A conceptual framework for Radio Design Project
Direct Digital Synthesis
There is no denying the fact that Digital Synthesis as the name suggests creates the source frequency from digitally generated samples of the output waveform which are then passed through a D/A converter. With sufficient processing speed and word-length in the device, it is practical to achieve very fine resolution of the output frequency (mill hertz accuracy) with incredibly fast switching times (s).
Current single chip DDS devices from Infineon, Analog Devices, QUALCOMM, etc costing only a few $ can generate frequencies up to 100MHz, using a clock rate of about 300MHz. Their main drawback with this clock rate is that they consume vast amounts of power, (approx.1W!). Further the stability is only as good as the stability of the clock source driving the DDS unit.
By using DDS in conjunction with conventional PLL based synthesis it is possible to get the benefits of each method, but with a cost, power consumption and size penalty that makes them currently much more suited to high power base-station applications than low power telemetry units.
Transmitter & Receiver
Architectures
Transmitter & Receiver Architectures:
Basic Building Blocks
Frequency Source (Carrier)
In order to make the information signal (e.g. data) pass through the air, it must be modulated on to a carrier signal whose frequency is well suited to the propagation environment, conforms to the licensed operating bands, and is sufficiently stable to allow detection by a tuned receiver in the presence of interference.
Modulation
The method of imposing the information signal onto the carrier signal is termed modulation and must be accomplished cost effectively and accurately for maximum range and minimum interference.
Amplification
The amplifier is a key part of the transceiver, and must be efficient (dc power in to RF power out), low cost, non-polluting, and possibly linear. Output power is dictated by regulation, range requirement, battery life, cost and linearity considerations.
Antenna
The antenna is often the most poorly engineered part of a radio system. Good design will ensure maximum range, high amplifier efficiency (good matching), good selectivity, minimal pollution, good interference rejection, good sensitivity, reduced design headaches.
Reception
Key to the sensitivity, dynamic range and strong signal handling properties of the radio is the receiver ‘front end’. The main task is to boost weak wanted signals, often in the presence of strong unwanted signals whilst introducing minimal noise and distortion. In many cases, some selective filtering is required to assist this task.
Demodulation
The process of removing the information signal from the carrier is termed demodulation. The challenge is to design a circuit (or algorithm) that will achieve this task optimally in the presence of noise, interference and varying signal strength, frequency and phase, whilst being compact, power efficient and cheap.
Data Processing
Pre and post processing of the information signal is often an after thought for low power radio applications, usually implying some form of microprocessor or DSP engine with the presumed complexity, cost, power consumption and size penalties. The benefits of matched filtering, error detection and correction (coding), channel equalization, etc, are however significant in terms of range, robust transmission, power conservation and data rate optimization.
The following sections cover each of these building blocks in greater detail.
Frequency Sources:
Requirements
The two most crucial factors affecting the design of a carrier frequency source are:
• Frequency Stability
• Phase Noise
Frequency Stability: The stability of the oscillator with temperature (and ageing) determines the channel spacing required to contain the modulated carrier signal. Conversely, for a given regulated channel spacing, the frequency stability determines the maximum data rate that can be supported without violating the channel boundary (spectral mask).
A UHF transmitter in the 868MHz band having to meet ETS300 220 for a 25kHz channel, may require a stability of better than +/- 3ppm, (better if maximum data rate is to be achieved), whilst for wideband operation at 433MHz, also under ETS300 200, the permissible error is up to +/- 800kHz or 1800ppm.
3ppm source accuracy at 868MHz gives a frequency uncertainty of:
868x106 x 3x10-6 = 2.604kHz
Phase Noise: The oscillator phase noise (phase jitter) results in a broadband component to the carrier signal which will extend into adjacent channels. If the phase noise is too high, this can corrupt the modulation source itself (if PSK or QAM based), and limit adjacent channel selectivity due to reciprocal mixing.
Tuned Resonators – LC, Ceramic, SAW, Crystal
A frequency stable oscillator requires a temperature stable tuned element in the feedback circuit.
A simple LC stabilised oscillator is generally not stable enough and will also suffer badly from microphony.
