27-11-2012, 11:55 AM
Efficient Implementation of SSB demodulation, using multirate signal processing
Efficient Implementation.pdf (Size: 1.81 MB / Downloads: 51)
Introduction
The main goal for introducing Software Defined Radio (SDR) is the reduction of the analog hardware
in the system. Software radios define an emerging technology, thought to provide a flexible
radio system, reconfigurable and reprogrammable by software [1]. The advantages of SDR can be
stated as follows.
• Flexibility: Changing the functionallity in the receiver can easily be done by a software updating,
instead of replacing or modifing existing physical hardware like analog electronics.
• Cost: The total cost are together with power one of the biggest technical issues facing developers
of SDR.
• Time-to-market: The prototype development of a SDR can be done with implementation of
FGPA’s, which makes the time-to-marked fast.
• Size: Analog components will be replaced by digital hardware, i.e. the physical size of the
receiver will be reduced.
• ADC: The ADC must have a high dynamic range and sampling rate, this are limiting factors
that determining the maximum achievable data rate of the receiver.
In software radio it is desirable to move the analog-to-digital conversion as close to the antenna
as possible, as in the ideal software radio [6]. This is shown in figure 1, but is unfortunately not
possible yet due to limitations in todays hardware technology [4].
Problem description
In the preceding, the implementation of software radio into an existing system has outlined that
there is a need for investigation. Since the tasks of transforming a part of the receiver from the
analog domain into the digital hardware is not a trivial task, there is a need for analysing these
tasks. These are two coherent tasks, the first is the mapping of the analog circuitry into the digital
domain, and the second is the design of a system architecture where the algorithm can be implemented.
Since this work is an existent system, the function basis is a priori information from the analog
domain. This implies that there is a need for functions to be synthesized into the digital domain.
Knowing about these existing approaches, considerations can be done in order to optimize the
algorithms instead of developing new ones.
The receiver should perform a real-time SSB demodulation using the traditional heterodyne
approach, and still be able to keep the specification within the limitations set by ETSI. A software
radio for a Super-Heterodyne receiver should be able to handle a large dynamic range, for which
reason an AGC is needed. Since the aim of this work is to consider a receiver design with the
ADC placed after the second mixer, the second IF must be chosen carefully. Before the final
demodulation re-produces the baseband output signal, the modulated signal is filtered to make the
last selection in the chain of the receivers. This is done in several steps through the chain to avoid
adjacent channel, and image response.
Design considerations
This section will state the design considerations as an intro to the algorithm chapter. We have the
possibility to lower the second intermediate frequency, from 455 kHz to a new wanted frequency.
There are mainly two things to be aware of when lowering this frequency, the bandwidth of the
spectrum which is assumed to be band limited (attenuated by 70 dB) to ±40 kHz with fc as
the center frequency of the spectrum. The other important consideration is the down-sampling
factor which is wanted, to be an integer multiplum of the samplings frequency. In the design
considerations, we consider the block diagram shown in figure 3.
Hilbert Transform
The Hilbert transform is applied to the sampled incomming signal, by using a two path half-band
filter that we modulate to the quarter-sample rate, as shown in figure
When applying the Hilbert
transform we zero out all negative frequency. Since the half of coefficients in a half band filter are
zero, we can cost free utilize the inherent properties of down-sample by a factor of two by skipping
the zero multiplications. This results in a throughput of four times the intermediate frequency.
FIR Polyphase decomposition
We can take advantage of decomposition the FIR filter, when using down-sampling, by utilizing
the properties of the noble identity. Before decomposition the filter, we need to known the downsampling
factor M, since the output rate need to be 12 kHZ, we use equation 5 to calculate M.
Two-path recursive All-pass half-band filter
The two-path recursive all-pass filter is able to down-sample by a factor of two in each section [2].
This will cause the down-samplings factor to be a power of two. When down-sampling by a factor
of two in each section, the down-sampling factor need to be a power of two as shown in equation 6.
We know that the spectrum is band limited within ±40 kHz. The output at audio has as sample
frequency of 12 kHz, which corresponds to that the second intermediate frequency is not allowed
to be lower than 96 kHz, and five section of down-sampling.