Part II
Front End Technology
Front End design – including RF Architecture, Data Conversion and Digital Front Ends – has
emerged as a key issue as SDR techniques are finding themselves increasingly embodied by
stealth into today’s new products.
The radical solution – ‘Pure’ Software Radio, with A/D conversion at the antenna – is not
yet feasible at GHz carrier frequencies. However, recent technology advances suggest it may
be nearer than had been thought.
Software Defined Radio
Edited by Walter Tuttlebee
Copyright q 2002 John Wiley & Sons, Ltd
ISBNs: 0-470-84318-7 (Hardback); 0-470-84600-3 (Electronic)
2
Radio Frequency Translation for
Software Defined Radios
Mark Beach, John MacLeod, Paul Warr
University of Bristol
In an ideal world, a software defined radio (SDR) would be able to transmit and receive
signals of any frequency, power level, bandwidth, and modulation technique. Current analog
receiver and transmitter hardware sections are still a long way from being able to achieve this
ideal behavior. It is the aim of this chapter to explain why this is the case, to present some
design techniques for synthesis of SDR RF translation architectures, and to consider where
the breakthroughs in technology are required if the RF hardware component of the ideal
SDR
1
is to become a reality.
This chapter is structured in four parts. Initially, we gather data to define the requirements
for the illustrative design of SDR hardware for commercial wireless applications. In the
second part, we attempt to define the problems that are associated with the design of SDR
hardware, both the receiver and the transmitter aspects. In the third, we consider techniques
which may be of value in solving these problems before finally drawing some conclusions.

does not introduce nonlinearities into the circuit performance and an introduction is presented
into the potential role that micro-electro-mechanical system (MEMS) technology could play
in future designs.
The body of the chapter concludes with an examination of the ‘Low IF’ design as a possible
compromise between the superheterodyne and zero IF architectures.
2.1 Requirements and Specifications
There are three driving forces for the development of SDR. The first impetus derives from the
requirement that a mobile phone can provide ‘world roaming.’ This means that the phone, as
well as being able to operate in Europe to the GSM radio standard, should be able to operate
in the United States to their IS54 and IS95 systems, and in Asia and Japan with their PDC and
PHS systems. The second stimulus revolves around trying to combine the performance
features of a radiotelephone (GSM, DECT, and UMTS), with the functionality of a personal
area network (PAN), (e.g. Bluetooth), and that of a local area Network (LAN) (e.g. HIPER-
LAN). The third motivation is that SDR could drive down the production costs through the
scale economies of a single radio platform for multiple standards and hence markets.
Improvements could be made via ‘software upgrades’ and the radio could be ‘future proofed’
to some degree.
In this section we review the radio specifications of some of the major European commu-
nications standards to define the performance that will be required of an SDR capable of
encompassing all these air interface standards. It would be possible to spread the net wider
and also embrace North American and Asian standards; however, the European standards are
sufficiently representative to highlight all the major issues involved. [1–8]
2
2.1.1 Transmitter Specifications
The most important design parameters when dealing with SDR transmitter design are:
X
output power level
X
power control range
X

2.1.2.2 Blocker Specifications
All air interface standards specify blocker signal levels with a mask type of specification.
Again, this is best summarized graphically and is included as Appendix A2 to this chapter.
2.1.3 Operating Frequency Bands
Table 2.3 lists the frequency bands for the air interface standards considered in this chapter;
this information is also shown graphically in Figure 2.1.
Radio Frequency Translation for Software Defined Radios 27
Software Defined Radio: Enabling Technologies28
Table 2.1 Transmitter power output and power control specifications
Air-interface standard Nominal maximum output power Nominal minimum
output power (dBm)
Power control
Terminal class Maximum power (dBm) Levels Power range (dBm) Step
GSM 900 2 39 5 0–239
337 3–15 37–13 2 dB
433 16–18 11–72dB
529 19–31 5
DCS 1800 1 30 0 29 36
224 30–31 34–32 2 dB
336 0–830–14 2 dB
9–13 12 – 42dB
14 2
15–28 0
Nominal output power
Level Power (dBm)
DECT 1 4
224
UMTS-FDD 1 33 2 44 Steps
227 1
324 2

