28.7 cm from the ProVision front array, and Scan Plane
2 was 59.5 cm away, the Scan Plane 2 power density data should decay approximately as follows:
|20 · log (28.7/59.5) = –6.3 db||(1)|
Comparing Figure 6.27 (Scan Plane 1) and Figure 7.28 (Scan Plane 2), it looks to follow this trend exactly as expected. The lower signal levels for Scan Plane 2 reduces the signal-to-noise further and, therefore, the uncertainty (or variation in the power density) for Scan Plane 2 should rise slightly relative to Scan Plane 1. The minimum specific power and maximum specific power lines are also adjusted downward to reflect the distance from the transmitter. Now, the upper value of the measured power density is seldom over −55 dBm/cm2 (or 0.00003 W/m2), which is 320,000 times below the standard exposure limit if the signal had been continuous. The limits in Figure 6.28 are the lines at approximately −48 and −60 dBm/cm2, representing 0.0002 and 0.00001 W/m2, and are 60,000 and 1,000,000 times below the standard exposure limit, respectively.
The last data set from DAY airport is the Scan Plane 3 data looking at the rear array of the ProVision system (see Figure 6.29). Because the apparatus weighs more than 150 lb, it was only possible to align the system within a centimeter in range. However, the calibration process took care of these small range differences, and the theoretical machine minimum and maximum values were adjusted to reflect the actual distances between the probe and array in the field measurements at each airport. The theoretical values are now for the rear facing array, which should be very symmetric with the front. As L3 confirmed, no two arrays are exactly alike. Therefore, subtle differences are expected between Scan Plane 1 and 3 for every airport case and among the different units that were measured. The average value of the measured power density was then approximately −51 dBm/cm2 (or 0.00008 W/m2), which is 125,000 times below the standard exposure limit if the signal had been continuous. The limits in the graph are the lines at approximately −43 and −54 dBm/cm2, representing 0.0005 and 0.00004 W/m2, and are 20,000 and 250,000 times below the standard exposure limit, respectively.
The floor of the ProVision portal has two footprint decals, and the operator instructs subjects to stand with their feet on the decals and their hands raised over
their head.5 This is intended to position each subject in the center of the portal. To appreciate the distance from the subject to the masts, consider the 95th percentile 40-year-old American male who has a bust depth of 28.2 cm and a foot length of 27.3 cm.6 A positioning error of 20 cm would require a subject to have significant parts of both feet outside the footprint decals, a case that would be clearly out of compliance with the operator’s instructions. This magnitude of positioning error would mean that the subject might be as close to the mast as 31 cm on one side, and as far away as 71 cm on the other side, which compares fairly well with the Scan Planes 1, 2, and 3 used in this report. Hence even an “out of position” person will receive a continuous power density during a scan that is no more than what is in Scan Plane 1—that is, 100,000 times below the applicable standard exposure limit—and most likely will receive much less standing in the center of the portal.
All of the DAY airport results reported above were produced with the rear reference antenna as described earlier. In subsequent airport measurements in CVG, MEM, and TYS airports, adjustments were made with the front reference antennas, and time base window adjustments—to allow a denser number of points in the curves above that—provide a slightly higher sampling of the power density in the frequency domain. Data sets for the other airports for each scan plane can be found in Appendix D and are similar to the DAY airport results.
While it is important to compare the actual measurement cases for individual airports, it is also meaningful to compare the four tested airport machines to one another. For this purpose, the average power density was extracted from each airport for each scan plane (normally the solid black curve for each case), and then these averages were plotted for comparison purposes (shown in Figures 6.30 to 6.32). It should be pointed out that there were slight variations (on the order of a centimeter or less) in comparable ranges between the airport probe distances, but it does not add too much variation (≈0.3 dB/cm). Looking at this data, it is clear the measurement results were extremely comparable airport to airport, despite continuing improvements executed in the RF signal paths and data processing as the measurement program progressed.
5 ProVision® 2 Operator Manual, CDRL D001, March 25, 2015, p. 21.
The measurements discussed above (e.g., see Figures 6.27 to 6.32 and also all measurements shown in Appendix D) provide a useful guide for assessing the risk associated with exposure to millimeter waves in the L3 millimeter wave AIT system. Based on these measurements, the committee has concluded that the peak intensity (the highest value observed during a scan) is in the vicinity of about −50 dBm/cm2 (or less, in most cases). For the purposes of estimation, the committee takes this value as a “worst-case” exposure scenario. Converting to more familiar units, −50 dBm/cm2 corresponds to 0.0001 W/m2. This value is 100,000 times smaller than the exposure limit of 10 W/m2 recommended by the Institute of Electrical and Electronics Engineers.7
However, one should also recognize that the L3 ProVision system does not continuously expose the passenger to radiation at this level. As can be seen, in Figure 6.18, for example, the system produces pulses of radiation and is, therefore, only radiating for a fraction of the time. In order to account for this, one must
7 IEEE, “IEEE Standard for Safety Levels with Respect to Human Exposure to Radio Frequency Electromagnetic Fields, 3 kHz to 300 GHz,” IEEE Std C95.1™-2005.
understand the temporal structure of the emitted radiation. The timing of the system is as follows:
- Each mast of the device contains ~200 transmit horns and ~200 receive horns. The horns are vertically offset, and each element (but one) is used twice in pairs. This results in a total of ~400 transmit pulses for a single line scan.
