In the Summer of 2011 we decided to prepare a couple of proposals for the European Satellite Navigation Competition.
With our surprise and delight we won the Italian prize and reached second place in the DLR special prize!
For the curious, the two awarded ideas were:
Interference free GNSS receiver (IFRX) and signal analysis tool
Description: Nowadays one of the most fashionable topics in the GNSS domain is interference. GNSS signals are buried under the noise and detectable only thanks to the processing gain obtained by code de-spreading techniques. Therefore they are very susceptible to interference. GNSS equipment manufacturers rely on the nominal specifications of the signal to tune the gain in their receivers as well as choosing the number of bits to use in the quantisation process. An Automatic Gain Control (AGC) is usually present in the Front End (F/E) in order to comply with different antennas which could possibly mount LNAs with different gain characteristics. In absence of interference, the AGC solves most problems. On the contrary, when the in-band power density differs from nominal, the number of bits used in the quantisation process plays a fundamental role as more bits lead to higher dynamic range. As of today, two main types of GNSS receivers exist: those targeted to mass market applications, and those targeted to professional applications. Fundamentally mass market receivers aim at maximising the availability of a solution versus size and power consumption, whilst high-grade receivers aim at maximizing accuracy and reliability with little regard to size and power. To maximise availability in a mobile phone where multiple kinds of transmitters share the same enclosure, manufacturers shrink the bandwidth of their passband filters to prevent unwanted noise to interfere with their ADCs, and they increase the number of bits of the quantisation process to improve robustness to in-band jammers. To maximise accuracy in a (typically) open-sky scenario of a survey grade receiver, manufacturers choose wide bandwidth for their filters and few quantisation bits for their high-speed ADCs. Considering how critical high-grade receivers are already in many civil applications (let go military) the allocation of new wireless services in L-band (see LightSquared) poses a tangible threat. On the other hand, considering the increasing availability of Personal Privacy Devices (PPD), mass market receivers have to adapt in order not to lose ground on their expected market penetration against other positioning technologies. Because of the reasons explained above, the need of moving to a higher number of bits in the quantisation process is a trend that can be anticipated in both markets (uBlox already markets a 5-bit digitisation engine as anti-jamming technology). But as the GNSS receiver uses more bits in the sampling process, its resemblance with a general purpose DSP with fixed arithmetic becomes more obvious. At 8-bit, the advantage of having dedicated silicon to do GPS correlation against the advantage of a flexible software implementation on a general purpose DSP can become questionable. At first the interest in SDR GNSS receivers was boosted by the prospective of being able to replace existing mass market hardware receivers in portable devices with a software application which would be very low cost and very easy to upgrade. This never happened as the market expectations for Location Based Services grew so much that silicon manufacturers invested heavily to bring down costs of their chipsets while improving their specs: that prevented SW receivers from being a reality.
Now that GNSS has become so popular and so many applications rely on it, the proposed idea anticipates a need for robustness to interference and therefore a trend in the market. With a difference: specialised hardware receivers will necessarily have to become more similar to general purpose DSPs because of the physical need for higher dynamic range in the spectrum of interest. The solution proposed uses a low-cost combined architecture FPGA+CPU, where a small FPGA is in charge of performing the raw signal monitoring and conditioning (e.g. spectrum analysis, filtering, frequency blanking) to combat (mostly) jamming and the embedded CPU runs a software receiver with 8 or 16 bits and addresses more clever techniques of interference (e.g. spoofing or meaconing). This system would be an ideal candidate receiver for security critical applications, but would also work very well as interference characterisation tool. Not from a
technical perspective but from the business case point of view, it is also of extreme importance to equip such instrument with a GSM/GPRS modem to support online updates. The technology enabler for the proposed idea is the availability of low-cost high-density FPGAs such as Xilinx
Spartan-6 series (and soon Artix-7), powerful embedded CPUs with SIMD instruction sets like the Intel Atom or the ARM Cortex-A8, and miniature GSM modems like the Telit GE865. A closer look at the application processors shows that an Intel Atom N270 (released March 2008) is capable of delivering 3300 MIPS in standard benchmarks, which is about twice as much as what can be delivered by an ARM Cortex-A8 (released Feb 2008) clocked at 800 MHz. Considering a realistic figure of about 100MIPS per tracking channel, they both can support all GPS constellation in view. The main characteristic of the Intel Atom is the availability of the SSSE3 instruction set (not to be confused with SSE3) which offers PSIGNB to do simultaneously 16 multiplications of a signal by a binary code, and PMADDUBSW which is basically a byte-wise MAC (Multiply-Accumulate). The main characteristic of the Cortex-A8 is the NEON core which supports with VMLA a SIMD MAC operation on 8, 16, and 32 bit.
