Random Demodulator
Random Demodulator
1
Article series of the Signals Research Laboratory, February, 2024 2
Random Demodulator
Nimisha T M,† and G. Abhilash‡
1 Introduction
Compressed Sampling (CS) deals with sparse or compressible signals and
N
attains the samples at rates O(Klog( K )), where K is the sparsity of the
signal and N its Nyquist rate (or the dimension of the signal under classical
sampling). Note that in the class, we have identified the minimum number
of measurements (m) required for the correct recovery with high probabil-
ity as 2K, where K is the sparsity of the signal. In practice, we keep the
N
number of measurements (m) more than 2K as O(Klog( K )). These samples
are obtained by projecting the signal to a measurement matrix which sat-
isfies the RIP and incoherence properties. It has been seen that a random
measurement matrix satisfies these conditions and hence is a good choice.
This concept is being made use of in Random Demodulator (RD), which is a
technique to convert analog compressible signals to compressed discrete form
[1]-[4]
The signal of interest, mainly wide-band in nature, is demodulated using
a pseudo random sequence which is generated at a rate equal to or above
Nyquist rate, such that the spectrum of the signal is shifted to the low pass
region. Now, a low pass filtering is done using an integrator and the signal
so obtained is sampled using low rate ADC. The set up of such a sampler is
shown in Figure 1 [4].
∗
This article has been prepared as part of the Signals Research Laboratory activity at
NIT Calicut.
†
Dr. Nimisha was an Algorithm Development Engineer at KLA-TENCOR, Chennai
at the time of preparing this article.
‡
Associate Professor, Department of Electronics and Communication Engineering, NIT
Calicut.
Article series of the Signals Research Laboratory, February, 2024 3
1 1
Figure 1: Single channel Random Demodulator [4]. Note: R
= N
in the text.
The main advantage of such a system is that it bypasses the need for high
rate ADC. But the price paid for this is a highly non-linear reconstruction
algorithm at the receiver. But this is acceptable because almost all acqui-
sition systems have power constraints (e.g. satellites). So our sampler has
to be less power consuming and easy to implement. But receiver complexity
and power consumption can be afforded since they are placed at the earth
stations.
2 RD in matrix form
A sparse signal (sparsity K) with bandwidth N2 Hz can be time-discretized
by sampling at N Hz as stipulated by the classical sampling theorem. But
N
the information content is only M = O(Klog( K )). Thus sampling rate can
be reduced to M . RD maps a continuous signal to discrete samples at a rate
as small as M .
It can be formulated in matrix form for discrete simulation. The first step
is to find a discrete representation of the analog signal space. This is done
Rt +1
by averaging the signal over a time interval of N1 given by xn = tnn N f (t)dt
Article series of the Signals Research Laboratory, February, 2024 4
y = Φx = ΦΨs = As (2)
Once random measurements are obtained, the next question is about the
reconstruction of the original signal from these measurements exploiting the
knowledge of sparsity. This is done, in general, using non-linear optimization
Article series of the Signals Research Laboratory, February, 2024 5
(a) (b)
1 1 −1 0 0 0
Φ=
0 0 0 −1 1 −1
Each column corresponds to each rail in Figure 4. The operation with
Φ is equivalent to keeping an integrator at the output of the adder shown
in Figure 4. The action of Φ on x gives the random measurements at sub-
Nyquist rate.
Article series of the Signals Research Laboratory, February, 2024 7
The results show that the Random Demodulator can efficiently sample
the sparse signals at sub Nyquist rates. A recovery algorithm can recover
the sparse set of representing coefficients from the measurement vector and
reconstruct the original signal from the recovered representing coefficients
with high probability.
(a) (b)
4 Summary
A method for compressively sampling analog signals is presented in this chap-
ter. It makes use of the concept of Random Demodulator which leads to a
Article series of the Signals Research Laboratory, February, 2024 8
References
[1] Jason N. Laska, Sami Kirolos, Marco F. Duarte, Tamer S. Ragheb,
Richard G. Baraniuk, Yehia Massoud, “Theory and Implementation
of an Analog-to-Information Converter using Random Demodulation,”
IEEE Proc ISCAS May 2007.
[3] Tamer Ragheb, Jason N Laska, Hamid Netaji, Sami Kirolos, Richard
G. Baraniuk and Yehia Massoud, “A prototype hardware for Random
Demodulation based conversion Analog-to-Digital Conversion,” Confer-
ence paper, IEEE Midwest Symposium on Circuits and Systems, August
2008.