- The period for each transmitted pulse is ~8 µs, during which the transmitter is active for about 6 µs, and then off for the remaining 2 µs. In Figure 7.18, five consecutive pulses are visible. During each pulse, the frequency of the emitted radiation sweeps from min to max frequency; however, in this simplified analysis, we are ignoring the frequency-dependence of the exposure limit.
- These ~8 µs pulses are repeated until the line scan is complete, which requires (8 µs/pulse)(400 pulses) = ~3 ms.
- Then, after a dynamically set time (to allow the mast to swing to its next position with different speed as to allow for gentle acceleration), the process repeats. A full scan involves ~200 such line scans, and requires about 1.3 s to complete.
- We can compute the total amount of time during which any of the horns is emitting radiation, by simply multiplying the duration of a single horn’s emission pulse (~6 µs) by the number of times all the horns are fired (~400 per line scan times ~200 line scans). The result is about 0.5 s.
Therefore, the “duty cycle” (fraction of time when millimeter wave emission is present in the chamber) is about (0.5 s)/(the full scan time) = 37 percent.
To compute the exposure of a passenger during a single scan, one must account for this fractional (less than 100 percent) duty cycle. A worst-case scenario would be an exposure of −50 dBm/cm2 for the entire time when the system is energized (obviously this is an over-estimate, because the emission does not reach this maximum value at all times during each pulse, as is clear from the temporal structure of the pulses seen in Figure 6.18). However, proceeding with this assumption, an intensity of 0.0001 W/m2 is computed, with a duty cycle of 37 percent. This gives an average exposure of 3.7 × 10−5 W/m2, which is about 270,000 times less than the 10 W/m2 recommended limit.
The committee notes that making a direct comparison to the IEEE exposure recommendation for millimeter waves is somewhat problematic, because this recommended maximum exposure value (10 W/m2) is defined with respect to a certain averaging time. For example, one version of the recommended MPE is an exposure of 10 W/m2 for a duration of 6 min. It is challenging to know how to compare a pulsed source (with an operation time of only 1.3 s and a duty cycle of
37 percent during that operation time) to a standard that specifies a much longer averaging time. Moreover, other versions of the recommendation specify even longer averaging durations, further complicating matters. Regardless of how one performs the averaging, it is nevertheless clear that any reasonable averaging time (i.e., longer than 1.3 s) will only reduce the average exposure relative to the value (3.7 × 10−5 W/m2) computed above. Thus, the safety margin of 270,000 is clearly a worst-case (corresponding to an MPE averaging time equal to the duration of a single scan, 1.3 s). Any longer averaging time would only increase this safety margin.
Another way to look at these numbers is to compute the total energy exposure, rather than the radiation intensity. As above, the committee estimated an exposure of 0.0001 W/m2 for a duration of 0.48 s (which, as noted above, is an over-estimate because of the temporal structure of the radiated pulses). This corresponds to an exposure of 48 µJ/m2 per mast during a single scan. Multiplying by 2 (since there are two masts), a total exposure of about 100 µJ/m2 is obtained. Because a human presents an area of about 1 m2 (very roughly), the committee predicts that a passenger would be exposed to about this much energy during a scan, spread out over the entire body. The amount of energy required to raise the temperature of a single drop of water by 1/1000th of a degree Celsius is about 100 µJ. As another interesting point of comparison, at sea level on a clear day, the Sun deposits more than 1,000 J of energy per square meter on Earth’s surface (and on the people standing outdoors) every second. Another way to view 0.0001 W/m2 is the fact that it is about 250 times less than the peak intensity of the light from a full moon as seen from Earth.
While at DAY airport, the BRC/ARA team attempted to execute the Scan Plane 4 and 5 measurements using the geometry set up in Figure 6.11. With the scanner mast in the middle position, the transmit and receive antenna patterns were so far into their respective sidelobes (i.e., wide angles) that the direct RF data measurements at DAY airport were below the noise detection limit of the equipment (See Figure 6.33). The direct path to the Scan Plane 4 or 5 position would come in the sidelobe of the probe antenna and transmit L3 elements. When the noise limit of the Keysight P8990B is plotted versus the predicted levels that were based on the angle and distance of the measurement, it is clear why there was no signal response in this configuration.
