GPS Antenna Placement Analysis for Receivers
By J. Blake Bullock
The placement of the GPS antenna on or inside a motor vehicle can affect the performance of the GPS
receiver. This article investigates the performance implications of mounting GPS antennas in the passenger compartment of a vehicle, with reduced coverage due to decreased satellite visibility.
Global positioning system (GPS) receivers are used extensively in intelligent transportation systems (ITS) for positioning and navigation. Positions derived from GPS receivers are the key pieces of information in emergency location systems, theft recovery systems, vehicle tracking systems, and in-vehicle
navigation systems.
GPS receivers use L-band signals from orbiting satellites to compute the three-dimensional (3D) position and the velocity vector components of the vehicle platform. The signals are line-of-sight and are blocked by metal and other materials, so proper antenna placement on a vehicle is critical for optimum performance.
GPS antennas are relatively small, so there are many possible locations for mounting them on a vehicle. In some cases, multiple antennas may be required for added safety and
security. The trend for vehicle mounted GPS antennas is to mount them in discreet or hidden locations to prevent detection and/or theft. Ease of installation and servicing is also an important consideration when selecting a location for the installation of the GPS antenna.
This article investigates the performance implications of mounting GPS antennas in the passenger compartment of a vehicle. The decreased coverage due to reduced satellite visibility is evaluated; the benefit of a dual antenna system
is considered; and the effect of antenna placement on dilution of precision (DOP) measures and satellite availability is analyzed. The analysis is based on simulated satellite alert data and actual road-test data recorded in both forested and urban canyon areas.
The results of the analysis show that mounting the GPS antenna on or under the front dashboard, or on or under the rear package shelf has minimal impact on the overall availability and accuracy of GPS positions. For both locations, the angle of
the glass is the most significant factor in the performance. There is little or no benefit from having antennas mounted in both locations.
Performance degradation
The placement of the GPS antenna on or inside a motor vehicle affects the performance of the GPS receiver. The antenna must have a clear view of the sky with no metal obstructions in the way. Since the antenna cannot be placed under metal body panels, the only remaining discrete location is in the passenger compartment.
To gain
the best view of the sky from within the passenger compartment, the antenna must be placed as far forward under the windshield as possible or as far back under the rear light as possible. To remain hidden, the antenna can be mounted underneath the nonmetal skin of the dashboard or the package shelf. A solution being considered is to use antennas mounted in both of these locations and switch the RF input from one to the other if a fix is not available.
The purpose of this investigation is to determine
the performance degradation when using an antenna mounted in the passenger compartment in one or both of the positions described, as compared to an antenna mounted on the roof of the vehicle with a clear view of the sky. Several performance measures will be considered, including accuracy and available coverage.
The test was accomplished in two phases. First, satellite alert computations were used to predict the satellite visibility and DOP due to the geometry of the satellites in view. Secondly, a series
of drive tests were performed to record the coverage available for antennas in various locations.
The most important performance characteristics to consider for using GPS in an emergency response system are coverage availability and accuracy. Both the simulation and live tests are used to generate coverage and accuracy measures.
Metrics
GPS coverage is defined as the area in which a position fix is available. A minimum of three satellites are required for a 2D fix and a minimum of four
satellites are required for a 3D fix. The satellites must be unobstructed by buildings or metal in the vehicle. A 2D fix is a horizontal fix with the height held fixed at the last known value. With some GPS receivers, a third type of solution is available, known as a propagation (prop) fix. When the number of available satellites drops to below three, the receiver will use the last known position and velocity vector to propagate a solution over the next 5 sec. After 5 sec, no fix will be reported. The
propagation solution works well when the vehicle has a constant velocity over the 5-sec interval.
The positional accuracy of a GPS fix depends on both the accuracy of the satellite measurements and the geometry of the satellites used in the solution. Antenna placement has little effect on the accuracy of the measurements as long as multipath is avoided. To reduce multipath, the antenna should be mounted flush or recessed in a location away from the edges of the vehicle so that multipath cannot arrive at the
antenna from below.
The strength of the geometry of the solution is measured using DOP values. The horizontal DOP (HDOP) is used for a 2D fix, and the position DOP (PDOP) is used for a 3D fix. The unitless DOP factor is a multiplier that is applied to the measurement accuracy to determine the overall estimated solution accuracy. Therefore, the lower the DOP the better, with a DOP of one being perfect. The position of the GPS antenna affects the DOP values due to the impact on the number of satellites
visible and the region of the sky where the satellites can be seen.
Alert planning
Alert planning software uses a satellite almanac to predict the positions of GPS satellites over a 24-hour period. A GPS satellite almanac consists of Keplarian orbital elements that are used to characterize the path of the satellites as a function of time. With a date, time, and position on the Earth, the almanac can be used to predict which satellites will be in view, based on azimuths and elevation angles for
each satellite.
