INTERNATIONAL TELECOMMUNICATION UNION
XX Yyyyy 199Z
Original: English only
Received: XX Yyyyyy 199Z
THE REDUCTION OF SPURIOUS AND UNWANTED EMISSIONS
This paper considers practical hardware and system measures to reduce interference from unwanted
A brief description of the mechanisms giving rise to unwanted emissions is followed by a description
of hardware measures that may help to reduce interference from unwanted emissions. Further work is
2. Unwanted Emissions - an overview
Unwanted emissions from transmitters generally arise in one of three main ways :-
(i) The desired modulation may not be sufficiently well filtered to constrain the output
spectrum to within the nominal channel bandwidth.
(ii) The linearity of the output power amplifier may be poor, which results in spectral
regrowth of the signal into adjacent channels.
(iii) There may be discrete mixing products generated by frequency translation stages within
In this paper, generic techniques for reducing the level of such unwanted emissions are reviewed.
2.2.1. Modulation sidebands
When ‘out-of-band emissions’ are referred to, it is most frequently the further sidebands resulting
from the modulation process that are intended. In the early days of telecommunications, when the use
of amplitude modulation was dominant, the modulating waveform was typically narrowband and
readily limited by simple filtering. The emergence of FM brought with it the possibility of sidebands
extending from the carrier frequency by many times the modulating frequency. The use of wideband
analogue modulation compounded this potential problem in systems such as FDM/FM and
In recent decades, the growth in the use of digital modulation (including systems, such as many types
of radar, using pulse techniques) has led to an increasing incidence of interference entries due to
excessive sideband power. If unfiltered, a digital system in which a carrier is modulated with
rectangular pulses may give rise to infinite sidebands following a law, as illustrated in Figure
Relative power density (dB)
P( f )
10 5 0 5 10
Norm alised frequency offset (df/R)
Figure 2.1: Power spectrum of unfiltered digital signal
As a result, the use of effective filtering is of prime importance, but it should be noted that the most
efficient way to achieve this is in the design of the modulating process, rather than by ‘brute-force’
techniques at the transmitter output.
It should also be noted that, following filtering of the transmitted signal (at IF, RF or by baseband
conditioning), sidebands may be re-generated owing to non-linear effects in the high-power
amplifier. These effects are discussed below.
2.2.2. Intermodulation products
When a signal consisting of more than one spectral component is passed through a non linear circuit
element, intermodulation products (IPs) will be generated. These products are due to the
multiplication of the signal components (equivalent to convolution in the frequency domain).
The ideal amplifier will exhibit a linear power transfer characteristic, with an invariant phase shift.
Plainly such an amplifier is not practically possible, and in the real world will exhibit AM-AM
distortion (non-linearity of power transfer characteristic) and AM-PM distortion (variation of phase-
shift with input power). Both these effects will become more severe near saturation, and the
amplifier will therefore require to be backed-off to operate at a suitably linear point in its
characteristic. If this is not done, the transmitted waveform will be distorted, causing increased inter-
symbol interference (ISI) and bit-error rate (BER) in a digital system, and baseband distortions in an
In the simple case of an amplifier, through which two signals at frequencies f1 and f2 are passed, IPs
may be formed at frequencies of the form: f1 - (2 x f2), (2 x f1) - f2, etc. As the generation of IPs is a
non-linear effect, their amplitude changes exponentially with input power.
Intermodulation products may be reduced (or effectively eliminated) by ensuring that all components
in the signal path operate with a linear characteristic, and this is discussed in a later section. In
addition, the majority of IPs will lie at frequencies far removed from the desired signal, and may
therefore be filtered readily.
In addition to the problem of linearising active circuits, passive system elements (such as combiners,
feeders and antennas) may introduce IPs due to the unintentional presence of non-linear electrical
junctions in these structures. Such effects must be eliminated by careful design.
A problem that may be largely outside the control of the equipment operator is that of externally
generated passive IPs (sometimes known by the unfortunate acronym ‘PIMPs’ , for Passive
InterModulation Products). This effect arises when high field strengths from a radio source
illuminate a non-linear element nearby, which can absorb and re-radiate energy. This is often referred
to as the ‘rusty bolt’ effect, as it commonly occurs in steel masts or structures near broadcast or other
high-power transmitter sites. Rust may be a problem, even in space systems, and in any case, great
care must be exercised in the choice of component materials that will be exposed to RF fields, if
PIMPs are to be avoided.
2.2.3. Mixer products
Traditionally, a final transmitted signal is derived from the mixing and multiplication within the
transmitter of signals at a variety of frequencies. This approach allows critical or complex circuitry
such as oscillators, filters and modulators to operate at low frequencies where stable or efficient
operation is more readily achieved. In such systems the suppression of unwanted output components
is dependent on adequate filtering and careful circuit design (to achieve adequate mixer balance, for
The choice of such internal frequencies is likely to be made, at least in part, to allow straightforward
filtering. It will, perhaps, seldom be the case that the system designer will take into account the
allocations made to the passive services in the Radio Regulations, except where there is a history of
With technological improvements (such as the development of MMIC devices - see section 2.3.3) the
performance of circuits at higher frequencies has become less of a constraint, and it is possible to
reduce the need for complex frequency translation, or to remove it altogether, in techniques such as
vector modulation (see section 3.1.1)
2.3. Related issues
2.3.1. Phased-array antenna systems
Much of the development of phased array antenna technology has roots in military radar systems,
where the possibility of implementing inertialess scanning antennas has obvious appeal. The
technology has since been developed for communications use, particularly in the satellite sphere,
where the ability to synthesise irregular beam shapes is valuable, as is the possibility of
implementing dynamically adaptive and steered beams. Further advantages include the possibility of
replacing a single high power thermionic amplifier, with elemental solid-state devices distributed
across the array, and the use of an array of active attenuators and phase-shift elements operating at
low-power in place of a heavy and mechanically-complex waveguide beam forming network.