The characteristics of the input signal are:
X
signal type real
X
low power down to –107 dBm
X
high dynamic range up to 2 15 dBm
X
spectrum band pass, with center frequencies varying from
876 MHz to 5725 MHz
Radio Frequency Translation for Software Defined Radios 29
Table 2.2 Input signal level specifications
Air interface standard Reference
sensitivity level
(dBm)
Maximum input level (dBm)
GSM 900 Small MS 2 102 2 15
Other MS 2 104
DCS 1800 Class 1 or Class 2 2 100/ 2 102 2 23
Class 3 2 102
PCS 1900 Normal 2 102 2 23
Other 2 104
DECT 2 86 2 33
UMTS (FDD) 12.2 kbps 2 92
64 kbps 2 99.2
144 kbps 2 102.7
384 kbps 2 107
UMTS (TDD) 2 105
Bluetooth 2 70 2 20
HIPERLAN/2 6 Mbps 2 85 Receiver class 1 ¼ 2 20

Software Defined Radio: Enabling Technologies30
3
Image signals are discussed further under ‘Image Rejection’ within this section on receiver design.
Table 2.3 Frequency of operation of major European air interface standards
Air interface standard Uplink (MHz) Downlink (MHz) Duplex spacing
(MHz)
GSM 900 890–915 935–960 45
E-GSM 900 880–915 925–960 45
R-GSM 900 876–915 921–960 45
DCS 1800 1710–1785 1805–1880 95
PCS 1900 1850–1910 1930–1990 80
DECT 1881.792–1897.344 1881.792–1897.344 Not applicable
– a TDD system
UMTS FDD (Europe) 1920–1980 2110–2170 190
UMTS FDD (CDMA 2000) 1850–1910 1930–1990 80
UMTS TDD 1900–1920 1900–1920 –
(Europe) 2010–2025 2010–2025
UMTS TDD (CDMA 2000) 1850–1910 1850–1910 –
1930–1990 1930–1990
1910–1930 1910–1930
Bluetooth USA, Europe, &
most other countries
2400–2483.5 2400–2483.5 –
Spain 2455–2475 2455–2475 –
France 2446.5–2435 2446.5–2435 –
HIPERLAN\2 5150–5350 5150–5350
5470–5725 5470–5725 –
application of appropriate technological ‘fixes’ such as image reject mixing, linearization,
and variable preselect filters.
Important commercial requirements, which place constraints on this, are:

simple filtering requirements
X
image signal suppression is easier (compared to multiple conversion architecture)
Its disadvantages are:
X
A local oscillator is required, in which the two output signals are accurately in phase
quadrature and amplitude balance, over a frequency range equal to the frequency range
of the input signal.
Software Defined Radio: Enabling Technologies32
Figure 2.2 Direct conversion receiver architecture
X
The mixers needs to be balanced and to be able to operate over a correspondingly wide
frequency band.
X
Local oscillator leakage through the mixer and LNA will be radiated from the antenna
and reflected back into the receiver from that antenna. The reflected signal will vary with
the physical environment in which the antenna is placed. This ‘time varying’ DC offset
caused by ‘self-mixing’ is a problem.
X
Most of the signal gain occurs in one frequency band creating the potential for instabil-
ity.
X
1/f noise is a major problem.
X
Second order distortion product mix down ‘in-band’.
All of these points are explained in more detail later in the chapter.
2.2.2.2 Multiple Conversion Architecture
A multiple conversion receiver is shown in Figure 2.3.
Its advantages are:
X