Using a transmission line in place of the LC circuit will be much less susceptible to microphony and usually exhibit less drift with temperature.
A ceramic resonator, widely used as a clock source for microprocessor applications, has a frequency tolerance of only approx. 3000ppm and hence is not adequate for most radio applications.
Co-axial resonators (often ceramic based) can provide very high Q elements and hence have excellent phase noise. Their temperature stability is however poor and on their own they are not sufficient to act as the carrier source. (Locking a coaxial resonator to a stable (e.g. crystal) source is however a very viable and widely used synthesis process).
Surface Acoustic Wave (SAW) devices can provide the accuracy needed for wideband operation, giving typically +/- 300ppm stability over –40oC to +60oC for frequencies up to 1GHz or beyond. The usually have an initial tolerance of approx. +/- 50kHz which can be tuned on manufacture (at a price). Because they provide a low cost, on-frequency source, SAW based oscillators are very popular for wideband low power radio solutions.
A crystal oscillator is the most widely used stable frequency source, relying on the natural fundamental (or overtone) vibration frequency of the quartz crystal. Depending on the quality of the crystal and the type and tolerance of the cut, accurate, temperature stabilised devices can be fabricated. For a very cheap crystal, a temperature tolerance of +/- 15ppm is typical, whereas for a Temperature Compensated Crystal Oscillator (TCXO), a +/- 3ppm tolerance, or better, is possible. The major drawback with crystal based oscillators compared with SAW devices is that the oscillation frequency is typically below 30MHz for a fundamental mode design, and possible up to 100MHz using a fifth overtone design. (Overtone crystals are more expensive and difficult to control in a low cost circuit design).
Frequency Synthesis:
In order to make use of the stability of a crystal reference, whilst generating a carrier frequency of several hundred MHz, some form of frequency multiplication or synthesis is required. This next section considers a range of options for realising this function.
Harmonic Generation.
Passing a signal through a non-linear circuit will result in harmonics of the waveform being generated. Turning a sinewave carrier into a squarewave will generate a signal rich in odd order harmonics of the fundamental frequency. Using a tuned circuit to extract one of the harmonics can deliver the required output carrier frequency. As the harmonic amplitude falls with order, it is usual to cascade a number of low order harmonic generation components, each giving perhaps x2 or x3 frequency multiplication.
An alternative to using a non-linearity to generate harmonics is the use of linear mixing to create multiples of the fundamental frequency. A perfect balanced multiplier will act as a frequency doubler, and if cascaded can generate any integer value of the fundamental source frequency. This approach to frequency synthesis is both relatively expensive and in-flexible. Today it has been superseded by the fully integrated phase locked loop based synthesiser, (see below).
Injection Locking
Injection locking is a common technique for providing stable microwave oscillators but finds little use in lower frequency applications due to its extreme design/layout sensitivity. The technique is based on the fact that a free running oscillator will (usually) lock onto a nearby low level frequency source (or harmonic of the source) and thus combine the benefits of the free running source – perhaps high output power or good phase noise, with the stability of the low level injected reference.
Phase Locked Loop Synthesizers
The phase locked loop (PLL) forms the basis of most modern carrier generation solutions. Like injection locking, the technique aims to get a high quality (good phase noise), high frequency oscillator to be locked to a stabile, low level, low cost, (often crystal derived), source.
The most simple form of PLL synthesizer contains a voltage controlled oscillator (VCO) operating at the required carrier frequency, (e.g. co-axial resonator or SAW design), a frequency divider (a digital divider circuit), a phase comparison circuit and a loop filter. The circuit acts to lock the reference frequency (fREF) to fVCO/N, resulting in an output frequency of N.fREF .
By adding a further divider between the reference frequency source and the PLL, it is possible to synthesise a whole range of output frequencies with value: N.fREF/M. With the divider ratio N/M under microprocessor control, a flexible channel synthesiser is realised. Today it is possible to buy single or dual synthesisers on a single chip containing all the circuitry required (excluding VCO and loop filter).
