Safety standards for setting acceptable levels of nonionizing radiation have been established for operating frequencies to 3 GHz.^1-3 While such standards cover a wide range of wireless devices, a growing number of consumer wireless devices operate in the range from 5 to 6 GHz, even in the absence of international testing standards. The key metric for assessing wireless devices used in proximity to the body is the specific absorption rate (SAR), which can be expressed in units of W/kg and for which limit standards for public exposure exist from 300 MHz to 3 GHz.^1-3

As frequency increases, absorption in the body is characterized by a more intense, but shallower influence on body tissues related to the wavelength of the transmissions in such media. At 300 MHz, penetration depths are typically 50 mm. At 6 GHz, it is around 5 mm. Most recent attention on SAR measurements has been in the cellular range from 835 to 1900 MHz. SAR probes, with tip diameters from 5 to 8 mm, are small relative to the exposure volumes at these frequencies, and suitable for examining field variations within the layers close to the body surface. But at 5.8 GHz, which is used for IEEE 802.11a wireless local-area networks (WLANs), the penetration depth is only a few mm and current-generation SAR probes are no longer smaller relative to the exposure volumes. Although errors that arise from using "large" probes in field gradients have been assessed and a correction scheme has been advanced,⁴ the main avenue for improvement is by producing SAR probes of reduced dimensions for measurements at 5 to 6 GHz. Other procedural problems related to the use of complex equipment for testing at frequencies above 3 GHz arise in the calibration of SAR probes and in the methods used for validating measurement set-ups to ensure that answers obtained in testing are correct.

At frequencies below 3 GHz, the recommended procedures for validating SAR measurement systems involve the use of balanced dipoles positioned at fixed distances below flat (box-shaped) phantoms. In the 5-to-6-GHz band, however, it is reported that the use of suitable balanced dipoles becomes difficult due to the reduced manufacturing dimensions and the increased accuracy of positioning that is required.⁵ Solutions have been proposed by using open-ended waveguides⁵ spaced away from the phantoms as alternative irradiation sources, but we have found it difficult to obtain measurements meeting the proposed reference values.⁶

This study proposes modifications to the waveguide validation technique, which employs the same waveguide recommended for SAR probe calibrations.^2,3 The waveguide are placed in direct contact with box phantoms using a dielectric matching window, which minimizes losses due to leakage and reflection and provides a reproducible geometry with minimal risk of misalignment.

A validation procedure has been defined utilising a flat phantom and a source based on the waveguide launcher with a dielectric matching window. This configuration has been modelled using the Falcon FDTD package at York University to establish reference values for key field parameters, which include the maximum 1-g or 10-g volume-averaged SAR values that should be obtained by measurement for a system validation.

The dimensions of the phantom used for the validation procedure are pre-defined by the requirements of the relevant standards for SAR assessment.^1-3 At test frequencies above 3 GHz, the phantoms do not need to be sized on account of the frequency—rather on account of the dimensions of the devices being tested (or at least their active parts). The configuration used for this study is shown in Fig. 1. The rectangular box phantom is more than adequately sized for the test frequencies involved. Potentially, different matching windows could be used for different frequencies in the band as shown in Fig. 1, but a single matching window of 5.2-mm thickness was selected for the tests reported here.

Version 1.6 of the York University Falcon FDTD package was used for this study. This package has been validated against COST 244 benchmarks where the results fall in the midrange. The model has previously been applied to several studies closely related to the present application.^7-9 The waveguide and box model was set up according to Fig. 1.

A relative permittivity of 2.56 was used for the 2-mm-thick phantom base. A uniform grid with a voxel size of 0.5 mm was used (less than one-tenth of the wavelength in the lossy liquid). The computations extended for a depth of 35 mm into the liquid and 85 × 65 mm laterally. The conducting waveguide walls were nominally considered to be lossy with realistic conductivity values used. A "monopole" excitation source was used with the length varied to give a match with 50-Ω source impedance. The time step was determined by voxel size and stability condition. A time duration of 8.34 ns was used with a time step of 0.834 ps. The computations typically took 2 h on a personal computer running the Windows 2000 operating system.

Since waveguide dimensions scale inversely with frequency, waveguide dimensions, which are impractically large at 300 MHz, are on the limit of being too small compared with probe tip dimensions above 5 GHz. Two different waveguide types have been referenced in previous work on SAR probe calibration and validation above 5 GHz. The ridged waveguide WR187 in ref. 5 has the advantage of slightly larger linear dimensions. On the downside, it is not one recommended for SAR probe calibration² and matching window designs for the 5-to-6-GHz band have not been published. Dimensions for each are in Table 1.

The waveguide selected for this study is the slightly smaller WG13 type, listed in the recommendations for higher-frequency probe calibrations¹³ and for which matching window parameters are suggested. The objective here is to reduce the component count required for SAR evaluation by using the same items for both probe calibration and system validation.