conversion architecture (see [9,10]). Having a low IF means that the image rejection require-
ments are not as onerous as with the superheterodyne structure, and the fact that the LO signal
is not the same frequency as the wanted signal minimizes the DC offset problems inherent in
the direct conversion architecture.
Its advantages are:
X
the DC Offset problems associated with direct conversion architecture can be overcome
while retaining most of the benefits of this architecture;
X
lower complexity than the superheterodyne approach (but slightly greater than the direct
conversion).
Its disadvantages are:
X
better image rejection is required from a low IF receiver than that required of the direct
conversion receiver.
2.2.3 Dynamic Range Issues and Calculation
In this section we develop equations that are essential for the design of receivers and trans-
mitters for SDR applications.
2.2.3.1 Third-order Distortion Components and Third-Order Intercept
Figure 2.4 shows the relationship of the output power of the fundamental signal component,
and the third-order distortion component, of an RF device, as the input signal power level is
increased. This plot is typical for most nonlinear devices that make up a receiver or trans-
mitter (although, of course, the power levels will be different). Two features may be observed
from this characteristic.
The first feature is that the third-order distortion component increases at three times the rate
at which the fundamental component increases. This is because the power in the third-order
component is proportional to the cube of the input power.
Software Defined Radio: Enabling Technologies34
4
Local oscillator balance and DC offset.

).
6
The ‘two tones’ refers to the two input signals, at frequencies f
1
and f
2
.
7
Other components such as 2f
1
1 f
2
will appear a long way out-of-band, and will thus not pass through the IF
filtering.
With reference to Figure 2.5 again, it can be shown that:
TOI dBmðÞ¼P
1
o
ðdBmÞ 1
AðdBÞ
2
ð1Þ
Equation (1) gives us a convenient way of calculating the TOI given a spectrum analyzer
display of the results of a two-tone test. The measurements required are the power of the two-
tone test signals, P
o
1
(dBm), and the difference between the power of the two test signals, and
the third-order distortion components, A (dB).
2.2.3.2 Cascading Devices with Known TOI and Noise Figure

2 1
G
1
:G
2
:G
3
1
…
ð2Þ
The worst case output TOI of such a cascade of amplifiers is given by
TOI ¼
1
1
TOI
1
G
4
G
3
G
2
1
1
TOI
2
G
4
G
3

reached that will allow a narrow band channel to be selected from the wideband channel.
The worst case situation arises when a narrow band channel, such as GSM, is exposed to
blockers over the wideband UMTS channel. A graphical interpretation of the blocking
specifications for a GSM channel is shown in Figure 2.7.
A situation that will lead to blockers having an effect on the wanted channel is illustrated in
Figure 2.8. Here, two high power blockers are assumed to be present in channels separated
from each other and from the wanted channel by 2.4 MHz
8
. A third-order product will be
produced within the wanted channel due to the blockers being distorted by the implicit
nonlinearity of the receiver hardware. It can be seen from Figure 2.7 that the blockers are
permitted to have an amplitude of up to 2 23 dBm.
Radio Frequency Translation for Software Defined Radios 37
8
2.4 MHz is chosen because it is the maximum separation of blockers from the wanted signal, and from each other,
that can fit inside the 5 MHz UMTS bandwidth (allowing for 200 kHz of the wanted signal). This number is not
critical, as the blockers need to be well outside the GSM frequency band before their allowed power jumps to 0 dBm.
Figure 2.6 Cascade connection of amplifiers
The cochannel interference specification for a GSM system demands that the carrier to
interference (C/I) ratio is at least 9 dB. This implies that, with an input signal of 3 dB above
the reference level ( 2 101 dBm in the case of a GSM mobile station), the cochannel
interference is required to be less than 2 110 dBm.
If the blockers are at the maximum level of 2 23 dBm, then the difference between the two
tones at a power level of 2 23 dBm and the distortion products (at 2 110 dBm), is 87 dB.
This figure can be now substituted into Equation (1) to derive the required input TOI of the
receiver as
Software Defined Radio: Enabling Technologies38
Figure 2.7 GSM blocker specifications
Figure 2.8 Scenario of blockers producing in-band third-order products
TOI