However, the use of phased array antennas when unwanted emissions are present in the output of a
transmitter system, can lead to severe sharing problems, as unintended antenna beams may be
synthesised in random directions. The output from a single exciter stage will be fed to each antenna
element with the appropriate value of phase-shift to synthesise the beam required at that frequency. If
unwanted emissions at random frequencies are present in this output, these will also be fed to the
antenna elements, with random phase-shifts, thereby synthesising unwanted beams at the frequency
of the unwanted emissions.
It is possible that, to a terrestrial user at the boresight of one of these unintentionally synthesised
beams, the unwanted emissions will appear at a higher PFD than the intentional satellite
Consideration must therefore be given to these effects at an early stage in the design of such systems.
2.3.2. Adjacent-band problems
The Radio Regulations prohibit emissions in certain bands, specified in S5.340. Radio astronomical
observations typically centre on these bands, though they will often, if spectral occupancy allows,
extend beyond them. A growing problem for the radio astronomers is posed by the increasing levels
of unwanted emissions falling within these bands; the subject of the bulk of this paper.
It should also be noted, however, that a considerable burden is placed on the Radio Astronomy
Service (RAS) by the need to accommodate the use of adjacent bands by active services. While most
filtering in an RA receiver is performed at IF frequencies, it is increasingly necessary to provide
filtering at the receiver front-end, to prevent powerful, adjacent band transmissions from driving the
receiver into a non-linear condition. Such filtering will, necessarily, imply insertion losses which will
increase the noise temperature of the receiver system.
An example of this problem is given by the recent commissioning, as part of a new UK television
network, of a transmitter in the English midlands, operating on UHF television channel 37 (598-
606 MHz). Interference from this transmitter is sufficiently severe at Jodrell Bank (some 70 km
distant) to require the addition of filtering to the front-end of the receivers designed to operate in the
610 MHz band (see section 3.1.2).
Such filtering is inevitably necessary, where active and passive services share spectrum, and is a
burden accepted by the passive services. In making frequency assignments, however, the problems of
implementing such filtering should be borne in mind. It is clearly feasible to achieve adequate
rejection of adjacent-band energy at UHF, where the percentage frequency difference is large, and it
is reasonable that the passive service should accept this burden. However, it is possible to imagine
cases where frequency assignments were made without consideration of technological practicability,
that would, in practice, preclude the operation of passive services.
The current concern regarding the allocation of ‘cloud profiling radar’ (CPR) systems illustrates this
point. These systems are carried on spacecraft and use radar techniques to determine the structure of
cloud formations. It is desirable to operate such systems in the 95 GHz range, adjacent to several RA
allocations. The devices used in RA receiver front-ends at this frequency (“Superconductor-insulator-
superconductor” or SIS) are extremely prone to damage by modest incident powers, and it is likely
that an RA receiver would be destroyed by the passage of a CPR satellite near the main-beam of a
radio-telescope. Filtering would therefore need to be implemented at the telescope, but it is currently
uncertain if this is possible ( a current study by ESA, in conjunction with Oerlikon-Contraves SpA is
addressing this question).
The potential exists, therefore, for a new allocation effectively to prevent the operation of an existing
service. There is, perhaps a subconscious tendency, within the ITU-R, to regard the absolute band-
edges in the Radio Regulations as being capable of duplication in the real-world by a finite amount
of filtering. Even if it is recognised that this is not the case, the onus is felt to be on the operators of
the receiver to provide sufficient rejection of adjacent-band power, whether or not this is physically
possible. The implication is that the ITU-R must explicitly consider available filtering technology,
when making frequency allocations, to prevent imposing unrealisable (rather than merely onerous)
demands on receiver designers.
2.3.3. Technological issues
Active circuits for use in the 1 - 100 GHz range are increasingly fabricated as Monolithic Microwave
Integrated Circuits (MMICs). In such circuits, active and passive components (such as spiral or
microstrip inductors) are fabricated on the same semiconductor substrate. The term ‘monolithic’
distinguishes such circuits from ‘hybrid’ circuits, where passive components are fabricated
separately, and later bonded to the substrate. The technology reached practical maturity in the 1980’s,
with the development of a wide range of transistor circuitry
A key benefit of MMIC technology is that of reproducibility and low cost for large volume
production. The technology also imposes restraints that may demand the use of sub-optimum devices
or circuit design, and as a consequence the technology may not be an appropriate choice for low-
volume production of critical components such as low-noise or high-power amplifiers. However,
parasitics arising in hybrid circuits from items such as chip components and bond-wires can be
mitigated in MMIC design to a very large extent.
A field in which the MMIC finds an important application is in the implementation of active phased-
array antennas. The ability to replace a thermionic high-power amplifier, and a complex, waveguide-
based beam-forming-network with a network of phase-shifters, attenuators and distributed
amplification allows great reductions in manufacturing complexity and a great increase in operational
flexibility. There are however, significant potential dangers in the use of this technology, from the
point of view of unwanted emissions (see 2.3.1 above).