Fractional-N Synthesis
A more recent extension of the basic PLL based synthesiser is the fractional-N device. This allows the generation of carrier frequencies that are no longer constrained to multiples of the integer divider ratio N/M without compromising on phase noise of the output waveform. These new devices allow the generation of pure high frequency sources with flexible channel spacing from a 100’s of hertz to 100’s of kilohertz.
The only current drawback with fractional N synthesis is the somewhat poor level of spurious components that can arise at some output frequencies. Careful design can play a big part in achieving a satisfactory solution. Again single or dual fractional N synthesisers are available on chip, (e.g. Motorola, Nat Semi), with a cost of around $3 and operating frequencies up to 3GHz.
Direct Digital Synthesis
Digital synthesis as the name suggests creates the source frequency from digitally generated samples of the output waveform which are then passed through a D/A converter. With sufficient processing speed and word-length in the device, it is practical to achieve very fine resolution of the output frequency (millihertz accuracy) with incredibly fast switching times (s).
Current single chip DDS devices from Infineon, Analog Devices, Qualcom, etc costing only a few $ can generate frequencies up to 100MHz, using a clock rate of about 300MHz. Their main drawback with this clock rate is that they consume vast amounts of power, (approx.1W!). Further the stability is only as good as the stability of the clock source driving the DDS unit.
By using DDS in conjunction with conventional PLL based synthesis it is possible to get the benefits of each method, but with a cost, power consumption and size penalty that makes them currently much more suited to high power base-station applications than low power telemetry units.
Modulation:
The process of imposing the information signal onto the carrier is termed modulation and can be achieved in a wide variety of ways depending on the type of modulation required.
Essentially there are three basic ways of changing the properties of the carrier signal to convey information: Amplitude Modulation, Frequency Modulation and Phase Modulation. Modern low power radio devices use all three of these methods and some of the more recent designs use combinations to achieve improved data rate and range.
A detailed discussion of digital modulation formats and their merits, performance and implementation is covered in ‘Digital Communications – Design for the Real World’ published by Addison Wesley Longman, and in the Digital Modulation section of this course. For now, a brief overview of methods for modulating a carrier, independent of the data type, is given.
FM Modulation
Varactor
The simplest method of varying the frequency of a carrier it to include a voltage controlled impedance element in the oscillator circuit. The Varactor (voltage controlled capacitor) is most widely used, costing only a few pence, and providing reasonably linear variation in capacitance (and hence oscillator frequency) with applied voltage.
PLL Based Modulation
To achieve stability of the modulated source, the PLL is again widely used, with the varactor based oscillator acting as the VCO. Frequency modulation can be achieved in one of three ways:
a) Modulating the reference source (typically using a varactor based crystal oscillator or DDS source), in which case the modulation must not exceed the tracking bandwidth of the loop otherwise the modulation will not be recreated at the output. (Note: the output frequency shift, (modulation index), is N/M times the input frequency shift).
b) Adding the modulation signal to the VCO control voltage within the loop. This only works if the modulation rate exceeds the tracking bandwidth of the loop such that the loop is unable to track it out, which it will attempt to do!
c) A combination of the above.
Vector Modulation
A vector modulator implements a complex mathematical sum, multiplying two input signals with in-phase (e.g. cosine) and quadrature (e.g. sine) versions of the carrier and summing the result. If the input signals are arranged to be cos(1t) and sin( 1t), where 1 is the desired frequency shift to be imposed on the carrier, then the output of the summing junction with be duly shifted by the required amount.
With single chip implementations of a vector (quadrature) modulator now available (RFMD, Analog Devices, HP, Motorola, etc). this approach is gaining support, making the design of the carrier source (synthesiser) much simpler. In addition, the vector modulator is not limited to frequency modulation but can impose any complex modulation format (amplitude, frequency, phase) onto a carrier.
AM Modulation
On-Off Keying
The simplest, (and most spectrally polluting), form of amplitude modulation for data transmission is to turn on and off (gate) the carrier source is sympathy with the data signal. Simply altering the bias or controlling the supply to the oscillator will achieve this. The reason that this approach is often frowned upon is that pulsing a carrier on and off creates a broad modulation spectrum as the harmonics of the square wave modulating signal extend many times beyond the data rate being conveyed.
