DSP Software
Radio Techniques:
by Andrew Bateman, Carole Mayhew and Graham Carter - DSP e-Newsletter Staff
Summary:
Delivering high-speed wireless connection to the Internet for business and
residential users is a massive business opportunity and technical challenge.
Governments around the globe are proactively allocating
new blocks of spectrum specifically to facilitate this service, trying to
ensure that their respective economies benefit from full exploitation of the
Internet potential.
The design challenge for the wireless terminal is to
achieve robust communication at the desired data rate, with good spectral
efficiency, (in order to maximise data rate or number of users for a given
frequency allocation), the minimum of set-up and maintenance overhead, at a
competitive price. From a manufacturing standpoint, it is also essential that
the wireless platform can be easily tailored to new frequency bands as they
become available, and be able to exploit and/or manage the characteristics of
the channel and variable data transfer demands in a flexible manner.
This article looks at using a software radio approach to
wireless Internet terminal design, harnessing the ever increasing MIPS/Watt
potential of modern DSP devices. The result is a highly configurable, low cost
solution with minimal frequency selective components, high efficiency and small
size. Practical architectures, and algorithms are highlighted.
Figure 1.1:
Point to Multipoint WLAN
architecture
Distributing the ultra high capacity of a fibre node
(multi Gigabit) to disparate users cost effectively often requires a radio or
free-space optical transmission link. Data rates in excess of 10Mbps are called
for, with upwards of 100Mbps a target for larger business users. Distribution
from a fibre hub will need to be via a point to multipoint network, Figure 1.1:, or a
distributed network.
The design challenge for the wireless terminal is to
achieve robust communication at these high data rates, with good spectral
efficiency, (in order to maximise the number of customers that can be served
for a given frequency allocation). Of course, minimal set-up time, low
maintenance and a competitive price are assumed. From a manufacturing
standpoint, it is also essential that the wireless platform can be easily
tailored to new frequency bands as they become available, and be able to
exploit and/or manage the characteristics of the channel, (interference,
multi-path, delay spread, etc) in a flexible manner.
Unlike the cellular phone market, which is tightly
controlled by standards (GSM/IS95) and 3Gpp (UTRAN), the WLAN market has no
dominant air-interface standard and literally dozens of proprietary system in
play. On the one hand, this offers the flexibility for designers to work with
the latest modulation, coding, access and equalisation formats, thus squeezing
every last drop of capacity out of the channel. On the other hand, the
manufacturer is forced to build in a degree of flexibility in his design (or
develop multiple products), in order to allow the product to operate on
multiple air-interface standards across a range of service providers and
network types.
Achieving flexibility in a wireless platform requires
three key components:
·
Minimal frequency selective (specific) components
·
High linearity to minimise self induced signal
distortion
·
Maximising the software defined functionality of
the unit.
Figure 1.2:
Core Digital IF Transceiver
Architecture
The core building blocks of a modern software defined
digital radio are shown in Figure 1.2, with the DSP elements highlighted. For maximum software configurability, the
point at which conversion between the Analogue and Digital world takes place
should ideally be at the RF carrier frequency. Unfortunately, this is not
currently viable at frequencies much above 500MHz , and with most WLAN spectrum
allocations in the 2GHz - 5GHz plus range, an analogue intermediate frequency stage is
still required. The actual digitisation
frequency to be used will be governed by:-
·
Component cost
·
Linearity, speed and dynamic
range of D/A and A/D converter technology
·
Power consumption
constraints
Linearity and speed of converters has improved
significantly in the last 12 months, with 14 bit, 200MSPS, 80dB SNR converters
now readily available. These can support 70MHz IF sub-systems with ease using
direct sampling, and will extend towards 600MHz IF solutions using sub-sampling
methods.
(For more information on sub-sampling, check
out the applications section of DSPStore – www.dspstore.com)
Power consumption of the high speed converters is
significant, (several 100’s mW), and further reduction is required in IC
feature size and/or fabrication technology to realise the power savings needed
to allow IF sampling in hand portable equipment. For fixed WLAN installations
however power consumption is less critical and linearity/sampling rate are
likely to dominate the choice of IF frequency.
With the digital IF architecture giving maximum design
flexibility for a WLAN solution, we now take a closer look at some of the
signal processing blocks involved.