At CVG airport, the BRC/ARA team attempted to measure the power density in the ProVision exit door adjacent to the TSA operators position (Scan Plane 4). As discussed above, this was unsuccessful at the DAY airport because the fixed center position of the ProVision scanner did not radiate sufficient energy into Scan Plane
4 to be detected by the equipment. At CVG, the ProVision scanner was operated in the normal scanning mode, rather than the fixed service/engineering mode, to attempt to capture the transient response at Scan Plane 4 with the scanner bar in motion. The split second after a right-to-left sweep is initiated (or when a left to right sweep ends), the scanner is in the position of Figure 6.34. To capture the reference pulse for triggering, the reference antenna was placed on a tripod and faced the opposite rear array in order to capture the transient response. The data acquired are excellent and are displayed in Figure 6.35. Essentially, the Figure 6.35 plot shows is the level of power at the exit door assuming the scanner mast was fixed in the position shown in Figure 6.34. The actual level of exposure on a single sweep if the TSA operator stood right in the exit door is at least 10 dB weaker than the energy that a passenger is exposed to. The exit door is only illuminated for roughly 30 percent of the mechanical sweep time—15 percent from the front array illumination and 15 percent from the back array illumination. If one factors in the 30 percent power duty cycle losses of the mechanical sweep, the expected exit door power levels of Figure 6.35 should be reduced on average by an additional 10*log(0.15) = −8.2 dB to account for the sweep duty cycle.
The BRC/ARA team also attempted to measure the cross-polarization level of the ProVision system in Scan Plane 3. As one can see in Figure 6.36, there was not sufficient sensitivity to detect the lowest level of expected cross polarization. With the probe antenna rotated 90 degrees, there is an expected decrease in the signal-to-noise of −5 dB on boresight for cross polarization. Therefore, for all practical purposes, the additional levels of cross-polarized RF in the chamber are at least 22 decibels, or more than two orders of magnitude, lower than the co-polarized
signal. As this was expected, further cross-polarization measurements were not attempted. Hence, with a noise limit below −70 dBm/cm2, and no signal seen, the 1.0 × 10−6 W/m2 represents an intensity at least 10 million times below the applicable standard, or if the pulsed nature of the signal is taken into consideration, 27 million times below the standard.
With an understanding of the RF subsystem used in the L3 ProVision units, it is not expected that any harmonic or subharmonic frequencies would be detectable, particularly given the low emissions at the in-band frequencies. The BRC/ARA team used a wideband spectrum analyzer at the DAY airport to test for subharmonic emissions. A standard 2 to 18 GHz dual-polarized horn antenna was positioned in close vicinity of the ProVision scanner in an attempt to detect frequency harmonics of the original millimeter wave (24 to 30 GHz) sweep band. Radiated emissions were sampled in the 25 percent harmonic band (6.0625 to 7.5 GHz) while the provision scanner was operating in the mid-scan position and in repetitive sweep mode. Radiated emissions were also sampled in the 50 percent harmonic band (12.125 to 15 GHz) of the normal millimeter wave band emis-
sion frequencies (24 to 30 GHz) while the provision scanner was operating in the mid-scan position and in repetitive sweep mode. Although the spectrum analyzer antenna was placed right up against the ProVision Lexan radome, no radiation was detected. During measurements of the in-band 24 to 30 GHz region when the ProVision scanner is not radiating within the millimeter wave bands—that is, in standby—no emissions were detected.
A discussion of the uncertainty analysis is provided in Appendix E. The analysis examined all of the measurement uncertainties associated with Equation 6.7, which was used to determine the measured power density in this study. While the uncertainty is a function of frequency, the committee points out that near the midrange of a frequency (f ≈ 27 GHz), the uncertainty in any particular measurement is ±1.2 dB (32 percent). This analysis is an estimated uncertainty of the measurement technique and would not account for any variations in the manufacturing tolerances of the ~200 transmitters found on each mast, which would be partially represented in data such as those in Figure 7.28. Most importantly, as shown in Figures 6.30 to 6.32, the comparisons show that the average power density (average of all waveforms) at each of the airports for each scan plane are very similar, show no large deviations, and are consistent with the estimated uncertainty.
Finding 6.1: The committee-led measurements at airports indicate that even an “out of position” person will receive an average pulsed power density during a scan that is 270,000 times below the applicable standard exposure limit of 10 W/m2 and most likely will receive even less standing correctly in the center of the portal.
Finding 6.2: The committee-led measurements at airports indicate that even at the entry position of the portal, the power density is several million times below the acceptable limit, even for a continuous signal.
Finding 6.3: The committee finds that during normal operation of the ProVision system, there is no risk to a person being screened to receive the applicable standard exposure limit of 10 W/m2; instead, the exposure is hundreds of thousands times less.