Mixer Based
Controlling the turn on and turn off rate of the carrier so that the amplitude grows and decays smoothly significantly reduces spectral pollution without degrading the performance of the system. Whilst it is possible to design a circuit that will achieve this for the carrier source itself, it is often simpler and cheaper to use a mixer to impose a pre-shaped data signal onto the fixed amplitude carrier.
Mixer Types
Whilst on the subject of mixers, it is worth looking at the properties of various types of mixer implementation:-
Ideal Mixer: The ideal mixer will perform the mathematical multiplication of the two input signals, creating components positioned at frequencies equal to the sum and difference of the input signals and no additional components. For this to be the case, the mixing (multiplying) device must be perfectly linear and there must be no leakage of the input signals to the output port.
Double Balanced Mixer: The term balanced mixer is used to imply that neither of the input terms will appear at the mixer output. In practice, suppression of these input components is never perfect in an analogue mixer circuit, (it can be virtually perfect in a digital implementation). A balanced mixer can be implemented using a transformer coupled diode arrangement (termed a passive mixer) of the type supplied by Mini Circuits and others, or using an active transistor based design, the most well know of which was invented by Barry Gilbert of Analog Devices, and called the Gilbert cell mixer.
Both types of mixer produce signals at odd harmonics of the carrier frequency, particularly the diode ring mixer. In most cases these can be easily filtered out.
Single Balanced Mixer: The difference between a single and double balanced mixer in practical terms is that the output will contain both even and odd harmonics. With 2nd order harmonic being the most difficult to suppress, double balanced mixers are the favoured solution.
Phase Modulation
Continuous phase modulation can be achieved using a frequency modulator with an integrator imposed between the modulating signal and the modulation circuit. Where discrete phase changes are required for conveying digital information, it is often much simpler to switch phases of the carrier source or use vector modulation – see next section.
Complex Modulation
To realise a combination of amplitude, frequency or phase modulation it is possible to use combinations of the techniques described above. The most elegant solution, however, (and today arguably the smallest and cheapest), is the vector modulator.
Vector Modulator
By applying the correct input to the two arms of the vector modulator, it is possible to impose any amplitude, frequency or phase shift directly onto a carrier in real time. If the carrier feeding the vector modulator is at the final output frequency for the radio, then the modulation is imposed directly onto the output signal in what is termed a direct conversion transmitter.
The function of the vector modulator can be realised in analogue or digital form. The digital form, whilst nearly perfect, has an upper frequency limit today of approx. 100MHz. The analogue form can be used with any frequency, (single chip devices now extend beyond 3GHz), but have imperfections in the mixing and summing process. These imperfections in practice result in an image of the modulation signal appearing at the output of the modulating circuit and residual carrier leakage.
In some cases it is possible to compensate for these imperfections by pre-distorting the input signal.
Combined Modulator/Amplifier
We shall see in the amplifier section to follow that it may in fact be advantageous, in terms of power efficiency, to split a complex modulating signal into its separate amplitude and phase/frequency components and independently impose these variations onto the carrier as part of the amplification process.
Amplification:
Objectives
The key design objectives for an RF amplifier are to boost the output signal level into the antenna with minimum cost, size, distortion, power dissipation, and spectral pollution. Optimising all of these parameters is a major design challenge and one that becomes increasingly difficult as the complexity of the modulation format is raised. The following gives a brief discussion of amplifier solutions to meet these challenges for different modulation types.
Constant Envelope Amplifiers
The simplest modulation formats to amplify are frequency modulation (FSK) or On-off keying. In both cases, when the amplifier needs to deliver power, it is at a constant level and hence at a fixed operating point on the amplifier power transfer characteristic. This means that the amplifier only needs to be optimised to deliver power efficiently at that output level and for a fixed input level. The linearity of the transfer characteristic is thus unimportant. As a result, most simple FM units use Class C biased amplifier designs with efficiencies of between 50% and 70%.