A typical feature of MMIC devices at frequencies below around 20 GHz is the extensive use of
lumped elements - a drawback of this technique is the limited Q-factor that can be achieved by such
technology. Recent GSM / PCN handsets operating around 2 GHz have, however, used distributed
elements and achieved better performance, especially using the new technology of multilayer
Improvements in active device technology also seem to offer the possibility of achieving reductions
in the levels of unwanted emissions. The linearity achieved by HEMT (High Electron Mobility
Transistor) devices, for example, is better than that achieved with GaAs MESFET technology. These
factors should be borne in mind in establishing future limits for unwanted emissions.
3. Reduction of interference due to unwanted emissions
This section presupposes that a problem exists, due to inappropriate adjacent band allocations.
Regulatory possibilities are considered in the following section.
3.1. Suppression at transmitters
3.1.1. Transmitter architecture
The RF architecture of radio transmitters often takes the form shown in the simplified block diagram
of Figure 3.1. The modulated input signal is generated at an IF, then frequency translated in one or
more mixing and filtering stages to the final frequency.
A common problem with this arrangement is that each mixing process will produce many spurious
products, as well as the main sum and difference frequency components. These arise through mixing
of the local oscillator (LO) harmonics with harmonics of the IF input, often referred to as ‘m’ x ‘n’
products. Although the LO harmonics are unavoidable due to the switching action of the mixer LO
port, the IF harmonics can be reduced by ensuring that the IF port is operated well below
compression. However, in practice, a compromise must be reached between linearity and signal to
noise ratio, so the spurii can never be completely eliminated. Spurious products which fall at offsets
far removed from the wanted frequency can be suppressed through filtering, but those close to the
carrier will not be attenuated.
RF O/P IF I/P
Power Frequency conversion stages
Figure 3.1 Typical upconversion transmitter architecture
One way of mitigating this problem is to generate the wanted signal directly at the final frequency
using a vector modulator, as shown in Figure 3.2. In this case, in-phase and quadrature (I and Q)
baseband signals are used to directly modulate a carrier at the output frequency. Although spectral
spreading of the signal into the adjacent channels can still occur, the harmonic mixing effect is
eliminated, since there is only a single carrier component applied to the mixers.
A drawback with this arrangement is that there will be a finite carrier leakage to the output, typically
suppressed by about 30 dB relative to the wanted signal. Usually this is of no consequence, but in
cases where better carrier suppression is required, it is necessary to adjust the DC bias on the I and Q
inputs to null the carrier.
RF O/P 90°
Figure 3.2 Vector modulator transmitter architecture
While the arrangement illustrated in Figure 3.2 utilises two bi-phase AM modulators, it is equally
possible to use four uniphase modulators, and four orthogonal channels.
A more complex, but more flexible, approach is to use a single channel incorporating a digitally-
controlled attenuator, and a digitally-controlled phase-shifter. These two components are driven by
the baseband input by way of a look-up table, allowing the direct generation of virtually any (digital)
modulation scheme. It might be noted that the carrier-frequency amplitude/phase shifter is the same
component required for use in active antenna beam-forming arrays.
Filtering (generally bandpass filtering) of the transmitter output can be used in conjunction with the
other techniques discussed in this report to reduce the residual spurious output levels. The choice of
the type of filter to be used is, as usual, a compromise between a number of interacting, usually
conflicting, requirements such as out of band rejection, passband attenuation, time domain response,
size, weight, cost, etc.
Filter designs are usually based on the classical analytically derived categories such as Butterworth,
Chebyshev, etc. Some of these categories are optimised for one of its characteristics at the expense of
others, and some provide compromises between characteristics as summarised below :
Category Optimised Parameter Sacrificed Parameter
Butterworth Passband amplitude flatness Out of band rejection
Chebyshev Out of band rejection Passband amplitude
flatness and attenuation
Bessel Passband delay flatness Out of band rejection
Elliptic Close-in out of band rejection Out of band rejection
(Caur) (theoretically infinite at spot away from spot
Other categories provide compromises between characteristics. For example, the so-called Linear
Phase filter can be designed to provide a passband flatness approaching that of the Bessel filter, but
with improved out of band rejection. Similarly, Transitional filters have a near linear phase shift and
smooth amplitude roll-off in the passband, with improved out of band rejection compared to a Bessel
filter (but still significantly less than a Chebyshev filter).
As well as the characteristics described above, another factor which defines the performance of any
filter is its order of complexity, which is related to the number of poles and/or zeros in its transfer
function. In general, increasing the order of complexity improves the performance of the optimised
characteristic at the expense of degrading the performance of the sacrificed characteristic(s).
Figure 3.3 shows examples of the out of band rejection (which is the main performance parameter of
interest in the context of this study) for Butterworth, Chebyshev and Elliptic filters of order of
complexity n = 3. Note that the low-pass response is shown; in a practical design the band-pass
response would be derived from this by suitable scaling of the frequency axis. The figure therefore
illustrates the relative performance of these filter types.
1 2 4 6 8 10
Figure 3.3 Comparison of Butterworth, Chebyshev and Elliptic filters, n=3
Figure 3.4 shows examples of the out of band rejection for similar filters of order of complexity n =
7. The improved performance of these filters compared with those in Figure 3.3 can only be obtained
at the expense of increased implementation complexity and, in practice, increased insertion loss in
the wanted frequency band.
