Design Features March 3, 1997

Time, position, and velocity?
Just ask your GPS chip set

Maury Wright, Technical Editor

GPS is poised to invade everything from cellular phones to backpacks with new chip sets that implement a complete receiver using just two ICs plus memory. Chip-set prices, meanwhile, have hit the $30 mark, making GPS suitable in virtually any application that needs position data, including remote embedded applications that need only an accurate clock or calendar.

Using 24 satellites in orbit more than 10,000 miles above the earth, the Global Positioning System (GPS) implements a radio-navigation scheme that allows receivers to accurately determine time, 3-D position, and velocity. Whether you are working on the latest car-navigation system or on a custom embedded system, you may need to consider using GPS. Proven applications include marine navigation, geological exploration, and surveying, but new, highly integrated chip sets make handheld receivers affordable for backpackers, too. IC vendors have reduced chip counts to two or three GPS-specific ICs and two or three memory components. In some cases, you can even use the microcontroller that's inherently found in the chip sets to handle the GPS task and your own application code.

GPS devices operate by receiving spread-spectrum RF signals from a minimum of four satellites. The receiver then uses triangulation techniques to calculate position, time, and velocity. The receiver needs four, rather than three, satellites to eliminate clock-synchronization error in the receiver. For more information on GPS operation, see box, "GPS primer."

The growth of GPS over the past few years is astounding. In Japan, more than a half-million cars will be sold with GPS-based navigation systems this year, and that number will cross the 1 million mark in 1998 (Reference 2). In the United States, Lincoln Mercury is including a GPS receiver in some cars for emergencies. In case of a breakdown or another emergency, the driver need only press a button to have a cellular phone dial a preset number and report the vehicle position according to the onboard GPS receiver. You can also find GPS receivers on golf carts, providing yardage and advice based on your position. Handheld receivers, meanwhile, can cost as little as $200. In perhaps the best recent example of GPS capabilities, Captain Scott O'Grady, who was shot down over Bosnia in 1995, used a Trimble (Sunnyvale, CA) Navigation receiver powered by a Motorola chip set to inform rescuers of his position.

Within a few years, you may find a GPS receiver in every cellular phone. In the United States, such phones will, by law, have to provide a position fix by 1999. The newest GPS chip sets are approaching the level of integration that can make GPS cost-feasible in every cellular phone. In fact, the prices are so low that GPS may prove to be the most economical way to keep time in remote embedded systems. The GPS satellites include atomic clocks accurate to within nanoseconds.

In addition to low prices, the chip-set vendors are making the technology available so that non-GPS experts can design a receiver into a system. Until recently, OEM designers had to turn to modules designed by GPS experts. Modular receivers, PC Cards, or even PC add-in boards, in fact, can still be viable choices in low-volume applications. Even if you choose the module route, however, carefully evaluate the chip architectures, choose the best one for your application, and buy a module from the chip-set vendor or one of its OEMs. If your application supports significant volume, the chip-set vendors stand ready to supply reference designs and development tools to support your effort. Think of GPS chip-set evolution as analogous to that of modem chip sets. Ten years ago, only the most experienced communication engineers designed modems at the chip level; others integrated modules into their systems. Today, you regularly find modem chips designed into motherboards, single-board computers, and even one-of-a-kind custom systems. GPS chip sets are just beginning a transition to mainstream status. Figures 1 through 4 from SiRF illustrate the ways you can design in a GPS chip set.

Evaluating GPS chip sets

Comprehensively evaluating GPS chip sets requires a mixture of feature scrutiny and live tests. Moreover, issues such as familiarity with a µP, software support, and reference designs may prove more attractive than hard-ware/software features in some cases. Remember, also, that system cost goes well beyond chip-set cost.

Today, you have complete chip-set offerings from only six vendors to consider (Table 1). Generally, each chip set includes an RF front-end IC that is essentially a down-converter from the GPS satellite transmission frequency of 1.575 GHz to a rate of around 4 MHz. SGS-Thomson is actually just sampling its monolithic front end, having previously required a discretely implemented down-converter.

The Rockwell Zodiac chip set uses a "MonoPac," which is essentially a package with two separate dice. The distinction is subtle, but competitors claim the design results in slightly higher prices. The price in Table 1 seems to confirm this claim, but Rockwell has actually not made public a current price for its Zodiac. The listed price was published at the Zodiac chip set's introduction in September 1995. Rockwell GPS Product Manager Vinay Gokhale points out that the chip set has been widely used in low-cost products. Without question, Rockwell currently holds a leadership position in market share. That position, however, can be attributed both to the fact that Rockwell was first to market with an integrated chip set and to the quality of the implementation.