The position is complicated a bit when the carrier needs to be pulsed on and off, (i.e. with On-off keying or Time Division Multiplexed Operation), and the rise and fall times of the pulse edges need to be carefully controlled to minimise spectral pollution. This requirement exists for some of the more stringent ETSI standards including ETS300-113 and the GSM digital cellular standard. In this case, more design effort is needed to control (linearise) the power transfer characteristic of the amplifier.
An additional problem that can complicate constant envelope amplifier design is the need for power control, either to ensure correct operation of a multi user system such as direct sequence spread-spectrum, or to again minimise spectral pollution or extend battery life.
Linear Amplifiers
As the modulation formats become more complex, including multiple amplitude and phase states such as QAM, then the amplifier must exhibit a more linear gain and phase response in order to preserve the integrity of the modulating signals and minimise spectral pollution caused by intermodulation products.
The traditional method of achieving linearity is to bias a transistor into a linear portion of it operating characteristic and not try to extract power outside of this linear region. Class A designs achieve this result, but are hopelessly inefficient and expensive due to the high bias currents needed and the fact that the transistor cannot be driven to its maximum output rating. Efficiency for a class A amplifier is about 25%. In some cases class A/B amplifiers will give sufficient linearity to meet a given specification with efficiencies approaching 40%. Of course 60% off the power is still being dissipated as heat!
Composite (Linearised) Amplifiers
In order to improve amplifier efficiency and reduce cost whilst still achieving the required level of linearity, a range of amplifier linearisation techniques have been developed, some more practical than others. In this section we will look at just one method, Cartesian Loop, which is proving popular in the market place.
Cartesian Loop
One of the most widely used linearisation techniques in the market-place is based on Cartesian feedback which controls the linearity of an entire transmit chain, including the up-converter and amplifier stages. This technique is well suited to integration with a few Cartesian ICs already available in the market-place.
The nature of the feedback process means that the amount of correction afforded by the control system decreases with increasing modulation bandwidth. Typically, intermodulation improvement in the order of 30dB is possible over a 25kHz modulation bandwidth, and more than 10dB improvement has been demonstrated for CDMA-type bandwidths in excess of 1MHz. Efficiencies of Cartesian loop transmitters can be as high as 70%.
Receivers:
Objectives:
The three most important attributes of a good receiver design are:
• Sensitivity
• Selectivity
• Low Current Consumption
Sensitivity: The ability of the receiver up weak signals close to the noise floor. In a good design, the sensitivity should be dominated by the noise figure of the front end.
Selectivity: Rejection of unwanted signals either in band (co-channel interference), close to the wanted band (adjacent channel interference) or well away from the wanted band. The out of band signals will, in general, be much stronger than the wanted signal. Selectivity is influence by many factors including:
• Synthesiser phase noise
• Channel filtering
• Receiver dynamic range
• Receiver linearity
• Demodulation method.
Low Current Consumption: Achieving good dynamic range, linearity and phase noise is often at odds with a low current receiver/synthesiser solutions.
The following considers a number of radio receiver architectures that meet some or all of these objectives.
Tuned Radio Frequency
The simplest of all receivers is the tuned radio frequency (TRF) receiver. Selectivity occurs at the incoming frequency using a tuned circuit (filter) followed by the detector. This method works well for AM with strong received signal strength at low carrier frequencies (e.g. below 1MHz), but sruggles to give adequate selectivity above a few 10’s MHz. It is also difficult to realise sufficient gain at the input frequency without introducing instability. There are however some receiver of this type on the market which use SAW based filtering to achieve selectivity at RF. An example is the Amplifier Sequenced Technique used by RFM. In this implementation, the problem of gain instability is overcome by applying gain at different in different time slots as the signal propagates through the RF chain. To assist with this, the receiver incorporates a SAW delay line between successive amplifier stages, which are pulsed on or off in sequence to achieve the overall system gain.
Super-Regenerative Receivers
The super-regenerative receiver actually makes use of the fact that when an amplifier is close to instability, the gain can be very large giving the potential for a highly sensitive receiver design. The trick with regenerative receivers is to maintain the receive chain on the point of instability. Typically this is achieved by repeatedly quenching the instability as it occurs and then allowing it to ‘grow’ back.
