Interchangeability Accomplished
Tri-Band Multi-Constellation GNSS in Smartphones and Tablets
This article presents a single-chip BeiDou/Galileo/GLONASS/
GPS/QZSS/SBAS architecture for use in cell phones and tablets. The authors explain the advantages to end users of multiple constellations. They also examine the details of system interchangeability, multi-system issues, and how assisted-GNSS data operates with all constellations, including BeiDou.
By Frank van Diggelen and Kathy Tan
With GPS, GLONASS, SBAS, BeiDou, QZSS, and Galileo there are over eighty operational satellites. Why do we need all these satellites in the first place? The answer is simple: in urban environments we want a few (six to eight) good satellites with an unobstructed line-of-sight (LoS) to the receiver and good horizontal dilution of precision (HDOP). In order to achieve this, we need many more satellites in space than any single constellation. In this article, we address the following issues.
- Receiver intersystem RF bias with a tri-band front-end. BeiDou uses a different RF section than GPS/Galileo/QZSS/SBAS and GLONASS. As a result, there is a receiver intersystem bias between BeiDou and each of these other systems—not just because BeiDou is on a different frequency, but because of the different RF path through the receiver. We explain how this bias is calibrated and removed.
- In the space segment there are intersystem biases primarily caused by differences in time standards. We discuss time management and show how the different systems can be made interoperable.
- BeiDou Assistance. In order to realize the benefits mentioned, we need infrastructure deployment for BeiDou assistance in accordance with 3GPP standards. We will discuss what is available, and what is left to do.
- Coverage outside of China. Europeans can see more BeiDou satellites than Galileos. At the time of writing (March 2014) they could see approximately twice as many. Thus, when used in a multi-GNSS receiver, BeiDou is far from being just a regional system. We will provide coverage analysis, and live-test data, including a focus on Europe.
- Finally, we will demonstrate all of the above in practice, explaining and showing how interchangeability is achieved, and where first fixes can be computed with no more than one of each satellite type.
Figure 1 illustrates the point referenced at the beginning, that we need many more satellites in space than any single constellation.
All of the lines in Figure 1 show signals that were actively tracked by the receiver at the position shown on the right. The orange lines are to satellites that are blocked, but the reflected signal is tracked. We do not want to use these measurements if we can help it, so we need many satellites to provide enough LoS signals.
Let’s look at the HDOP of the LoS signals. In this example, the HDOP for the three LoS GPS satellites was 50. For the three LoS GLONASS satellites, the HDOP was 45. However, with the combined GNSS constellation, the HDOP for the six LoS satellites was 2.2. In other words, we expect about a 20x accuracy improvement by using the combined constellation.
There are many places and times in cities where we see just one or two direct LoS signals from a particular constellation, and we need more than just GPS and GLONASS to get the desired number of good signals, thus explaining the desire and need for all available constellations.
We’ll now look at the coverage provided by BeiDou2, which has five Geostationary satellites (GEOs), five inclined Geosynchronous satellites (GSOs), and four Medium Earth Orbit satellites (MEOs). With this 14-satellite constellation, the global coverage is as shown in Figure 2. This figure shows the percentage of time in a day that four or more BeiDou satellites are visible above a 10-degree mask angle. In the Asia-Pacific region, where the GEOs and GSOs are positioned, the coverage is predictably 100 percent. In fact, there are seven or eight BeiDou satellites visible in much of this region most of the time.
As shown in Figure 3, outside the Asia-Pacific region the coverage is also interesting. We see that at least four BeiDou satellites are available over Europe about half of the time. This is quite significant given the previous discussion; even one or two extra satellites can make all the difference in an urban environment. Another notable fact is that, for now at least, Europe can see more BeiDou than Galileo satellites.
Technical Requirements
There are five significant technical requirements that we want to satisfy when creating a multi-GNSS receiver for consumer applications:
Three Separate RF Paths. To acquire and track all of the satellites already mentioned, we need three separate RF paths. Details follow in Section 3 (Front-End Architecture).
Search and Track capability for all visible GNSS satellites. The receiver must have the ability to search a very large number of code-frequency bins at once.