At first glance, it's tough to differentiate the RF chips. The various implementations use different frequency plans (differing numbers of IF stages and choices of intermediate frequencies), but no differences in the feature set jump off a data sheet. GEC Plessey Marketing Manager David Richardson claims that the company's GP-2010/2015 offers several advantages. For example, Richardson says that the three-stage down-converter offers superior rejection of out-of-band signals, and it allowed the designers to adopt a frequency plan that minimizes interference from other wireless devices. He points out that the frequency plan is optimized to allow the GPS RF device to operate alongside a cellular phone with no signal-jamming problems.

Not everyone agrees that three IF stages provide an advantage. SiRF Marketing Manager Greg Turezky points out that his company's single-stage converter, which produces a single IF, is ultimately less likely to cause jamming problems. The other vendors' implementations use two-stage converters, but all of the companies claim no problems operating around cellular devices. Philips is the newest supplier of GPS chip sets, with samples due around press time. Its SA1570 RF device features a two-stage design, but the IFs can be programmed for maximum frequency agility. Ultimately, you should test a reference design in the target environment if you suspect frequency conflicts might create a problem.

Another major difference in the RF chips is in the A/D-converter implementation. With the exception of Motorola's products, the chip sets all integrate the A/D converter in the RF chip. GEC Plessey and SiRF include 2-bit quantization, which yields a 2-dB advantage in S/N ratio during operation. The other vendors either rely on 1-bit quantization--as simple as a comparator--or don't specify that detail. The accuracy demands of some applications, such as aviation navigation, require 2-bit quantization.

DSP-based correlators

All of the chip sets include CMOS chips with a signal-processing block that accepts the digitized bit stream. Some of the vendors label the block "DSP," but, in reality, the so-called correlator functions use hard-wired state machines to implement the signal-processing functions. The receiver in SGS-Thomson's ST20-GP1 requires a dedicated correlator for each satellite channel (Figure 5). The circuit at-tempts to match the incoming signal with a stored copy of the GPS pseudorandom-noise (PRN) sequence. By correlating the sequences, the receiver can calculate the amount of time and, therefore, the distance the one-way RF signal travels from satellite to receiver.

The implementations vary in how the companies integrate the DSP block (Table 1). Philips, Rockwell, and SGS-Thomson offer ICs that combine the DSP function with a processor. The other three vendors use an external processor. All the chip sets require external ROM and RAM. For the most part, the vendors have chosen different processors, although Motorola and SiRF both use 68K-compatible µPs.

The choice of processor, DSP partitioning, and development support for the chip set could be important factors in your choice. These three factors significantly affect development time and costs. First, consider both chip-set and system cost. The chip sets with integrated processors would presumably offer lower chip counts and potentially lower costs. In the case of Philips SC1575, the $30 (10,000) price in-cludes turnkey binary GPS software. Moreover, the company claims that complete OEM system costs can be as low as $45 (250,000). Philips, however, hasn't fully characterized its design and isn't sure how much spare processing power might be available on the XA microcontroller for user code. Moreover, the company is not currently planning to sell a development system. Should you need more functionality than the Philips reference design offers, you may have to add a second processor. Philips partnered with GPS vendor Ashtech (Sunnyvale, CA) in developing Philips' chip set and refers OEMs that need to modify or enhance the reference software to Ashtech.

Rockwell doesn't quantify how much spare processing power can be harnessed for a user application in the Zodiac chip set. The company does point out that existing customers have built handheld, automotive, and marine GPS receivers using no resource other than the on-chip AAMP2 microcontroller. SGS-Thomson, meanwhile, boldly states that GPS-specific code requires less than 50% of the computation cycles available in the company's ST20 RISC core, thus leaving computer power for almost any user application. The company also offers a complete, low-cost development kit, although its GPS binaries and code libraries are pricey, starting at $7000.

Spare MIPS for user applications

Motorola, SiRF, and GEC Plessey all have ample processing power for GPS and user code with their external processors. Moreover, should designers using these products require more or less performance, they can choose a faster or slower processor. Motorola and SiRF, in fact, may have a significant advantage because of their 68K-centric designs. Many designers would prefer to work with the well-known processor architecture and instruction set. Motorola offers binary GPS software and a complete reference design with its $1200 Oncore evaluation kit. SiRF currently offers an evaluation kit but is still in the planning stages of a development kit and reference design.

Having processing power to spare can also reduce costs in other parts of the system. For example, Motorola uses a temperature sensor, a low-cost crystal, and temperature-compensation software to generate the precise clocks that GPS requires. The software scheme saves costs and pc-board real estate compared with designs that require an external temperature-compensated oscillator.

An external processor can also pay off in cost savings when you might not expect it. SiRF's Turezky points out that a typical GPS receiver might require an LCD, a keypad, and other peripheral functions. The immense library of Motorola microcontrollers allows designers to choose a processor with functions such as LCD and keyboard controllers integrated on chip. GEC Plessey may develop the same sort of advantage over time as the ARM processor family proliferates.