Whilst this approach yields a very low cost and simple receiver solution, it has poor selectivity as for other TRF designs, and is easily blocked by other signals. Further, radiation from the high gain oscillations on which the technique relies can often violate regulatory limits unless very careful design and screening is used.
Superhetrodyne
The main problem with selectivity in the above receiver types can be overcome using a super-heterodyne design. Heterodyne in this context simply refers to the inclusion of a mixer in the chain to convert the incoming high frequency RF signal to a low frequency (usually a fixed intermediate frequency (IF)), where much improved selectivity is possible.
To achieve this mixing function, a local oscillator (LO) running at or near the incoming carrier frequency is needed. The difference between the LO and input signal frequency results in the intermediate frequency (IF) desired. By making the IF a fixed frequency and tuning the LO to select a given channel, the IF selection filter and additional amplification can be carefully optimally for good selectivity with small size and low cost. Typical IF’s used are 455kHz, 10.7MHz and 45MHz. Filters implemented using ceramics, SAW technology or resonant crystals are all available at these frequencies suiting a range of channel spacings.
A further advantage of using an IF is that more gain can be introduced into the receiver chain without instability occurring due to feedback into the input circuits.
Because the input of the receiver is exposed to strong adjacent and out-of-band signals, linearity and intermodulation performance of the front-end amplifier and mixer stages must be carefully considered.
A drawback with the simple superhet design is that the receiver cannot distinguish between signals appearing either above the receiver LO or below the receiver LO. In practice, the unwanted ‘image’ signal on one side of the LO must be filtered in the front end before reaching the mixer.
To alleviate the image problem and aid selectivity, multiple frequency conversion stages are often employed, with a high IF chosen as the first stage to make image rejection easy, and a low IF chosen for the second (and third) stages to make channel selection more precise. Commercial FM receiver IC’s typically incorporate two IF’s at 10.7MHz (45MHz) and 455kHz for this very reason.
In some cases it may be possible to use a frequency band where there are no signals on the image frequency and so do away with the requirement for image reject filtering. There will however be noise in the image band which will be mixed to IF doubling the noise power and reducing the sensitivity by 3dB.
Image Reject Mixing: One further solution to the image problem is an image reject mixer. This device uses careful phase and gain balancing with a pair of mixers to remove the image by vector cancellation rather than filtering. Active image reject mixers are now appearing on the market and are used in some modern cellular handsets. Typical image rejection is however only about 20dB, which is sufficient to eliminate the noise doubling effect, or relax the specification on image reject filters, but not sufficient to suppress a strong signal if present in the image band.
Digital IF Sampling: With recent advances in technology, it is now possible to digitise the waveform at an IF up to 70MHz and then implement further channel filtering and data detection in the digital domain. This brings enormous benefits in flexibility of the receiver unit, but also brings a power consumption penalty with the high sampling clocks required.
Zero IF (Direct Conversion)
A special case of the superhet receiver is the zero IF solution where the LO coincides with the incoming carrier frequency giving an IF of zero Hz. This is also called a direct conversion receiver for obvious reasons.
Because the signal in a zero IF design is mixed down about zero Hz, quadrature versions of the signal must be generated in order to allow the detector to differentiate between in-band components above the LO frequency and those below the LO frequency.
A big attraction of direct conversion is that there are no image problems, and that the IF selection filter becomes a pair of low pass filters at baseband rather than a bandpass filter at say 455kHz or 10.7MHz. This allows the filter to have even greater selectivity with better gain and phase response.
Direct conversion receivers are often used in conjunction with a digital signal processor to implement the channel filtering and data detection. In this case, the two baseband analogue outputs from the mixers are digitized using A/D converters. The advantage of using DSP to implement channel filtering is that near perfect gain and phase response can be realized with very high order (highly selective) filters with variable pass band. This in turn means that the channel spacing can be changed under software control.
There are unfortunately two design challenges with the direct conversion receiver solution. The first is the problem of local oscillator re-radiation from the antenna. Very careful design is needed to ensure that the LO, which is now at the frequency of the incoming signal, does not leak back through the front end mixer/amplifier/filter chain and violate spurious emission regulations. (Most ETSI specifications call for LO radiation levels to be below 2nW (-57dBm).
