The measurements as described above were performed on the most recent optimized array and on several older nonoptimized control arrays, which lack these design improvements. The results are presented below.
Measured Characteristics of Microstrip Antenna patches
Measurements made with HP 8753C network analyzer with 85047A S parameter test set using 3ft coaxial cable and smapmmx adapter into 9mm bolus over bag of liquid muscle phantom.
Using TDR the reflections from at the connector all the way to the patch were measured.
Data From NonOptimized Antenna Arrays:
Table 1 Network analyzer measured data from two different nonoptimized arrays.
Data from Optimized Antenna Array
Table 2 Network analyzer measured data from the optimized array.
Comparing the overall averages (third line from the bottom) from the two tables, it is evident that the variation in all measured quantities for the optimized array is much less than for the nonoptimized arrays. While the averaged real part of the input impedance for the nonoptimized array happens to fall slightly closer to the ideal 50 ohms, its variance is seven times greater than in the optimized case and thus is clearly inferior. Although an ideal aperture would exhibit zero reactance at its resonant frequency, the averaged imaginary part of the input impedance is positive in both cases but the nonoptimized value is almost twice as large, with three times the variance. This significant reduction in variance indicates that the line length matching technique is a significant design improvement. The optimized array has a return loss that is 1.6 times less than the nonoptimized array on average, and the variance in RL as a percentage of the average value is only 10% of that for the nonoptimized case. The SWR of the tested antenna arrays is not dramatically different. The average SWR of the optimized array is 72% of the SWR for the nonoptimized array but the variance in SWR values for the nonoptimized arrays are three times that of the optimized array.
Representative Analyzer Results
The following representative analyzer measurements were performed on two antennas one with optimized feedlines and one with nonoptimized feedlines with as similar geometries as possible. Both antennas were 3cm apertures with 2.5mm gaps and have approximately the same feedline length.
Figure 1 TDR graph of an antenna with nonoptimized feedlines.
Figure 2 TDR graph of antenna with optimized feedlines.
Comparing the two previous graphs of figure () it is seen that the optimized antennas feedline is much closer to the ideal 50 ohms than the nonoptimized. The optimized feedline varies a maximum of 50mU away from ideal while the nonoptimized feedline stays at a nearly constant 200mU away indicating that the nonoptimized feedline has a impedance that is lower than the ideal 50ohms. The magnitude of the disturbance from the coaxial connector is 25% smaller for the optimized case. A measure of effective impedance matching is the variation of the TDR graph with distance. The variation in impedance from the 1^{st} Tjunction to the 2^{nd} Tjunction to the patch is only 50mU for the optimized case while the same interval in the nonoptimized case has a variation of 300mU indicating a better match in the optimized case. Also, the maximum distance away from ideal is 450mU in the optimized case where it is 650mU in the nonoptimized case, indicating an overall smaller departure from ideal for the optimized case.
Figure 3 Return Loss as a function of frequency for the nonoptimized antenna.
For this nonoptimized case it is seen that the RL is only –4.3dB at 915Mhz and the shape of the RL versus frequency graph does not appear to have any definite resonance over this frequency band. This is a very poor match with 40% of applied power reflected back to the source.
Figure 4 Return loss as a function of frequency for the optimized antenna.
For the optimized feedline case the RL is –18.4dB at 915MHz. We see that there appears to be a definite resonance around 915Mhz. There is a relatively sharp dip at 900MHz. This sharp dip that appears on all the optimized antennas tested in approximately the same place. This valley, so close to the intended matching frequency, is indication that the matching techniques are having a positive effect. These results indicate that there was a systematic error in the calculations producing a slight offset from the desired 915Mhz resonance. The most probable cause of this offset is a miscalculation of the effective wavelength in the PCB substrate. If the assumed effective wavelength was incorrect, the quarter wave transformers would not be exactly the correct length for destructive interference at 915Mhz. From the data above, it appears that the calculated ¼ wavelength in microstrip was slightly shorter by about 11% than actual. Future designs will correct this.
The following two graphs display the Standing Wave Ratio versus frequency for the optimized and nonoptimized antennas.
Figure 5 The SWR versus frequency for the nonoptimized antenna.
From the above graph it can be seen that the SWR for the nonoptimized antenna is 3.36 at 915Mhz and varies considerably over the frequency range.
Figure 6 The SWR versus frequency for the optimized antenna.
In contrast, the SWR for the optimized antenna is only 1.3 at 915Mhz and over the entire frequency range it varies from a low of 1.1 to a high of 2.1.
Another indicator of antenna performance is the overall output efficiency of the antennas is the powertovoltage ratio. Before scanning the antenna array the antennas were scanned individually to quantify uniformity of the heating patterns. The power output from each antenna is adjusted to the same level. This balancing provides information on how much power was required to achieve a particular voltage reading from the electric field probe. This input power to output voltage ratio is a good indicator of aperture efficiency. The following data was recorded with the electric field probe positioned over the center of each aperture, 1cm away from the surface of a 9mm bolus. The power balance information is from the same antenna arrays that were used in the network analyzer measurements.
Average Power Required to Generate 2 Volt Output 1cm Deep in Liquid Muscle Phantom
NonOptimized Array 15.1 ± 4.2 watts
Optimized Array 7.3 ± 0.37 watts
Comparing the two cases it is seen that the power required for the optimized array is less than half of what is required to obtain the same output with the nonoptimized array. In addition, the percent variance for the nonoptimized array is 5.5 times as large as the same quantity in the optimized array. This data is another indication that the overall matching and balancing techniques have been successful at producing a wellbalanced high efficiency antenna array.
The following four graphs show the SAR patterns from a single optimized antenna and a single nonoptimized antenna. The electric field was scanned 1cm away from the surface of a 9mm bolus with approximately the same output power from both antennas. The scale shows both voltage and relative SAR. The colored bands represent %SAR where purple is 100%, yellow is 75%, red is 50% and green is 25% SAR.
Figure 7 SAR contour and surface plots for the nonoptimized feedline antenna.
Figure 8 SAR contour and surface plots for the optimized feedline antenna.
In both cases the SAR pattern has essentially the same shape The 50% SAR contour is well outside the square patch antenna perimeter. The region with higher than 50% SAR is relatively broad and flat, with no sharp peaks or dips in power deposition. The SAR pattern follows the shape of the antenna and is roughly square in the region outside the antenna. Both of these patterns would heat well with no localized hotspots and good addition of fields between adjacent antennas on the same array. This result was expected. No change was made to the actual radiating antenna geometry, only in the feedline network which delivered power to the aperture. This data indicates that the feedline optimization did not have any significant effect on the SAR pattern, as desired.
The following SAR scan is included to show the shape of the additive field with nine antennas powered simultaneously. With the 50% SAR contour level extending approximately 7.5mm outside the perimeter of each square aperture, it is no surprise that the maximums are also located inbetween antennas for this aperture spacing of 1.5cm. The overall 50% SAR contour is also just larger than the physical array boundary. The area inside the 50% contour is also reasonably flat with no extreme peaks or dips. Thus this array can effectively heat an area, the area inside the 50% contour, of 196 square centimeters.
Conclusions >>  