1 2 4 6 8 10
Figure 3.4 Comparison of Butterworth, Chebyshev and Elliptic filters, n=7
Transmitter output filtering nearly always requires the use of resonant elements such as tuned circuits
or transmission lines to form filter structures. Although SAW (surface acoustic wave) filters have
been produced for operation at up to 2 GHz, these have relatively low power handling (< 20 mW
typically). The insertion loss of SAW filters also tends to be quite high, up to 6dB for SAW
resonator filters, and up to 30 dB for transversal (delay line) filters.
At frequencies up to a few hundred MHz, LC (inductor capacitor) filters are usually used to achieve
bandwidths of 10% or more. Narrower bandwidths are possible, but the unloaded Q, tolerances and
temperature stability of the components generally precludes significant further reduction.
At higher frequencies, up to a few GHz, the commonest filter technologies are printed microstrip and
silver plated ceramic. Microstrip filters are generally limited to bandwidths no less than a few
percent, due to tolerances of the dielectric constant, substrate thickness and etching variability. The
unloaded Q of microstrip resonators (typically < 200) also limits the minimum practical bandwidth
due to insertion loss considerations.
The use of silver plated ceramic technology can achieve better performance owing to the higher
unloaded Q and excellent stability of the materials used. The digital cellular and cordless telephone
industry in particular, has prompted the development of very high dielectric constant, low loss
ceramics for use in miniature coupled resonator filters. A typical 2-pole 1.9 GHz filter for example,
can achieve an insertion loss of 0.8dB with a bandwidth of 1%.
At frequencies of several GHz and above, the resonant elements tend to be cavities or transmission
lines with an air dielectric. A common configuration is the interdigital filter, where several resonant
‘fingers’ are positioned within a single cavity to give the desired coupling, and hence overall filter
response. Performance is comparable with that of silver plated ceramic filters, with bandwidths
available as low as 0.2%.
An example of the mitigation available, and the cost at which it is achieved is given by the filtering
used at some UHF television transmitters to protect the Radio Astronomy Service. As described
above, interference is possible to RA receivers operating at 610 MHz from the adjacent channel,
recently assigned to high-power analogue television transmitters.
The transmitter operators have, therefore, installed high power filtering at certain transmitters to
reduce the emission of further modulation sidebands and intermodulation products. In the case of the
transmitter site closest to Jodrell Bank it has been necessary to install a 12-pole filter, allowing a
rejection, 2 MHz below the band-edge, of some 80 dB. This degree of filtering, however, is only
achieved at a cost amounting to some 25% of the entire transmitter installation.
3.1.3. Modulation techniques
In transmitters intended for single carrier applications, the choice of modulation mode can
significantly affect the level of adjacent channel energy. Paradoxically, schemes which can
potentially give the most constrained spectrum, often result in the worst performance, in this respect.
Figure 3.5 shows the theoretical normalised power spectral densities of various modulation schemes.
From this it can be seen that in the simplest case, binary phase shift keying (BPSK), the adjacent
channel energy reduces very slowly with offset from the carrier frequency.
Unfiltered quadrature phase shift keying (QPSK) and offset quadrature phase shift keying (OQPSK)
have a narrower main ‘lobe’ but otherwise show only a marginal improvement in suppression of
adjacent channel energy. OQPSK can give much lower out of band energy by filtering the baseband
signals before modulation. A root raised cosine filter for example, can theoretically give infinite
adjacent channel rejection. However, in practice, the filtering has a limited stopband and, more
importantly, since OQPSK is a non-constant envelope scheme, power amplifier non-linearity causes
spectral re-growth through AM to AM and AM to PM conversion.
Minimum shift keying (MSK) without baseband filtering has an improved rate of reduction of out of
band energy. This can be further improved by the addition of Gaussian baseband filtering (GMSK).
The degree of improvement depends on the parameters of the filter used, the example shown in
Figure 3.5 is for the case where the time-bandwidth product is 0.3 (as used in the GSM cellular radio
system). It can be seen that this scheme gives only moderate adjacent channel performance (typically
-40dBc at offsets comparable with the symbol rate), but since it is a constant envelope technique it
has the advantage that a limiting power amplifier can be used.
GMSK can be regarded as a special case of a class of constant envelope modulation techniques
known as continuous phase modulation (CPM) [1, 2]. As in the case of GMSK the details of the
power spectral density of the CPM signal depends on various parameters. The example shown is the
case of a 4-level signal, modulation index 0.33 and raised cosine baseband filtering of 3 symbol
duration. (M=4, h=0.33, 3RC)
In practice, limitations in the accuracy with which these advanced modulation schemes can be
implemented restrict the degree of suppression of out of band energy that can be achieved. The signal
envelope is nearly but not exactly constant, so the power amplifier non-linearity can still cause some
spectral re-growth, although this effect is not as severe as in the case of OQPSK.
