SIMULATIONS AND RESULTS
This physical layer was simulated in version 4.5 of the
Signal Processing Worksystem (SPW) from Cadence Design Systems, San Jose,
California, to determine the detector sensitivity (Eb/NO), and susceptibility to
interference from Bluetooth transmissions, which may also be present in the
2.4-GHz band. SPW is described in Appendix A.
3.4.1 Simulations
As illustrated in Exhibit 2, the BER simulation is both simple and
flexible. Both I and Q channels are generated and detected independently, so
that studies of possible differences in performance between them (due to
auto-correlation differences between PN sequences, for example) may be made.
This results in some slowdown of simulations, but because the packet size in
these systems is so small (< 800 bits), there is little need for BER
information below BER = 10−5. This means
that the long simulations needed (to achieve a statistically significant number
of bit errors) for lower BERs are unnecessary, so simulation speed is not of
critical importance.
Exhibit 2: Code Position Modulation Simulation
in SPW
The modulator block (Exhibit 3) produces the desired shifted PN sequence,
based on the incoming data, from the 16 possible choices. After combining the I
and Q channels, the transmitted signal is sent over an AWGN channel block representing receiver noise. This block is a
large file of random noise, used so that repeatable results could be obtained
even when modifications to the system were made.
Exhibit 3: SPW Simulation of CPM
Modulator
The range of a wireless sensor network communication link is
usually limited by design to 10 meters or less, so the one-way flight delay of
the RF signal, even if via an indirect (e.g., reflected) path, is 30 nanoseconds
or less. Because the duration of a chip is one microsecond, the effects of
multipath interference were not considered significant and so were not
simulated.
The demodulator block (Exhibit 4) is a correlator to each of the 16 possible
transmitted sequences on each channel, matched to the half-sine-shaped chips.
Because the purpose of this study was to investigate the performance of the
proposed physical layer during data transmission, instead of the performance of
potential symbol synchronization algorithms, the symbol synchronization was
fixed (i.e., "hard-wired" from transmitter to receiver).
Exhibit 4: SPW Simulation of CPM
Demodulator
To investigate the effects of potential interference from
Bluetooth devices, which will also occupy the 2.4-GHz ISM band, a Bluetoth
interfering scenario was simulated (Exhibit 5).
Exhibit 5: SPW Simulation of Bluetooth
Interference Scenario
The Gaussian minimum shift keying (GMSK) modulator block in
the SPW communication library was modified to reduce its modulation index from
0.5 to 0.315, the nominal Bluetooth value. An adjustable gain block was placed
after the GFSK modulator, to control the ratio of undesired to desired signal
level. The GFSQ modulator was fed with 1-Mb/s random data, and no attempt was
made to offset or randomize the phases of the 1-Mb/s Bluetooth data bits and the
1-Mc/s direct sequence chips. Similarly, no attempt was made to model any
higher-protocol stack levels (e.g., frame structure) of either Bluetooth or the
sensor network, so neither packet error rate (PER) nor message error rate (MER)
was simulated.
3.4.2 Results
The BER versus Eb/NO performance of the proposed
modulation method, as simulated by SPW, is given in Exhibit 6, and is compared with the
theoretical prediction of Equation 5 for k = 1 and k = 4. The SPW Grouping parameter script was run to link the
number of samples simulated with the Ec/NO parameter (and, therefore, the
expected number of bit errors), so that each data point was simulated just long
enough to obtain 100 bit errors. This minimized simulation time while ensuring
statistically valid results. The Eb/NO ratio, required in this
simulation to achieve 1 percent BER using the proposed modulation method, is
approximately 5.2 dB. This is within 1.25 dB of the theoretical value calculated
using Equation 5
with k = 4, and is approximately 4 dB better than that obtained with k = 1 (i.e., binary orthogonal signaling).
Exhibit 6: BER Performance of Code Position
Modulation versus Eb/NO
The 1.25-dB difference between the simulation results and Equation 5 is
largely due to the nonideal autocorrelation and cross-correlation properties of
the PN sequences chosen for the I and Q codes. The correlation between a code
and, itself shifted in time, is close to, but not exactly, zero; similarly, the
correlation between the 45 (octal) code on the I channel and the 75 (octal) code
on the Q channel is not zero. This means that the signals used are not precisely
orthogonal, as assumed by the theory; this trade of sensitivity was made in
exchange for the simple circuit implementation that results.
The effect of Bluetooth interference on the sensor network BER was
also simulated. The scenarios simulated were interference to a relatively weak
sensor network signal, with an Eb/NO
level of 4 dB, a somewhat stronger sensor network signal, with an Eb/NO level of 7 dB, and a very strong
sensor network signal, with an Eb/NO
level of 50 dB. The amount of Bluetooth interference was parameterized and
allowed to vary in amplitude relative to the desired signal level, from - 10 to
+ 15 dB. The results, presented in Exhibit 7, demonstrate clearly the effect of
spread-spectrum processing gain, and the ability of the system to survive,
albeit at an increased BER, even in the presence of interfering Bluetooth
signals stronger than the desired sensor network signal.
Exhibit 7: BER Performance of Code Position
Modulation in the Presence of Bluetooth Interference
Sections
