Microwave Absorption in Humans

djjx.jpg (15256 bytes)
By David J. Jefferies G6GPR

juanx.jpg (8737 bytes)


Juan Fanals

School of Electronic Engineering
IT and Maths, University of Surrey, Guildford GU2 7XH, Surrey, England, UK

micro_o.gif (1415 bytes)ne of the authors (DJJ) was approached by a company in Scotland, UK, for advice on the following problem. The idea had been floated that a miniature camera could be made in the form of a pill, swallowed by a patient, and could transmit pictures of the inside of the gut to antennas located on the patient's skin. The technology exists now to make this feasible from an equipment size point of view, but it was not clear what frequency should be used for the link. The trade-offs here are that as the frequency increases, the attenuation of the body rises steeply. However, the transmit antenna has to be carried on the pill and it will be very small in overall linear dimensions compared to a wavelength; at higher frequencies the wavelength is less and so we may expect more power to be coupled from the transmitter to the radio wave in the body.

An initial theoretical estimate by DJJ indicated that the process would be possible towards the lower end of the 1GHz to 10GHz frequency range. The free space wavelengths (in air) range from 30 cm at 1GHz to 3 cm at 10GHz. However, the microwave dielectric constant of water is about 80 so we might expect the wavelength inside the human to range from about 3.4 cm at 1GHz to 3.4 mm at 10GHz. Thus a loop antenna having a perimeter of one wavelength would vary in diameter from about 1 cm to about 1 mm over the frequency range. Now this is a very rough estimate, because the antenna will be printed onto a dielectric substrate and the coupling calculations and simulations are very difficult to do, for the environment found in the gut.

It therefore seemed advisable to support the estimates of body attenuation with a measurement. The other author, JF, undertook these as his final year undergraduate student project in the Electronic and Electrical Engineering department at the University of Surrey, and it is these results that we report here.

Meanwhile, the original application has been successfully implemented and the project has been reported in the press.

We thought that there were a number of issues in these measurements which would be of interest to the amateur radio antenna community. These include

Constructing the Antennas
The important property of the antennas for this application is that they should spread the radiation over a significant proportion of the human torso, and not radiate in unwanted directions when placed against the body.

This was achieved using a very simple arrangement of aluminium foil stuck to cardboard, in the form of pyramidal horns. Here are a pair of the larger antennas.

large-antenna-pair.jpg (17275 bytes)

It was discovered that the smaller versions of these antennas had rather lower free space insertion loss across the small distances in the laboratory. Coupling to the fields in the antenna mouths was made by a 7cm length of copper wire connected to a coaxial BNC socket. This was chosen to be about a quarter wavelength at 1 GHz. The rear of the antennas was made from square cross section waveguide about 7cm on a side, and the spacing from the rear wall was also 7 cm. The cutoff frequency of this guide section will be about 2GHz which was rather larger than the maximum frequency used for these experiments (1.5 GHz max); the philosophy was to avoid a resonant waveguide cut-off transition within the band of frequencies used and to make a reasonable reflection of power in the forward direction. Three typical polar radiation plots for these antennas, in the lab, are shown below.


polar.gif (22106 bytes)

The front-to-back ratio is at least 17 dB; the measured rear lobe is probably due to reflections off objects in the rear-facing direction. No special efforts were made to optimise the match of the antennas, there being plenty of signal at the generator power level of +10 dBm.

Initial measurements were made of path loss against distance. In air, the antenna separations were such that for the most part the antennas were in each other's near field region. This remained so when they were placed in position in front of, and behind, the human torso as is shown below.

measurex.jpg (58831 bytes)

The loss vs distance measurements in air revealed that there was much multipath radiation, arriving at the receive antenna having scattered off objects in the laboratory, despite the directionality of the horns.

The width of the human bodies used was in the range 35-45 cm, and this was comparable with a wavelength in air at the centre frequency of the measurements. It was realised that there would be significant contribution to the received power from diffraction around the body, and possibly surface wave coupling. There were some microwave absorber screens available in the lab; these had been designed for use at 10GHz but attenuation measurements at normal incidence at 1GHz showed that they had transmission loss of about 10dB.

attenwallsx.gif (4983 bytes)

The screens were placed either side of the subject and measurements made of the effects on the path loss at various frequencies and power levels.

channel-wide-Juan.jpg (42414 bytes)

It was eventually found that the absorbers worked best when placed close either side of the subject. This was not unexpected; they suppressed most of the diffracted field contribution and helped to eliminate any surface wave.