Various masking curtains were used to simulate the blockage of the roof when the antenna is mounted in the passenger compartment. The number of satellites in view given a certain set of masking curtains can be determined. Also, the DOP factors that result from the constellation in view can be computed. The effect of the vehicle's travel direction was also investigated.
Drive testing
A typical four-door sedan was used for the drive testing portion of the analysis. The
vehicle had windshield and back light (rear window) slopes that are about average (30ý to 35ý). The testing was done with three
8-channel GPS receivers recording data simultaneously. The factory default settings were used for the test with the exception of the 2D to no fix threshold, which was set to 20 instead of 12. The default settings include a 10ý mask angle, a 3D to 2D threshold of 6, and an all-in-view position computation. Standard active GPS antennas were used with 6m of RG-58 cable.
The three
antennas were placed on the roof, in the passenger compartment on the dashboard, and on the package shelf. The vehicle was driven in various urban settings for several hours throughout a 24-hour period in the city of Chicago, IL. The antenna on the roof was considered to be the best possible situation in terms of both fix availability and accuracy. The data recorded using the roof antenna was used as the baseline for comparing the other two antenna locations.
For the live satellite-test data, several
metrics are used to compare the performance. The total number of fixes available and the types of those fixes (3D, 2D, prop) are used to compute GPS coverage. The overall GPS coverage is the total number of fixes available as a percentage of the total possible. The maximum number of possible fixes is the number of seconds of the test duration.
In order to estimate the impact of the antenna placement on the accuracy of the positioning, the DOP factors can be used. For the 3D fixes, the PDOP values are
used, while for the 2D fixes, the HDOP values are used. For this analysis, the maximum, minimum, and average DOP values are determined.
Of particular concern with an emergency system is the longest duration where there was no fix available, which can be measured as a gap time. The number of gaps, the maximum gap time, and the average gap time are computed from the data.
In order to evaluate the benefit of using two GPS antennas simultaneously for the system, the number of fixes available when one
antenna does not have a fix is recorded. So, for each of the three antennas (when there is no fix available), the number of fixes available on each of the other two antennas is counted.
Alert simulation
Masking curtains were entered into the software to represent various glass slopes. Obviously, as the slope increases, the number of visible satellites decreases. With a 10ý mask angle, there are always five to eight visible satellites . With a simulated windshield slope of 35ý, the number of
visible satellites drops by an average of one (i.e., seven to six). Refer to
Figure 1
for a graphical comparison of the number of satellites in view with a roof mounted antenna (dark) and a windshield mounted antenna facing east (light).
Figure 2
shows the PDOP values over the same 24-hour period. The PDOP values increase on average by less than one when the antenna is placed in the windshield. However, in some cases, the PDOP increases by
a larger amount. In the cases where the PDOP exceeds six, the GPS receiver automatically holds the height fixed and computes a 2D fix. This information was computed using a 24-hour window in Chicago with fixes every 2 minutes.
The effect of the direction was also analyzed. When the windshield is facing south, the number of visible satellites is highest. When facing east, west, or north, there is a slight drop-off in the number of visible satellites. Similarly, the DOP values are best when facing south.
This is expected when in northern latitudes due to the GPS constellation. The plots in
Figure 2
are based on a windshield facing east, which would be worse than a windshield facing south.
The largest factor in the antenna placement performance trade-offs is when there are blockages from buildings and trees, which cannot be properly simulated. For this reason, the road tests were performed.
Forest though the trees
Table 1
contains the statistical data from the forest preserve test. The total time of the test was 1,260 sec
(21 minutes). As shown in the first block of data, only the front antenna had a gap, which lasted only 1 sec. The number of 2D fixes for the front and back antennas was higher than for the roof antenna by a few percentage points. A corresponding decrease in the number of 3D fixes for these two antennas was also observed.
The maximum, minimum, and average PDOP values for the forest test were nearly the
same across the board. The maximum HDOP was slightly higher for the front and back antennas as compared to the roof antenna.
These results show that using an antenna in the passenger compartment does not compromise coverage. Based on the DOP values, the accuracy degradation is also minimal.
Figures 3, 4, and 5
, showing the vehicle path as computed using the roof, front, and back antennas respectively.
Urban canyon test 1
The statistical results from the
first urban canyon test, which lasted 2,520 sec (42 minutes), are presented in
Tables 2
and
3
(and the plots are shown in
Figures 6, 7, and 8
. At first glance, it appears as though the front antenna was much worse than the roof antenna. Looking more closely at the breakdown of fix types reveals that there was a much larger number of propagation fixes and that the number of 2D and 3D fixes was much closer.
Another major factor for the difference in coverage is evident in the plots of the data. There are four long north-south runs in this test run. On the second north-south run from the left, there is a long gap in
Figure 7
, which is for the front antenna. At this point, there are tall buildings on both sides of the street and at the south end of the street where the road has a dead end. The vehicle was traveling south on this street. It is clear that the front antenna was
unable to obtain fixes for a long time due to the severe blockage. The roof and back antennas were still able to get a few fixes since they were better able to see satellites in the northern direction. This situation is a concern since the last known fix may have been a couple of blocks behind the actual vehicle position. Aside from this incident, most of the longer gaps were common in all three receivers.