TOI
LNA
G
IF
1
1
TOI
IF
¼
1
1
11:22 £ 10
4
1
1
100
¼ 99:9W< 50 dBm
Radio Frequency Translation for Software Defined Radios 39
Figure 2.9 Mitigation of the effects of nearby blockers by using a channelization filter high up in the
RF chain
where TOI
LNA
is the TOI of the LNA (W), TOI
IF
is the TOI of the IF amplifier (W) and G
IF
is
the power gain of the LNA (linear ratio). This yields a distortion component of 1 11 dBm
and a SINAD of 2 50 dB, which would be unworkable. Note the TOI of the LNA is bigger
than it needs to be for the signal path on the left-hand side of Figure 2.9

2
3
TOI
in
22174 1 10log B
w
ÀÁ
1 NF
ÀÁÀÁ
ð6Þ
where TOI
in
is the input TOI of the receiver. Equation (6) essentially restates Equation (5).
2.2.4 Adjacent Channel Power Ratio (ACPR) and Noise Power Ratio (NPR)
The broadband nature of the signals used in modern radio systems, combined with the close
spacing of the channels, has produced important changes in the way of characterizing distor-
tion. The TOI figure of merit is often replaced, or at least augmented, by parameters that
employ measurement techniques more directly related to the system that the engineer is
trying to characterize.
The adjacent channel power ratio (ACPR) is one such parameter. This parameter measures
the effect of a signal from one channel appearing in the adjacent channel. ACPR is the ratio of
the average power in the adjacent channel to the average power in the desired channel. Figure
2.10 shows how measurement of ACPR is calculated. This is conveniently done using a
spectrum analyzer. P
DC
and P
AC
are measured by integrating the respective desired channel
and adjacent channel powers over the channel bandwidth.
Software Defined Radio: Enabling Technologies40

Such a signal budget diagram for a hypothetical receiver is shown in Figure 2.11. (For
clarity, the maximum input signal and its associated automatic gain control (AGC) charac-
teristics have been omitted from the diagram.)
Radio Frequency Translation for Software Defined Radios 41
Figure 2.10 Different ways of quantifying the IMD distortion for wideband modulated or multi-
channel signals
A 100 kHz signal bandwidth has been assumed for this receiver, thus the effective thermal
noise input is 2174 1 10logð10
5
Þ dBm. The signal level increases by an amount equal to
the gain, or decreases by an amount equal to the loss, as the signal progresses down the
receiver chain. The noise level increases or decreases by an amount equal to the gain or loss
plus an amount given by the progressive noise figure (see Equation (2)). The difference
between the input signal-to-noise ratio and the output signal-to-noise ratio gives the overall
receiver noise figure (4.62 dB in this example).
An ADC will have a noise floor set by the quantization noise of the converter itself.
Quantization noise occurs because the converter output only ever approximates the analog
signal that it is converting. It can be shown that the signal-to-noise ratio of an ADC is given
by
SNR
QF
dBðÞ¼6:02b 1 1:76 1 10log
F
S
2B
C

dB ð7Þ
Software Defined Radio: Enabling Technologies42
Figure 2.11 Signal and noise levels throughout a hypothetical receiver chain

0
for the modulation scheme used, i.e. AGC range ¼ [(P
in(max)
2
P
in(min))
2 (P
ADC(min)
2 n
ADC
) 1 E
b
/N
0
] dB.
†
The noise floor at the output of the receiver can be determined by either the noise floor of
the ADC or the thermal noise floor. For narrow band systems it tends to be the noise floor
of the ADC which determines the receiver noise floor. For wide band systems, it is the
thermal noise floor (in which case the ADC has more resolution than required).
2.2.5.1 An Approach to Receiver Design
To make a start on the design of the receiver, we need to know certain parameters. From
knowledge of these parameters, the design calculations (also listed) can be made.
1. The maximum signal input level to the ADC. This establishes point A in Figure 2.12.
2. The maximum blocker level likely to be present in the signal at the ADC expressed in dB
relative to the minimum signal level (usually the reference signal level plus 3 dB). This
establishes the separation of point A from point B and hence establishes point B in Figure
2.12. The required net gain of the receiver channel can now be calculated.
3. The minimum signal-to-noise ratio for the narrowest band air interface standard being
used. This will establish the separation of B and C in Figure 2.12 and hence establish point

