Comparison of Narrowband and Wideband GNSS Signals Acquisition
Jérôme Leclère, René Jr Landry, LASSENA, École de Technologie Supérieure (ÉTS), Canada; Cyril Botteron, ESPLAB, École Polytechnique Fédérale de Lausanne (EPFL), Switzerland
Among the GNSS signals, those of the L1 bands and the L5 bands are quite different. The L5 band signals have a much higher chipping rate implying a higher sampling frequency. On the other hand, the L1 band signals have longer pseudo-random noise (PRN) codes and more complex modulations, except the GPS L1 C/A signal. Finally, the power level and the power repartition between the data and pilot channels differ according to the signals.
Therefore, if one should choose between acquiring the L1 band signals and the L5 band signals, the choice would not be straightforward because there are advantages and drawbacks on both sides. The goal of this paper is therefore to compare the acquisition implementation of signals of both bands. In particular, the acquisition of the L5 signal is compared to the acquisition of the L1 C/A signal, and the acquisition of the E1 signal is compared to the acquisition of the E5a/b signals.
For GPS, the L5 signal has significant advantages over the L1 C/A signal: a higher power (1.5 dB is significant), which allows reducing the integration time for the same detection performance; a pilot channel, which allows extending the coherent integration time; and a secondary code that makes the navigation data synchronization much easier. On the other hand, the L5 signal has also some drawbacks in terms of computational burden : PRN codes 10 times longer and bandwidth (and consequently sampling frequency) 10 times higher, increasing the number of samples to process by a factor of 10, the acquisition search space by a factor of 10, and the memory to store the correlation results by a factor of 10 also. Therefore, it is interesting to make a clear comparison between both acquisition implementations, to determine which one is more complex in the end, and by which factor. Here, the complexity corresponds to the computational burden, which is characterized by the processing time and the memory requirements.
The case of the L1C signals will also be discussed.
For Galileo, the same applies, however the E1 and E5 signals are closer in terms of complexity than the L1 C/A and L5 signals, which means that the comparison is tighter. Indeed, the E5a and E5b signals are similar to the L5 signal from the acquisition point of view, except that the secondary codes of the pilot channel are longer. Whereas the E1 signal is more complex than the L1 C/A signal, with a PRN code four times longer and a BOC modulation requiring a higher sampling frequency and smaller steps for the search in the code space.
Therefore, in this paper, we investigate the implementations of the acquisition for these four signals to determine the complexity of each one, in the general case and in two specific cases as example, with a focus on hardware implementations. For this investigation, we will consider different types of implementations, all based on the parallel code search: we will consider the different level of parallelization for the correlation with the secondary code for the pilot channels or the accumulation of the correlation results for the data channels, namely serial, semi-parallel and parallel. We also consider the case of coherent integration only and the use of non-coherent integration, and discuss the fact of performing the accumulations of the coherent integration before or after the FFTs computing the primary code correlation.
Two example cases will illustrate the comparison, a cold start with moderately high sensitivity (about –151 dBm) and a hot start with high sensitivity (about –157 dBm).
In these two examples, it is shown that the L5 signal is 12 to 27 times more complex to acquire than the L1 C/A signal. 27 times more complex means that the L5 acquisition can use the same amount of memory as the L1 C/A acquisition but the acquisition time will be 27 times longer; or the processing time can be the same for both acquisitions but the L5 acquisition will require 27 times more memory; or it can be a mix, the L5 acquisition can be about 4 times longer and requires 7 times more memory.