Juan in the screens

Measured Results
A number of different subjects were measured. Individual attenuation figures varied by several dB; the subjects were instructed to exhale and hold their breath for three seconds or longer while the measurements were taken. A typical subject increased the loss on the link by about 40dB at 1GHz under these conditions. A representative set of raw data is displayed below. The effects of the absorber screens can be seen clearly. The insertion loss falls steeply below 1GHz as the wavelength becomes larger than the width of the subjects.

path loss vs frequency

Various measurements were taken to eliminate experimental uncertainties. Both the large and small antennas were used. The geometry of scattering objects in the lab was altered. The effect of varying the transmitter power was investigated, to see if the detector calibration was accurate.

Data Analysis
Despite the variability of the path loss from measurement to measurement, the results of this experiment, when averaged, are reasonably consistent. Various factors contribute to the path loss. We have discussed above the various ways that energy could get from the transmitter to the receiver without going through the human. The absorber screens were necessary, and it is felt that they were reasonably effective in stopping most of the leakage as, in the raw data plots above, the ripple in the insertion loss due to resonant effects is largely absent.

It is possible to estimate the wave impedance mismatch at the human-air boundary by the following argument. The human body, for microwave purposes, is mostly salty water. The dielectric constant is about 80 at these frequencies, and that puts a single surface reflection coefficient at about 0.8 and a single surface transmission loss at 4.4 dB. Thus, the reflection losses on the path are around 8.8dB.

If we discard the measurements below 800 MHz (for which the free space wavelength is 37.5 cm or larger - about the width of DJJ) and subtract the total interface loss from the remaining measurements, we obtain the following data table.

        800 MHz        -20 dB (+/- 2dB)
        900 MHz        -28 dB (+/- 2dB)
        1.0 GHz        -28 dB (+/- 2dB)
        1.1 GHz        -29 dB (+/- 2dB)
        1.2 GHz        -31 dB (+/- 2dB)
        1.3 GHz        -38 dB (+/- 2dB)
        1.4 GHz        -43 dB (+/- 2dB)
        1.5 GHz        -47 dB (+/- 2dB)

This data is fit best by an empirical formula that the loss (dB) in 30 cms average thickness of body is given by

loss(dB) = (28 +/-2)x[frequency(GHz)]^1.2

This formula is supported by another measurement of the loss through DJJ's hand at 10 GHz, which would put the loss for 30 cms thickness of body at 470 dB +/- 100 dB compared to the formula prediction which is 444dB +/- 32 dB.

Going back to the original application, the path loss at 1GHz from the middle of a human torso to the surface, including one reflection loss at the single surface, is no more than about 20dB. Therefore using a transmitter power of one milliwatt (-30dBW) and a noise floor estimate of -120 dBW for a bandwidth of 10MHz there would still be headroom of 50 dB for the transmitter antenna coupling loss, should we need a receive S/N of 20dB. The upper frequency at which this application would work is somewhere between 1GHz and 10GHz; a reasonable estimate from this experiment's data would put it at about 3.5 GHz.

However, the data from the experiment yields some thoughts of interest to the amateur. The attenuation in human torso at the mobile phone frequency of 1.8GHz is about 2dB/cm, and the skin depth (the distance at which the field strength has fallen to 1/e of its value at the surface) is about 4.6 cms. If we assume the attenuation in the head is comparable to that in the midriff, then most of the power from a mobile phone antenna is absorbed within a couple of cms of the head surface, assuming that it gets past the reflection loss at the interface with air.

Extrapolating our empirical formula to lower frequencies, the wavelength in a human at 146 MHz is about 23 cm allowing for the slowing effect of the dielectric constant of the human. That means that in the 2 metre band, parts of the human, if lossless, would stand a good chance of being a resonant structure. Our empirical formula, if trustworthy, would suggest that the loss at this frequency might be 0.1 dB/cm or 2.3 dB/wavelength. Thus the human may be considered to be a lowish-Q resonator close to the 2 metre band with a capacity for absorbing nearly all the coupled power.

By the time we have got to 10 metres, the human will support resonant currents and may re-radiate. Of course, these resonant effects may well be masked in reality as there is a reflection coefficient of 2/3 of the incident power from the human impedance mismatch, and it is for this reason that the signal strength in a band II broadcast receive scenario varies wildly as people walk around the room containing a receiver with a rod antenna.

We have presented the results of a simple and imprecise measurement of the attenuation of microwaves close to 1GHz through a variety of humans. From this measurement we have derived an empirical formula, which should not be extrapolated too far beyond the frequency limits of the original measurements without further investigation. We have considered the implications for frequencies used by mobile phone operators and by hams. These implications require further investigation, in our opinion.

We have also showed that it is not surprising that the gastro-pill transmitter can be made to work. -30-

David Jefferies PhD CEng CPhys, D.Jefferies email
Department of Electrical Engineering
University of Surrey
Guildford GU2 5XH
Click here for David Jefferies' Biography

Send mail to webmaster@antennex.com with questions or comments.
Copyright © 1988-2004 All rights reserved worldwide - antenneX©
Last modified: December 27, 2003