Normalised power spectral density, (dB)
QPSK and OQPSK
-3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0 0.5 1.0 1.5 2.0 2.5 3.0
Normalised Frequency Offset fromcarrier (f-fc)/R (Hz/bit/s)
Figure 3.5: Power Spectral Densities of some Example Modulation Schemes
A recent development is the use of COFDM (Coded Orthogonal Frequency Division Multiplex) in
digital broadcasting (audio and video) . This modulation technique produces a comb of carriers,
usually separated by a few kHz, where each carrier is modulated at a low symbol rate by orthogonal
data steams. The overall spectrum is therefore almost rectangular. However, the amplitude
distribution of such a signal is virtually noise-like, consequently considerable back-off is required in
the power amplifier to allow for the peak to mean ratio. Headroom of 12 to 15 dB would be typical.
Clearly, amplifier non-linearity is also a problem with this technique.
In multi-carrier systems, where a single power amplifier is used to amplify several carriers, the
problem is compounded by intermodulation products between carriers. In this case, spurious
products can be generated at multiples of the carrier spacing.
Linearisation of a transmitter system may be accomplished by a number of methods:
Feedforward linearisation: This technique compares the amplified signal with an appropriately
delayed version of the input signal and derives a difference signal, representing the amplifier
distortions. This difference signal is in turn amplified, and subtracted from the final HPA output. The
main drawback of the method is the requirement for a second amplifier - the technique can, however,
deliver an increase in output power of some 3 dB when used with a TWT.
Feedback linearisation: In audio amplifiers, linearisation may readily be achieved by the use of
feedback, but this is less straightforward at high RF frequencies due to limitations in the available
open-loop amplifier gain. It is possible, however, to feedback a demodulated form of the output, to
generate adaptive pre-distortion in the modulator. It is clearly not possible to apply such an approach
in a ‘bent-pipe’ transponder, however, where the modulator and HPA are rather widely separated.
Predistortion: Rather than using a method that responds to the actual instantaneous characteristics
of the HPA, it is common to pre-distort the input signal to the amplifier, based on a priori knowledge
of the transfer function. Such pre-distortion may be implemented at RF, IF or at baseband. Baseband
linearisers, often based on the use of look-up tables held in firmware memory are becoming more
common with the ready availability of VLSI techniques, and can offer a compact solution. Until
recently, however, it has been easier to generate the appropriate pre-distortion function with RF or IF
RF amplifier linearisation techniques can be broadly divided into two main categories :-
1. Open-Loop techniques, which have the advantage of being unconditionally stable, but
have the drawback of being unable to compensate for changes in the amplifier
2. Closed-Loop techniques, which are inherently self-adapting to changes in the amplifier,
but can suffer from stability problems.
This following sections review linearisation techniques:
This involves placing a compensating non-linearity into the signal path, ahead of the amplifier to be
linearised, as shown in figure 3.6. The signal is thus 'predistorted' before being applied to the
amplifier. If the predistorter has a non-linearity which is the exact inverse of the amplifier non-
linearity, then the distortion introduced by the amplifier will exactly cancel the predistortion, leaving
a distortionless output.
Predistorter Power Amplifier
Figure 3.6: Predistortion concept
In its simplest analogue implementation, a practical predistorter can be a network of resistors and
non-linear elements such as diodes or transistors. Several examples of this technique have appeared
in the literature, eg. , where the reduction in third order intermodulation distortion that has been
reported is typically in the range 7-15dB. The poor performance is due to the fact that the amplifier
characteristics are not constant, but vary with time, frequency, power level, supply voltage and
Better results have been reported in , where a pair of FET amplifiers are used as the predistorter,
as shown in figure 3.7. In this arrangement, the input signal is unequally split between the two
amplifiers, such that one of them is driven into compression. The compressed output is then scaled
and subtracted from the linear output to produce the inverse of the compression characteristic, as
required. Reduction in intermodulation distortion of around 20dB has been measured using this
technique, but only when the main amplifier is operated with at least 1dB of back-off.
Atten Amp 1
Phase Amp 2 Atten
Figure 3.7: Soft-limiter Predistortion
Although adaptive predistortion schemes have been reported, where the non-linearity is implemented
in digital signal processing (DSP), they tend to be very computationally or memory intensive, and
Feedforward  involves comparing the power amplifier input and output signals to derive an error
or distortion term in a signal cancelling loop. This residual error is then amplified in a separate, low
power amplifier before being subtracted from the main amplifier output in an error cancelling loop.
This is shown in Figure 3.8. If the low power, auxiliary amplifier is perfectly linear and the error
cancelling loop is perfectly balanced, then the overall result is distortionless amplification. However,
in practice the cancellation loops are only partially effective, and the technique is compromised.
Figure 3.8: Feedforward
In a practical feedforward implementation, there will be imbalance in the error cancelling loop which
will limit the distortion reduction. For example, a 1 dB gain error and a 10° phase error limits the
distortion suppression to just 14 dB. To improve this to say, 30 dB, would require the balancing to
be within 0.3 dB and 1° . Even if such stringent requirements can be met, the overall linearity can
never be better than that of the auxiliary amplifier, which must therefore operate in Class-A and will
consequently be inefficient. These problems are further compounded by errors in the signal
cancelling loop, which will increase the power handling requirements of the auxiliary amplifier. A
gain error of 2 dB and a phase error of 10°, for example, demands that the second amplifier output
power is only 12 dB below that of the main amplifier.
An example of a practical application of feedforward is given in  which concerns a 30 W HF
amplifier. Here, the auxiliary amplifier had the same power rating as the main amplifier, yet the
distortion reduction achieved was still no more than 15 dB across the band. Interestingly, when the
two amplifiers were connected in parallel with each one operating at half power, the results were
It may be seen therefore that the improvements achieved through the use of feedforward are modest,
and that the overall power efficiency is seriously degraded.