A problem that arises uniquely with software radio RF design is how to accommodate image
signals. An image signal is a signal of such a frequency that, along with the wanted signal, it
Software Defined Radio: Enabling Technologies44
Figure 2.12 Receiver signal levels
will be mixed down to the set IF frequency. Image signals differ in frequency from the wanted
signal by twice the IF frequency.
10
Image signals are removed in a conventional receiver via the use of preselect filters. The
receiver in a GSM mobile station, for instance, should be set to receive signals in the range of
935–960 MHz. The preselect filter could thus have a bandwidth of 25 MHz based on a center
frequency of 947.5 MHz ((935 1 960)/2). Assuming an IF frequency of 80 MHz, the image
frequency bands would be 1095–120 MHz (for high side mixing) and 775–800 MHz (for low
side mixing). Both of these bands would be eliminated by any preselect filter of modest
performance.
Such a comparatively narrow band image filter is not a simple option with a proper SDR
receiver as the frequency band of signals it will be set to receive should be software
determined, although switchable or electronically tuneable preselect filters are one possible
solution.
Radio Frequency Translation for Software Defined Radios 45
Figure 2.13 Received signal levels for W-CDMA
10
Twice the IF frequency higher than the IF frequency, in the case of high side mixing, and twice the IF frequency
lower than the IF frequency, in the case of low side mixing.
Image reject mixing is another way of dealing with image signals. It is typically used with
low IF, or zero IF, receivers.
11
To get an image reject mixer to function satisfactorily, the two
local oscillator signals need to be in precise phase quadrature, and have precise amplitude
balance. Zero IF and low IF receivers can get away with relatively poor image rejection
performance. Image rejection performance of the order of 40 dB is satisfactory for a direct

conversion, the image signal will be at 540 MHz. If 80 dB image rejection is required, then
the first IF filter must be 80 dB down at 540 Hz. This places quite a restriction on the first IF
filter. These specifications can be met with an SAW filter. However, the effect is described
because it is all too easy to assume that because the image signals have been removed from
the first down-conversion, then subsequent down-conversions will not produce image
problems. The effect could be summarized by saying that the filter prior to any down-
conversion acts to provide an RF preselection function to remove image signals and therefore
this filter should be designed accordingly.
2.2.7 Filter Functions within the Receiver
To summarize the points raised in this section, in any superheterodyne receiver architecture,
filters are required to perform three functions.
X
First, they band limit the signal to the frequency of interest. This function is often referred
to as ‘channelization’ and is achieved, for preference, in the baseband of the receiver.
X
Second, filters are used to allow the image signal to be separated from the wanted signal.
This function is performed at the first opportunity in the receiver chain.
X
Third, filters should prevent nearby but out-of-band ‘blocker’ signals generating suffi-
cient ‘in-band’ power to interfere with the wanted signal. It should be noted that if the
receiver amplifier were perfectly linear, then it would not be possible for out-of-band
signals to generate in-band products, and a filter to achieve this function would not be
required. In practice, some nonlinearity exists in all amplifiers and mixers that make up
the receiver chain. This means that some degree of channelization needs to occur at a
fairly early stage in the amplifier-mixer chain.
2.3 Transmitter Design Considerations
The design of the transmitter is somewhat similar to the design of the receiver in that there are
elements in the receiver design which appear, slightly disguised in format, within the design
of the transmitter. We first discuss these features before moving on to issues more particularly
related to transmitter design.