Host-Based. As much as possible, we want to make use of the host application processor (AP) and memory. This allows for tight integration with assistance data (which is coming from the host), other sensors, and other wireless data (such as Wi-Fi and Bluetooth for indoor locations). A host-based architecture also keeps size and cost as low as possible.
With Host-Offload. A significant trend in location applications is the need for “always-on low power” location. The host AP cannot be used for continuous position updates, since it draws too much power. So, while we want host-based location when the host AP is active (such as when navigating with turn-by-turn directions and a map), we also want a host-offload capability so that the GNSS chip can compute positions internally while the host is asleep.
Interchangeability. The ultimate requirement for multi-system GNSS is the ability to use any combinations of satellites as if they were all in the same constellation. This is summarized as “any four satellites will do.”
Front-End Architecture
From a cell phone/tablet perspective, the signals in space are all in the L1 band, with frequencies as shown in Figure 4. The key architecture feature of the GNSS front-end is that it should have three separate RF chains for the three separate frequencies-of-interest; see Figure 5.
Baseband Architecture
The preferred architecture of a chip, as shown in Figure 6, is host-based to take advantage of the large host CPU when it is active. When the host CPU is asleep, a small, low-power, on-chip CPU is leveraged for background “always on” location. This enables applications such as geofencing to run without significantly reducing battery life.
When the host is active, such as when you are actively using the phone for turn-by-turn navigation, the host AP is on and we want to make as much use as possible of the host AP and memory. This allows for tight integration with assistance data coming from the host, other sensors, and other wireless data (such as Wi-Fi and Bluetooth for indoor locations). A host-based architecture also keeps size and cost as low as possible, even with host-offload capability, which adds very little to the size of the chip.
Receiver Intersystem RF Biases
With the three different bands of frequencies, we will get RF group delays in the receiver front-end. These must be calibrated out by the receiver’s designer as part of the chip’s system design. If the group delay between BeiDou and GPS is not calibrated, it will lead to approximately three meters of bias between the two systems (Figure 7). Once it is calibrated, there is essentially no bias.
Satellite Intersystem Biases
Different GNSS constellations run off their own master clocks; referenced to different realizations of UTC. GPS is referenced to UTC (USNO), QZSS is referenced to UTC (NICT), GLONASS to UTC (SU), BeiDou to UTC (NTSC), and Galileo to UTC (INRIM). GLONASS UTC (SU) differs from the others by 3 hours.
Furthermore, different systems treat leap seconds differently. This is indicated by the red arrows in the clocks in FIGURE 8. GPS, QZSS, BeiDou and Galileo system times are continuous and ignore leap seconds. Thus, each system time is ahead of UTC by a number of leap seconds. GPS time started in 1980 in synch with UTC; there have been 16 leap seconds since, so now GPS is 16 seconds ahead of UTC. QZSS and Galileo system times were started in synch with GPS. BeiDou system time was started in 2006 in synch with UTC; there have been 2 leap seconds since, so now BeiDou is 2 seconds ahead of UTC. GLONASS system time, on the other hand, includes leap seconds.
Apart from this, each of the different realizations of UTC is within several nanoseconds of the others.
To combine measurements from these different systems and avoid any time-induced intersystem biases, we need to resolve the time offsets. Each system transmits the delta-time between its system time and the systems that preceded it, as listed in Figure 8. To combine the systems, we either need to decode these data messages or obtain the delta-time values from Assisted GNSS.
Note, however, that in the BeiDou broadcast Nav message the intersystem time-offset data values are all set to zero (even though the true offsets are not zero).
Assisted-GNSS Including BeiDou
Assisted GNSS, or A-GNSS, increases sensitivity and decreases the time-to-first-fix of a receiver by providing assistance data in the form of the receiver’s approximate position, time and frequency, as well as all data that the receiver might decode from the broadcast signals. The assistance data may also include data beyond what is broadcast, in particular, let’s focus on BeiDou time offsets. The BeiDou time offset to the other systems is included in the BeiDou broadcast Nav message as shown in Figure 8; however, at present these data values are all set to zero (even though the true offsets are not zero). Thus, in order to get these offsets and integrate BeiDou properly into a combined GNSS system, one must compute the offsets at a reference station and provide them as part of the assistance data, as shown in Figure 9.