Evaluating performance

The final evaluation criterion for the GPS ICs centers on performance, and performance is highly subjective. None of the chip sets offers significantly better positioning accuracy than any other because the Department of Defense ultimately controls the accuracy. You may find that, in some cases, carrier-phase tracking can improve accuracy. Motorola, for example, offers carrier-phase augmentation for the more typical code-phase GPS tracking. Primarily, however, performance differences come down to how fast the receivers can acquire a satellite and begin tracking position--the time to first fix (TTFF), measured in seconds. TTFF is typically specified in four ways:

Hot start: TTFF when a GPS receiver has been tracking a satellite previously and has stored the following information about the satellite in battery-backed memory: time, location, almanac, and ephemeris (parameters defining the satellite's orbit). For example, a receiver in a car can perform a hot start after the car is parked for an hour or two, but the ephemeris data is good for only two to three hours.

Warm start: TTFF when the receiver has no current ephemeris data but has time, location, and almanac data either supplied by the user or stored from previous operation. A receiver in a car parked overnight could likely perform a warm start. Note that the difference between hot- and warm-start specs is at least 30 sec, because it takes that long to receive updated ephemeris data.

Cold start: TTFF when the GPS receiver has no data from previous operation in battery-backed memory and no user data.

Reacquisition: TTFF after a momentary blockage of a satellite due to circumstances such as when a car with GPS receiver passes under a bridge.

Table 1lists TTFF specs based on the above criteria. Unfortunately, the vendors don't necessarily measure results in the same way. Moreover, the listed specs are average and not worst-case times. The specs are also software- and system-dependent. If TTFF is key to your application, evaluate each technology in your target environment.

The chip sets that feature faster TTFF specs achieve those results in one of several ways. Motorola's chip set, for example, features the second fastest cold-start time. The company claims that its architecture is heavily software-intensive during initial satellite acquisition, and the spare power in the 68331 µP results in the 90-sec cold-start performance.

Wide PRN windows speed TTFF

SiRF's chip set is even faster, and the company's designers added more signal-processing hardware to achieve the 60-sec result. The design uses 10-bit-wide windows in the correlator channels when trying to acquire a satellite signal. Moreover, you can cascade all 12 channels to make a 120-bit window. The PRN sequence is only 1024 bits long, so the SiRF window quickly searches through the received bit stream.

SGS-Thomson, meanwhile, offers the fastest hot-start performance and matches Motorola's cold-start performance. SGS-Thomson also takes a hardware approach to accelerating TTFF. During acquisition mode, the receiver operates with a sampling rate four times faster than the rate used during routine tracking operations.

The number of channels is another differentiating factor among the GPS chip sets, although all but Philips' and Motorola's support 12 channels. Strictly, no more than eight satellites are ever visible to a GPS receiver at any time, and it's a rare occasion in a flat area when eight are visible. Some of the vendors claim 12 channels can help accelerate TTFF, but the eight-channel Motorola Oncore chip set sports some of the fastest specs.

Differential GPS (DGPS) provides one possible future use for extra channels. DGPS increases the accuracy of GPS receivers, and all of the chip sets include the hardware necessary to support DGPS as users typically implement it today. In most cases, a DGPS receiver must accept differential data from a source such as a Coast Guard beacon or via an FM subscription service. These signals typically require a dedicated radio and an antenna that transmit the correction data to the GPS chip set via a serial connection. And, although the chip sets all include a UART to support such serial connections, the standard software from the chip-set vendors doesn't necessarily support DGPS. Moreover, the differential radio can cost significantly more than the GPS receiver, because the radio is typically a specialty, low-volume product. Down the road, however, other satellites will provide differential corrections over GPS channels (see box, "Looking ahead"). SGS-Thomson's chip set is already designed to support such an environment.

Hands-on evaluations

With so few differences among the available products, you may find that you must base your choice on a hands-on test with evaluation units. Based on the company's experience, Rockwell's Gokhale recommends two comparative tests that designers should perform in the target environment. First, you should measure each receiver's performance as it moves around a test track. Specifically, you should measure the percentage of time the GPS receiver spends in navigation or tracking mode relative to the percentage of time spent in acquisition mode. The test determines how well the receiver handles obstacles, such as mountains, trees, bridges, and buildings. Second, Gok-hale recommends testing ground-track aesthetics. For example, a receiver moved along a square test track should yield square position results.

References

1. Gallant, John, "GPS receivers: System revolutionizes surveying and navigation," EDN, Jan 7, 1993, pg 31.
2. Quinnell, Richard A, "Directionally dyslexic? Don't worry: The car knows the way," EDN, Dec 21, 1995, pg 37.

You can reach Technical Editor Maury Wright at (619) 748-6785, fax (619) 679-1861, ednwright@mcimail.com

| EDN Access | Feedback | Subscribe to EDN | Table of Contents |