The DSP tasks for a high speed wireless internet modem can
be categorised into core modem functionality (source/channel coding, pulse
shaping, modulation, demodulation), and the more advanced software radio
management tasks such as quadrature frequency translation, PA linearisation, receiver
dynamic range extension, I/Q calibration, automatic power and frequency control,
direct digital synthesis.
Unfortunately there is not time to cover all of these
software functions in detail here, and we will be focusing on the highlighted
tasks for the remainder of this article.
Further information on a wide range of
software radio algorithms can be found in the DSP Handbook [1], Digital
Communications – Design for the Real World [2] and DSPStoreTM (www.dspstore.com)
[3].
Pulse Shaping:
To maximise the data transmission rate over a wireless
link with finite bandwidth, it is essential to shape the data pulses modulating
the carrier signal. For FSK based modems, this has traditionally involved
Gaussian filtering whilst for the more advanced QAM modems, the Root Raised
Cosine (RRC) filter is commonly employed. Whatever pulse shaping is used, there are two core algorithms for
implementation – the classical filter, and the look-up table method.
The filter approach is normally realised using an IIR
(Infinite Impulse Response) filter for a Gaussian filter shape and an FIR
(Finite Impulse Response) filter for the RRC shape. Focussing on the FIR filter
solution, the RRC filter shape is always an approximation of the true Nyquist
filter response, trading off filter length (and hence delay and processor load)
for stop-band attenuation and roll-off rate. Most filter design packages now
offer RC & RRC filter options, making it easy to explore the trade-off between these parameters. In many cases, it may be preferable to
cascade an RRC filter with a second FIR filter (possibly half-band for ease of
implementation). The second filter achieves the desired level of out of band
attenuation to meet a given spectral mask, whilst relaxing the requirements on
the first, (and more processor intensive), RRC filter.
The alternative approach to pulse shaping makes use of a
look up table, Figure 1.3. This table holds the pre-calculated values for the
pulse response of the desired filter, based on all possible input state
transitions, (e.g. 0®0, 0®1, 1®0, 1®1 for
a binary input).
Figure 1.3:
Look-up
Table Based Data Pulse Shaping
The state transition is used to index the correct stored
pulse response from the chosen filter which is then summed with the pulse
responses from previous transitions to form the composite pulse shaped
waveform.
The lookup-table method is preferred where execution time
is at a premium, as look-up table indexing carries very low overhead.
Conversely, the real time filter realisation of pulse shaping is used when memory
space is at a premium and storage of the multiple filter pulse responses is
impractical.
Data Demodulation:
Efficient algorithms for data demodulation are key to the
success of software defined wireless modems. The topic is of course an
extremely large one, with multiple modulation formats, coding strategies,
carrier and symbol timing recovery mechanisms, equalisation methods, etc in
widespread use.
We shall look at just one demodulation algorithm for
frequency discrimination here which finds widespread application and is a good
example of a DSP optimised solution. Frequency discrimination has two primary
uses – demodulation of the FSK family of waveforms, such as GMSK used in GSM
cellular, and in automatic frequency control loops.
Fig 1.4 shows the block diagram of the DSP frequency
discriminator algorithm.
Figure 1.4:
Block
Diagram of DSP Frequency Discriminator Algorithm
With general complex input signals of the form
I(t) =
r(t) · sin θ(t)
and
Q(t) =
r(t) · cos θ(t)
where r(t) is the signal envelope and θ(t) the angular phase/frequency, the signals at the outputs of the two
differentiators can be represented as,
By cross multiplying and subtracting these signals as
shown, an output signal is obtained, given by:
Further division by the (envelope)2 term yields
a normalised real time measure of the instantaneous frequency variations of the
input signal. (In practice, it is much more efficient to use a look up table to
generate 1/r2(t) which is multiplied with the top path signal). A major attraction of this algorithm is that
it does not involve any feedback process (as with conventional PLL based
frequency discriminators). It also
introduces little or no bandwidth expansion into the signal, thus ensuring that
the Nyquist sample rate limit is not violated.