The average gap time in this urban canyon test shows an increase from 12.6 sec for the roof, to
19.3 sec for the front antenna, and 13.5 sec for the back antenna. The maximum gap time also increases - this time from 79 sec to 153 sec and 132 sec, respectively. Thus, the maximum time without a position fix was about 2.5 minutes for the front antenna versus
1.3 minutes for the roof antenna.
Again, there was very little change in the average, maximum, and minimum DOP values from antenna to antenna. In fact, in some instances (such as the maximum HDOP), the values improved for the antennas in the
passenger compartment. Note that the maximum PDOP is always slightly below 6.0. This is because if the PDOP is 6.0 or higher, the receiver will automatically compute a 2D fix and hold the height fixed, which tends to minimize the horizontal error.
Table 3
contains a cross comparison of the antenna locations. The number 40 represents the number of fixes that the back antenna did not have that the front antenna did. If an RF switch were implemented, these 40 fixes might be
recoverable. However, when a new antenna is selected, the receiver will take a few seconds to reacquire the satellite signals and compute a position fix. In the event that a new satellite has risen, the ephemerides will not be immediately available and may take 30 sec to download. In most applications, it is probably better to use the last known fix, which may be a few seconds old, rather than wait for the new fix to be computed, which may or may not even be available.
Urban canyon test 2
The
second urban canyon test lasted for 3,840 sec (64 minutes), and the results are shown in
Tables 4
and
5
(and
Figures 9, 10, and 11
). The coverage for this test ranged from 86.4% to 90.4%, which is a very narrow margin.
The average gap time changed by less than 2 sec from antenna to antenna, and the maximum gap time increased from 70 sec to 109 sec - just under 2 minutes. Finally, the DOP values showed
little movement from the roof to the passenger compartment antennas.
The front antenna missed a total of 524 fixes. The cross comparison in Table 5 shows that of the 524 missed fixes, the back antenna had 224 fixes, which is 5.8% of the total test time. Of the 441 fixes that the back antenna missed, the front antenna had 141 fixes, which is 3.7% of the total test time. This shows that adding a second antenna to a system will, at best, improve the total coverage of the system by only a couple of
percentage points. These numbers are best case, assuming that the receiver would be able to re-acquire satellites and resume computing fixes with zero delay.
This test run was not as severe as the first urban canyon test. The first test had more time in the most severe portion of downtown Chicago in the financial district. The statistics for the second test are more representative of the urban canyons in major city centers.
Impact on ITS
Many navigation systems today make use of dead-reckoning
sensors that augment the performance of GPS by bridging the gaps when GPS is not available. An odometer, compass, rate gyro, and even ABS wheel sensors can be used to propagate positions using speed and heading when the GPS signals are blocked. In navigation systems with augmentation, a single GPS antenna can be mounted in either of the locations considered here without degrading the system performance.
In vehicle tracking or theft recovery systems, it is not critical to maintain a continuous position. For
these systems, gaps in coverage are tolerable, so again, there is flexibility in the selection of the GPS antenna location.
For emergency call systems, the position is reported to a monitoring location. This position is then passed on to the required emergency dispatchers so that the vehicle can be located efficiently. In this type of system, an accurate position determination is more important. A position fix that is at most a block away is desired. The largest gaps in the urban canyon environments
last for more than one block. This clearly illustrates the advantage to establishing voice contact with a driver in the event of an emergency so that the position can be clarified if necessary. In the event that the driver is unable to confirm the position, an emergency response unit would have to search a distance of a couple of blocks for the vehicle.
With vehicle electronics, an important consideration is the system installation process and the serviceability. The location of the equipment and the
wiring must be easily accessible in order to keep installation and servicing costs down and reliability high. Installing the GPS antenna under the back package shelf is ideal for installation and servicing.
The results of the testing clearly show that the best location for a GPS antenna is on the roof of a vehicle where there is a clear view of the sky. However, there is not a significant degradation in the performance of the GPS receiver when the antenna is in the passenger compartment as compared to the
roof.
The results also show, that using a dual antenna configuration does not improve the GPS coverage significantly. Finally, the results show that there is little difference in the performance between a front and back mounted antenna. Hence, the choice can be made using other criteria. It is best to have the antenna as close to the base of the glass as possible. It is also best to have the antenna located under the glass with the lowest slope angle. In the event that one of the locations allows a
better mounting position, it should be used.
J. Blake Bullock is an applications engineering manager for Oncore GPS products with Motorola's Automotive and Industrial Electronics Group in Northbrook, IL. He received his BS in surveying engineering and his MS in geomatics engineering from the University of Calgary. He can be reached at g12645@email.mot.com.
Puzzled by a network processing design issue?