Commercial Implementation
The preferred architecture described in this article has been implemented in a commercial GNSS receiver that is now available for commercial host-based products, such as cell phones and tablets. The chip, Broadcom’s BCM47531, is the first consumer GNSS chip with a tri-band front-end capable of acquiring and tracking satellites from GPS, SBAS, QZSS, GLONASS, and BeiDou constellations, simultaneously; and operating in host-based mode for navigation and in host-offload mode for Always-On location.
Broadcom has collaborated with leading smartphone manufacturers to launch the first wave of BeiDou enhanced consumer smartphones. Figure 10 shows one of these smartphones being tested in Europe. Note the number of BeiDou satellites in view. As predicted by the availability plots shown earlier, there are many BeiDou satellites in view (in this case, six).
Interchangeability: Any Four
Now that we have addressed all of the major issues related to integrating different GNSS systems (in particular BeiDou), we can demonstrate the payoff.This is the achievement of interchangeability, where any GNSS satellites can be used together, as if they all belong to a single constellation. Figures 11 and 12 show assisted cold starts, where first fixes are obtained with no prior knowledge other than that provided by A-GNSS data. In each case, we show a different combination of satellites; including one satellite from each of four different constellations, and all four from BeiDou.
Multi-Constellation Robustness
While this article was being edited, the GLONASS system provided us with the most dramatic demonstration yet of the need for, and benefits of, multi-constellation receivers. On April 2, 2014, the GLONASS system failed spectacularly for a period of 11 hours. Receivers that used GPS and GLONASS had very large position errors, or no positions at all. While the receiver discussed in this article, the BCM47531, operated seamlessly. This receiver tracked GPS, GLONASS, QZSS and BeiDou satellites, correctly identified the faulty GLONASS satellites, and automatically stopped using them.
The details of the incident are as follows: The GLONASS control system uploaded incorrect orbit data to several satellites. When receivers used these satellites they had position errors of hundreds of meters, or no positions at all. At that time, the BCM47531 was being tested alongside a GPS/GLONASS receiver, and we have the data to show what happened. The receiver using only GPS/GLONASS suffered position errors of ten thousand meters, and long periods with no position at all; at the same time the multi-constellation receiver produced continual positions with normal accuracy. Figure 13 shows the test data — the left most image shows the route being driven, the middle image shows the data from the GPS/GLONASS receiver, and the right image shows the data from the BCM47531 multi-GNSS receiver. Figure 14 shows the details of the multi-GNSS receiver, you can see that no GLONASS satellites are being used.
This incident may raise the question: Why use GLONASS at all, why not just GPS? The answer is that in urban canyons, such as where this test was done, GPS alone does not have enough satellites to give the performance now expected in consumer products — for the reasons explained in the beginning of this article. Also, GPS, although it has been more reliable than GLONASS, is not immune to failures or jamming itself. The lesson of this incident is that reliability and accuracy comes from the combination of all the available constellations, with a receiver that can use the signals interchangeably.
Conclusion
We have shown the preferred architecture for a consumer GNSS receiver that includes all of the available constellations. We have addressed the major requirements of such a receiver for the consumer market, in particular, for cell phones and tablets. A receiver that meets these requirements is now available, the Broadcom BCM47531, has been designed into a new generation cell phones and tablets for 2014. Finally, we have shown how, with this receiver, the ultimate GNSS goal of interchangeability can be achieved.
Frank van Diggelen is vice president of technology at Broadcom Corporation, a consulting professor at Stanford University, and inventor of coarse-time GNSS navigation, co-inventor of Long Term Orbits for A-GNSS, and author of A-GPS: Assisted GPS, GNSS, and SBAS.
Kathy Tan is a senior principal engineer at Broadcom Corporation. She has worked on GNSS development and Assisted GNSS for Ashtech, Magellan, Global Locate and Broadcom. She received her MS and BS in electrical engineering from Fudan University, China.
Fascinating and engaging presentation, especially with the real-world GNSS receiver responses to the April 2 GLONASS event. Thanks!
I am curious what the power consumption is for a tri-band receiver relative to single band. Also what the size (volume) difference is.