PA Linearisation:
All of the effort put into clever pulse shaping and
multi-symbol modulation is wasted if the waveform is overly distorted after
passing through the analogue TX/RX functions. The usual culprit for introducing distortion is the RF Power Amplifier. With designers trying to maximise
PA efficiency and power output, linearity is often sacrificed. The conventional solution of backing off the PA drive to operate within a
linear portion of the PA characteristic is very wasteful of power. A much more efficient solution is to use DSP
techniques to pre-distort the waveform of the signal driving the PA, in a complementary
manner to the PA non-linearity, Figure 1.5.
Figure 1.5:
DSP
Based Adaptive Base-band Pre-Distortion of Amplifiers
The source waveform is passed through a look-up table,
which stores the correction factor for the amplitude (and phase) of the
waveform at any given PA drive level. For most applications, the PA
characteristic is not sufficiently stable with temperature, supply voltage,
output load, etc for this process to be purely open loop. A feedback path from the PA output is thus
usually employed to provide a means of measuring the residual PA distortion and
updating the lookup table coefficients accordingly.
For the high bit rate solutions required in WLAN and the
LUT specific nature of the processing tasks involved, implementation of
adaptive pre-distortion is often best performed using dedicated ASIC or FPGA
solutions.
For more information on adaptive pre-distortion and other
amplifier linearisation techniques, see [2, 4].
Software Radio – Hardware Issues
A/D & D/A Conversion:
A range of manufacturers, including Texas Instruments,
Analog Devices, National Semiconductor, Fujitsu, and Intersil, are competing
aggressively in the area of high speed digital IF converter solutions and new
devices are appearing on the market almost weekly.
An up to date list of these devices can be found in DSPStoreTM
Digital Up/Down Conversion:
The initial task for the digital processing unit is to
convert the IF signal to a complex base-band form (down-conversion) and from
complex base-band to IF (up-conversion). This ensures minimum sampling rate
processing for the remaining radio functions. These tasks involve mixing
(multiplication) with quadrature versions of a digital oscillator. In addition,
a process of interpolation and decimation is needed in the up-converter and
down-converter respectively to optimise the sampling rate between the digital
IF requirements and the complex base-band requirements, Figure 1.6:.
Figure 1.6:
Sample Rate Conversion
Because these two functions, (frequency mixing and sample
rate conversion), are common to all digital IF solutions, a number of
manufacturers have produced dedicated IC’s optimised for this task. The
advantage of this approach is that the high sample rate processing associated
with the digital to analogue interface can be accommodated in these hardwired
devices, allowing the slower (comparatively) digital signal processing of the
wireless signal content to be undertaken in cheaper, lower power, software
programmable DSP devices.
An alternative route, maintaining full flexibility of
design, is to implement the mixing and sample rate conversion processing in
software, using high speed DSP or programmable gate array blocks. Companies
such as Xilinx and Altera now offer FPGA devices suited to this task, and TI,
Analog Devices, etc offer DSP units with sufficient processing speed.
Additionally a range of third party suppliers are providing custom algorithms
or development tools (such as the Celoxica Handel-C compiler for FPGA’s).
Digital Signal Processing Engines:
Once in complex base-band form, the option again exists to
process the signals using a dedicated ASIC (very limited flexibility), an
off-the-shelf DSP solution (or DSP core), or FPGA device(s). Again, the maximum
flexibility sought in our software programmable WLAN solution is achieved using
the DSP/FPGA option. A growing range of high speed DSP devices now exist which
can handle data rates of several Mbps, including the TI C5X range of ultra low
power devices for portable use, and the TI C6X range for very high speed
applications
Concluding Remarks:
This is a very exciting time for the wireless modem design
engineer. There is unprecedented flexibility in the range of wireless functions
that can now be implemented or influenced in the digital domain, and new
algorithms are being devised to exploit this advantage to the full. Although
there is limited opportunity here to do justice to this exciting area, a wealth
of information is available on manufacturers web sites and has been bought
together under the DSPStoreTM [3] umbrella. Many of the basic algorithms
underpinning software radio solutions are contained in the new DSP Handbook [1]
References:
[1] The DSP Handbook, Bateman & Stephens, Prentice
Hall, ISBN: 0-201-39851-6
[2] Digital Communications – Design for the Real World,
Bateman A, Addison Wesley, ISBN 0-201-34301-0
[3] DSPStoreTM http://www.dspstore.com
[4] High Linearity
RF Amplifier Design, Peter B.
Kenington, Artech House;
ISBN: 1580531431