Negative feedback  is the most well known linearisation technique and is widely used in low
frequency amplifiers, where stability of the feedback loop is easy to maintain. With multi-stage RF
amplifiers however, it is usually only possible to apply a few dB of overall feedback before stability
problems become intractable . This is mainly due to the fact that, whereas at low frequency it can
be ensured that the open-loop amplifier has a dominant pole in its frequency response (guaranteeing
stability), this is not feasible with RF amplifiers because their individual stages generally have
Of course, local feedback applied to a single RF stage is often used, but since the distortion reduction
is equal to the gain reduction, the improvement obtained is necessarily small because there is rarely a
large excess of open loop gain available.
At a given centre frequency, a signal may be completely defined by its amplitude and phase
modulation. Modulation feedback exploits this fact by applying negative feedback to the modulation
of the signal, rather than to the signal itself. Since the modulation can be represented by baseband
signals, we can successfully apply very large amounts of feedback to these signals without the
stability problems that beset direct RF feedback.
Early applications of modulation feedback used amplitude (or envelope) feedback only, applied to
valve amplifiers , where amplitude distortion is the dominant form of non-linearity. With solid-
state amplifiers however, phase distortion is highly significant and must be corrected in addition to
the amplitude errors. The first successful practical implementation of simultaneous amplitude and
phase feedback was demonstrated by Petrovic and Gosling , and is known as the Polar Loop
220.127.116.11. The Polar Loop Technique.
The Polar Loop Technique is based around the principle of Envelope Elimination and Restoration
(EER), first proposed in 1952 by L R Kahn , but modified to allow feedback to be applied. A
block diagram of the Polar Loop technique is shown in figure 3.9.
Figure 3.9: The Polar Loop Technique
The RF stages of the system are particularly simple. They consist of a voltage controlled oscillator
(VCO) running at the output frequency, which generates the phase component of the output signal,
an amplitude modulated stage which generates the amplitude component, and the main power
The input signal to the Polar Loop is first generated at intermediate frequency (fi) and at low power
level (shown as the exciter block in the diagram). It is then resolved into polar coordinate form by
envelope detection to produce the amplitude component, and hard-limiting to give the phase
component. The envelope detection is conveniently achieved by multiplying the input signal by the
limiter output in a double balanced mixer (a process equivalent to full-wave rectification). A sample
of the final RF output is translated (usually down) to the same frequency as the input signal, and is
similarly resolved into its polar coordinates. The two envelope signals are then compared in a high
gain differential amplifier which in turn controls the amplitude modulator, forming an envelope
feedback system. The two phase modulated signals are phase compared in a phase sensitive detector
(PSD), and the amplified error signal controls the VCO forming a phase locked loop (PLL). The
overall effect is that two orthogonal feedback loops are formed, which by suitable choice of loop
gain and bandwidth, attempt to make the amplitude and phase of the output signal closely approach
that of the IF input.
The two main limiting factors in the performance of a Polar Loop system are :-
(i) The balance between the two polar resolving circuits (limiters + mixers).
(ii) The relative bandwidths of the feedback loops and the amplitude and phase spectra (which
determines the amount of negative feedback available).
In practical Polar Loop transmitters designed for narrow-band (5 kHz) applications, it has been found
that the balance of the resolving circuits is the main problem, and this sets a minimum value to the
residual third order intermodulation distortion of around -60 dBc . For wider bandwidth signals,
it is the finite amount of feedback which is the main restriction. This is particularly true for signals
where the envelope can fall to zero, as the zero-crossing often results in a sharp discontinuity in both
the envelope and phase waveforms, and consequently produces envelope and phase spectra which are
considerably wider than the composite signal bandwidth.
An alternative approach to modulation feedback, which overcomes both of the above problems is
known as the Cartesian Loop technique invented by Smith and Petrovic  and this is covered in
the next section.
18.104.22.168. The Cartesian Loop Technique.
The Cartesian Loop technique makes use of the fact that a modulated RF signal can be represented in
complex (I and Q) baseband form as well as by amplitude and phase functions.
If negative feedback is applied to I and Q rather than A and , this leads to the configuration shown
in Figure 3.10.
The principal of operation is as follows :-
Complex baseband signals, Imod and Qmod, are used to modulate in-phase and quadrature local
oscillator signals in double balanced mixers, and the combined output forms the input to the driver
and power amplifier. A sample of the PA output is fed to a second pair of mixers configured as
demodulators which use the same local oscillators. The RF output is thus coherently demodulated
back down to I and Q baseband. These signals, Ifb and Qfb, are then fedback and compared with the
input signals, Iin and Qin, in high gain differential amplifiers, the outputs of which form the inputs to
the modulators, Imod and Qmod. Just as in the Polar Loop, two orthogonal feedback loops are thus
formed which attempt to make the I and Q demodulated outputs closely approach the I and Q inputs.
Note that because of the coherent nature of the feedback, the technique is identically equivalent to
RF feedback, but because dominant loop poles are introduced by the differential amplifiers, a good
phase margin of stability may be easily maintained, even when very large amounts of feedback are
Figure 3.10: The Cartesian Loop Technique
The delay element shown in the diagram is to ensure that the RF output and the demodulating
carriers are at the correct relative phase. Perfect alignment is not necessary owing to the
compensating action of the loops.
