Furthermore, it inevitably degrades the front-end noise figure of the GPS receiver, thereby offsetting performance gains relative to alternative approaches.
CDMA Mode Switching.
In the case of a CDMA handset, the main alternative is to switch modes between GPS and the handset. The two subsystems cooperate so the receiver "listens" only during timeslots when the handset is not permitted to transmit.
For example, during the 16-second response period, the transmitter may only be able to transmit for a fraction of the time in bursts of a few hundred milliseconds. While this process allows roaming to continue, it may well represent an unacceptable limitation on speech communication during that period. Meanwhile, the imposition of antenna switching constraints on GPS signal processing also is significant as integration periods must be sized to fit within these constraints.
GSM Signal Blanking.
For GSM, the problem is less extreme because the technology uses timeslots of only a few milliseconds in width. Signal blanking can be used during these slots, thereby efficiently avoiding the need for cooperation between the software of the two subsystems.
Nevertheless, the blanking itself eats into the GPS integration periods, thereby degrading sensitivity and/or increasing typical acquisition times.
The first limitation of this hardware search engine is the coherent integration period is limited to much less than a navigation message bit. If not, the probability of bit transitions occurring within integration periods will be high, and excessive random losses will result. The effect of this limitation is to ensure the squaring losses prior to the non-coherent integration are relatively large. The end result is relatively long overall integration periods are required to achieve the desired sensitivity.
CDMA Mode Switching.
In the case of a CDMA handset, the main alternative is to switch modes between GPS and the handset. The two subsystems cooperate so the receiver "listens" only during timeslots when the handset is not permitted to transmit.
For example, during the 16-second response period, the transmitter may only be able to transmit for a fraction of the time in bursts of a few hundred milliseconds. While this process allows roaming to continue, it may well represent an unacceptable limitation on speech communication during that period. Meanwhile, the imposition of antenna switching constraints on GPS signal processing also is significant as integration periods must be sized to fit within these constraints.
GSM Signal Blanking.
For GSM, the problem is less extreme because the technology uses timeslots of only a few milliseconds in width. Signal blanking can be used during these slots, thereby efficiently avoiding the need for cooperation between the software of the two subsystems.
Nevertheless, the blanking itself eats into the GPS integration periods, thereby degrading sensitivity and/or increasing typical acquisition times.
Receiver Architectures
In tailoring a GPS receiver for embedded AGPS applications, several factors need to be considered including correlator design, memory requirements, and microprocessor control issues.
AGPS receivers need to acquire weak signals quickly. To meet the demands of the market and exceed the demands of the AGPS standards, more-sophisticated baseband hardware is needed than was required of conventional GPS receivers of only a few years ago. Essentially, this means many more correlator taps or "fingers." How those fingers are organized and the type of signal processing employed are critical factors.
AGPS receivers need to acquire weak signals quickly. To meet the demands of the market and exceed the demands of the AGPS standards, more-sophisticated baseband hardware is needed than was required of conventional GPS receivers of only a few years ago. Essentially, this means many more correlator taps or "fingers." How those fingers are organized and the type of signal processing employed are critical factors.
Limited Coherent Integration.
Figure 1 illustrates a common search engine architecture. This design performs multiple rounds of coherent integration for each of the n fingers per frequency bin, and integrates the results non-coherently. The architecture lends itself to efficient hardware mechanization through reuse of the arithmetic elements. Chips have been produced using this design to incorporate multiple bins and 20,000 fingers or more. Other designs, such as u-Nav Microelectronics' uN8130 baseband chip (with which the author is very familiar), combines a 2,048 finger × four-bin search engine with 12 four-finger correlators.
Figure 1 illustrates a common search engine architecture. This design performs multiple rounds of coherent integration for each of the n fingers per frequency bin, and integrates the results non-coherently. The architecture lends itself to efficient hardware mechanization through reuse of the arithmetic elements. Chips have been produced using this design to incorporate multiple bins and 20,000 fingers or more. Other designs, such as u-Nav Microelectronics' uN8130 baseband chip (with which the author is very familiar), combines a 2,048 finger × four-bin search engine with 12 four-finger correlators.
Figure 1: A two-bin hardware search engine. The first set of mixers mix the final local oscillator signals with the incoming signal to downconvert the selected satellite signal to near baseband. The first set of summing blocks perform coherent integration on the downconverted, despread signal. The squaring blocks compute the magnitudes of the complex integrals resulting from the coherent integrations. The second set of summing blocks perform non-coherent integration on the squarer outputs. A shift register produces multiply delayed versions of the locally generated pseudorandom noise code.
The first limitation of this hardware search engine is the coherent integration period is limited to much less than a navigation message bit. If not, the probability of bit transitions occurring within integration periods will be high, and excessive random losses will result. The effect of this limitation is to ensure the squaring losses prior to the non-coherent integration are relatively large. The end result is relatively long overall integration periods are required to achieve the desired sensitivity.
The second limitation of this approach is that, when it is possible to constrain the search to a small range of code phases, the rest of the fingers are effectively wasted. When precise time assistance is available this means most of the potential of the search engine hardware is wasted all of the time. When only coarse time assistance is available, it means the full potential of the hardware is being utilized for acquiring the first satellite signal — but, again, most of its capacity is wasted when acquiring subsequent satellite signals.
0 comments:
Posting Komentar