"Miniaturised pseudolites for mass wildlife tracking"
Description: Whenever there is the need for tracking mobile subjects with meter-level accuracy, GNSS equipment comes at hand. In wildlife tracking, for example, most solutions use low-power GNSS receivers with some kind of transmitter (GSM, Bluetooth, ZigBee, ISM ...). GNSS technology made great progress in the last five years so nowadays receivers measuring less than 1 cm2 have power consumption in the order of a few tens of mW, long term ephemerides for warm start beyond almanac validity, acquisition engines capable of sub-second full constellation search, and sensitivity levels around -160dBm. With the above capabilities is possible to implement miniature tracking solutions which exhibit a battery life of months or even years. As the number of subjects to track grows, the investment necessary to equip them with such tags becomes significant. The price of a GPS receiver is in the order of 10$, and cellular modems can be several times more expensive than that. The problem that the authors see is that such tags need a GNSS receiver (positioning sensor) plus a RF transmitter (communication channel). And in most of the cases, the information to be transmitted is negligible compared to the one to be received (NMEA, versus several mega-samples of GPS signal) and carries a lot of unnecessary overhead (modem connection protocol versus 80 Bytes of a NMEA sentence). Moreover, the users of the system need more (cellular) receivers to continuously monitor the status of the group under study. Given this scenario, the authors propose to build miniature transmitters which combine the transmission capability with ranging information (a unique wideband sequence), in the same fashion as GNSS satellites do. Terrestrial satellite-like transmitters are often called pseudolites and many different kinds of pseudolites exist. In this case they should have as little intelligence as possible, and the receivers should perform the triangulation instead. To determine the position of a transmitter on the ground, three or more receivers will be needed. The system would work as a conventional GNSS but reversed, having synchronised receivers and asynchronous transmitters. The fields of applications are all of those which need a large number of subjects or objects to be tracked in a confined area for a long time by means of non-invasive miniature tags. Since transmission in the GPS bands is prevented by international regulations, the 433MHz ISM band is proposed instead. The authors consider this frequency a good compromise between the size of the antenna (higher frequencies have shorter wavelengths and need smaller antennas) and the penetration capability in a partially obstructed environment. Being a popular band, off-the-shelf ICs are easily available and this reduces the cost, power, and size of the final solution. Having almost 2 MHz of useable bandwidth, this allows for enough resolution in positioning with code only. The spreading sequence for the BPSK modulation will be at about 1Mchip/sec. Since the spreading code already identifies the tag, there is no need for data modulation therefore the PRN sequence can be stored in flash memory and played back at wakeup time. Between transmissions, the tag can go to a power save-mode where consumes about 5uA. The major drawback of the 433MHz ISM band is the interference, which is mostly pulsed and narrowband (ASK, FSK). Here is where the receiver architecture makes a difference. First of all, receivers should all be synchronised (for example with GPS) so that measurements can be referred to a common instant. Since the modulation used is non-standard for that band the authors propose the Software Defined Radio (SDR) approach: signal processing will happen on a standard CPU. It is well known that hardware GPS receivers use 1 or 2 bits for the A/D conversion, but this is only permissible in an interference-free environment where the nominal signal dynamic is tuned and predictable. In the above system’s case distances between transmitters vary greatly (there is a near-far problem) and external interference is allowed by design. So, a multi-bit A/D Converter (ADC) can address both issues because of its large dynamic. It is worth noting that when signal processing is done in silicon having 2-3 bits translates in a saving in terms of transistors, whereas most Digital Signal Processors (DSP) use with 16-bits integer arithmetic anyway, therefore such ADC has the dual purpose of exploiting the capabilities of the software while providing interference immunity. An Automatic Gain Control loop should also be implemented in the FPGA doing the signal conditioning. The coverable range of the system is a trade-off to be evaluated. If the maximum transmissible power in band is 10dBm, experience tells that several hundreds of meters can be achieved with narrowband signals and more should be expected by CDMA signals. Extending the length of the spreading codes (at the expense of battery life), more processing gain is possible at the receiver level and longer distances can be covered.
You will find more and more references to both ideas on our pages as we try to bring them to the market.
And many thanks to everyone for the contribution and the interest in GNSS applications!