The effectiveness of the Cartesian Loop depends on two factors :-
(i) The ratio of the feedback loop bandwidths to the I and Q input bandwidths (determines
the amount of feedback).
(ii) Linearity of the demodulators (since the I and Q demodulated outputs must be a linear
representation of the RF output).
Note that unlike the Polar Loop, the RF output bandwidth is simply twice the I and Q bandwidth.
We do not have the problem of generating wideband A and signals.
Practical Cartesian Loop transmitters have been constructed which operate at up to 900 MHz  for
relatively narrow band signals (< 5 kHz bandwidth) and these have achieved excellent results. On a
two-tone test, third order intermodulation products are typically reduced by 40 dB, compared to the
same transmitter with the power amplifier run open loop.
With the increasing use of on-board processing and re-modulation in satellite systems, the use of
modulation feedback to improve HPA linearisation has become possible. There are still formidable
bandwidth limitations associated with these techniques, however, and RF pre-distortion methods are
currently more widely used.
3.2. Mitigating measures by the passive services
Many RA observatories are located in remote areas chosen to provide reasonable protection from
terrestrial interference. In densely populated parts of Western Europe this is often not possible, and
the location of sites such as Jodrell Bank has been determined as much by historical accident as by
any other considerations.
In the case of millimetre-wave observations, it is essential to locate telescopes above as much of the
atmosphere (especially water-vapour) as possible, to avoid absorption of cosmic emissions. As a
consequence these sites are often on mountain-top sites such as Pico Veleta in Spain (2870m) and
Plateau de Bure in France (2552m).
In specific cases of terrestrial interference it might be possible to construct artificial screens to
mitigate terrestrial interference. It should be borne in mind, however, that the feed point of a large
telescope might be at a considerable height above ground level, implying the need for an extremely
large structure. For smaller instruments, however, this method can offer useful protection. The
cosmic anisotropy telescope at Cambridge employs aluminium-lined earth banks to reduce
interference to the 70 cm diameter interferometer elements. The method, however, restricts
observations to elevation angles greater than 30.
Filtering is routinely used at the front-end of radio astronomy receivers to reject adjacent-band
energy (see the discussion above, relating to the filtering of television carriers at 610 MHz and cloud-
profiling radar systems at 95 GHz). The suppression of unwanted emissions falling within the
passive bands is obviously impossible by filtering.
3.2.3. Antenna pattern
The majority of interference events experienced at an RA site will be due to the entry of unwanted
power via the sidelobes of the telescope antenna. If it were possible to reduce the level of sidelobe
response, so the amplitude and duration of the interference events would decrease.
However, the attainment of excellent antenna characteristics is already one of the key drivers in the
design of radio telescopes. A reduced sidelobe response will, by reducing antenna noise, increase the
sensitivity of the overall instrument.
In the design of Earth Station antennas, the use of an offset feed is often preferred to avoid the
degradation of the sidelobe pattern due to the feed or sub-reflector supports, and to reduce loss of
gain due to aperture blockage. For RA use this is not generally appropriate, as the offset feed
arrangement will degrade the cross-polarisation performance, and the overall symmetry of the
antenna response is an important consideration in radioastronomy. It is interesting to note that the
25m x 38m elliptical telescope (the Mark II) at Jodrell bank is normally only used with the
symmetrical part of the aperture illuminated, for this reason.
An antenna designed to minimise sidelobe levels has been described , for use in propagation
measurements at 20/30 GHz. This 7 m antenna achieved sidelobe levels (at 20/30 GHz) of some -40
dB at 1 from boresight, by the use of an offset feed, and a reflector surface tolerance of 0.01 (0.1
mm rms). This surface tolerance incidentally permitted use of the antenna up to 300 GHz for
radioastronomical measurements. Good cross-polarisation performance was obtained by the use of a
large secondary focal-ratio, which ruled out the use of a corrugated horn feed, a quasi-optical feed
being used instead.
A limitation on the performance of large aperture antennas is imposed by the deformation of the
paraboloid due to gravity. As the elevation angle of the telescope changes, so the curve will deform,
degrading the pattern. It is possible to counter this distortion by mounting the reflecting surface on an
active substrate that can be computer controlled to optimise the reflector geometry. Such active
techniques are finding increasing application in optical astronomy. An alternative approach,
implemented at the 100m telescope at Effelsberg (the largest fully-steerable in the world) is to use
the principle of ‘homologous deformation’ . In this scheme, the mechanical design of the
telescope ensures that as the elevation angle is varied, the reflector curve deforms from one
paraboloid curve to another. It is then possible to adjust the position of the feed, under computer
control, to accommodate the change in focal point.
The design of antennas for satellite radiometers is patently not influenced by considerations of
gravitational deformation, but by the severe constraints on available space (at launch, at least). The
antenna for the UK designed AMSU-B sensor (operating at 89, 150 and 183 GHz)  uses a single
reflector, with a quasi-optical feed employing dichroic plates for frequency de-multiplexing (as used
optically in television cameras to separate the three colour channels). Sidelobe levels are constrained
by the careful consideration of reflector and sub-reflector illumination, to avoid beam truncation:
Fourier optics demonstrate that such truncation will be transformed to an increased sidelobe
3.2.4. Operational measures
It has been proposed by one MSS operator, that a system of time-sharing would allow the
‘simultaneous’ use of frequencies at 1.6 GHz for both MSS downlinks and radio astronomy
observation. It is intended that observatories be furnished with a blanking device, that would inhibit
the RA receiver for the period of the MSS time-division duplex (TDD) frame used for spacecraft
While this might be technically feasible, it would be applicable only to a small number of
astronomical observations. It would be inapplicable to interferometric observations, and to the
observation of pulsars (accounting for some two-thirds of time on the Lovell telescope at Jodrell
bank) for which the study of the temporal characteristics of the emission is vital.
For the remaining category of single-antenna RA observations, the loss of half the available
integration time would necessitate longer observations, and reduce the utilisation of expensive
research facilities. Whether this is acceptable or not is a matter more of public policy than of
4. Further work
Further study might usefully be directed towards an investigation of the actual levels of unwanted
emissions achieved by a wide range of operational equipment. This would be followed by theoretical
(and laboratory) work to determine what levels of suppression might be achieved for the same
systems, were the reduction of UWE to be a primary design driver. A specific example of such work
is a project associated with the need to reduce interference from the GLONASS system . It is
suggested that a campaign of this sort, on a more general basis, might usefully inform the
determination of more appropriate limits.
The passive services suffer both directly and indirectly from interference: the direct effects have been
described in section 5, above. In addition, however, interference imposes a burden on these services
in terms of the staff time and resources that need to be directed to controlling and investigating
interference, in place of core research work. Such resources are necessarily constrained, and this
work is therefore normally carried out in a reactive manner, when specific problems become
intolerable. There might be a case for funding to be made available specifically to investigate the
interference environment and matters relating to unwanted emissions, using the facilities and
expertise available within radio astronomy (and other) research groups.
The expertise available to such groups might also inform the development of the sensitive measuring
techniques required for the verification of equipment compliance.
While the passive services are often the only victims of interference due to unwanted emissions at
present, such pollution will ultimately limit the use of the spectrum by all services. Current and
proposed limits for unwanted emissions offer inadequate protection to the passive services.
The passive services have limited opportunity to introduce general measures to mitigate interference
problems. Such mitigation techniques will generally involve assumptions regarding the character of
the observations that are to be made. Due to the research nature of these services (particularly in the
case of radio astronomy) it is not generally possible to make such assumptions.
Particular attention should be addressed to the effect of wideband digital modulation schemes, which
seem likely to give rise to an insidious rise in background noise levels across the spectrum. Because
the spectral density of such signals is so low, the increase will generally be imperceptible.
 Aulin, T. and Sundberg, C-E.: “Continuous phase modulation - part 1: Full response
signalling”, IEE Trans. on Comm., COM-29(3), March 1981
 Ponsonby, J.E.B.: “CDMA and DSSS: their impact on Radio astronomy and other
spectrum users and a proposal to reduce their impact”, COMSPHERE 95,
Proceedings of, pp.115-121, Eilat, Israel, January 1995.
 Shellswell, P.: “The COFDM modulation system: the heart of digital audio
broadcasting”, Electronics and Communication Engineering Journal, Volume 7
Number 3, IEE, June 1995
 Gray, L.F., 'Application of broadband linearisers to satellite transponders', IEEE Conf.
Proc. ICC 80.
 Aghvami, A.H., Robertson, I.D., ‘Power limitation and high-power amplifier non-
linearities in on-board satellite communications systems’, Electronics and
Communication Engineering Journal, April 1993.
 Black, H.S., 'Translating system', U.S. Patent No. 1686792, Oct. 1928.
 Bennet, T.J., Clements, R.F., 'Feedforward - An Alternative approach to Amplifier
Linearisation', Radio and Electronic Engineer, May 1974.
 Black, H.S., 'Wave Translating system', U.S. Patent No. 2102671, Dec. 1937.
 Mitchell, A.F., 'A 135MHz Feedback Amplifier', IEE Colloq. Broadband High
Frequency Amplifiers, Nov. 1979.
 Arthanayake, T., Wood, H.B., 'Linear amplification using envelope feedback', Elec.
Lett., 8th April 1971.
 Petrovic, V., Gosling, W., 'Polar Loop Transmitter', Elec. Lett., 10th May 1979.
 Kahn, L.R., 'SSB Transmission by Envelope Elimination and Restoration', Proc. IRE,
 Smith, C.N., 'Application of the the Polar Loop Technique to UHF SSB Transmitters',
Ph.D. Thesis, University of Bath, 1986.
 Smith, C.N., Petrovic, V., 'Cartesian Loop Transmitter', Internal Research Report,
University of Bath, School of Electrical and Electronic Engineering, 1982.
 Cole, R.A., 'Linearisation of a Power Amplifier using Cartesian Loop Feedback', Report
No. 72/89/R/451/C, Dec. 1989, Roke Manor Research.
 Chu, T.S et al.: “The Crawford Hill 7-Meter millimeter wave antenna”, Bell System
Technical Journal, Volume 57, Number 5, May-June 1978
 Hachenbrg, O, Grahl, B. and Wielebinski, R..: “The 100-Meter radio telescope at
Effelsberg”, Proc. IEEE, Volume 61, Number 9, Sept. 1973
 Martin, R.J. and Martin, D.H.: “Quasi-optical antennas for radiometric remote-sensing”,
Electronics and Communication Engineering Journal, Volume 8 Number 1, IEE,