The second design challenge is the problem of dc-offset within the two baseband signals which, if present, can corrupt wanted information that has been mixed down around zero Hz. Causes of dc offset are either drift in the baseband components (e.g. op amps, filters, A/D converters), or dc from the mixer output caused by the LO mixing with itself or with the mixers acting as square law detectors for strong input signals . Again careful design can minimise this problem, but it can be the reason why direct conversion will not work for every application.
A number of pager designs already use the direct conversion approach, as do some proprietary radio modem products. Most recently, Analog Devices have released a single chip GSM receiver using a direct conversion architecture.
Demodulation:
Having selected the wanted signal from all the other signals in the ether, the next task is to demodulate the information signal from the carrier. This process is often more complex that the corresponding modulation process, partly because the signal level is variable and partly because of high levels of noise or in-band interference. Again, a range of detection methods exist for the various modulation types.
FM Detection
Many FM demodulators (or discriminators) have been developed, with perhaps the most common being the quadrature mixer technique and the phase locked loop detector.
Quadrature Mixer Technique: Here, the signal is limited to remove any amplitude variations due to the transmission path and then split into two paths. One path is fed directly to a mixer port, the other passed through a resonant element to impose a 90 degree phase shift and into the second mixer port. The resultant mixer output contains a dc component that is proportional to the rate of change of phase of the input signal and hence its instantaneous frequency. This type of detector can give a good linear response with frequency making it attractive for high quality reception.
PLL Discriminator: A second widely used technique is the PLL discriminator. Here a PLL is fed with the incoming FM waveform and attempts to force the VCO to track the frequency, (actually phase), changes in the input. Assuming that the VCO has a linear transfer characteristic, the control voltage to the VCO should be a true representation of the modulation signal used to impose the frequency modulation onto the source.
A benefit of the PLL detector is that the filtering action of the feedback control loop can be used to suppress unwanted out of band interference.
Other detectors based on filter slopes and zero-crossing counters can also be used, as well as more sophisticated DSP based detection algorithms.
AM Detection,
Almost any non-linearity will act as an AM detector, and for hearing aid users with GSM phones this is a real problem as they pick up the 200Hz frame pulse. In many cases, a simple diode based detector will suffice, although this has limited dynamic range. Logarithmic envelope detectors are now widely available in IC form, (used extensively for Received Signal Strength Indication (RSSI) purposes) which can extend the dynamic range to 70dB or 80dB.
For optimum sensitivity, coherent (or synchronous) AM detection is needed where the input signal is mixed with a local oscillator that has the exact frequency and phase of the source carrier. The known phase relationship between TX and RX carriers allows the detection process to eliminate half of the noise which in not phase aligned with the carrier, giving a minimum 3dB improvement in sensitivity.
Complex Detection.
More complex detection methods, best suited to DSP implementation, can be used and are discussed in more detail in the context of digital modulation techniques.
Data Processing.
In the majority of low power radio modems on the market there is little or no pre or post processing of the data signal to improve performance. In some cases, e.g. key fobs, it is not needed or desirable as the range requirement is small and cost, size and battery consumption dominate. In other applications such as control of plant and machinery or remote monitoring where improved sensitivity and tolerance to multipath is highly beneficial, then the processing techniques outlined below can be of great help:
Matched Filtering: A matched filter is the term used to define a transmit and receive data modulation and demodulation process that give optimum detection in the presence of noise. For example, with square (unshaped) data pulses used to modulate the carrier, the optimum filter to extract the data from the noise is an ‘integrate and dump’ filter. If however the data is shaped to reduce spectral pollution, perhaps using a Gaussian or Root Raised Cosine Filter, then the integrate and dump detector will not be optimum and a different filter type should be used. Implementing a matched filter system can improve sensitivity by up to 3dB.
Coding: Forward Error Correction (FEC) coding can give even greater improvements in sensitivity of up to 6dB for some modulation formats and coding methods. Although mathematically complex on the surface, the hardware/software implementation of a good coding algorithm is not too difficult and once in place will add very little cost or increase in battery drain.
Channel Equalization: Equalizers’ are used in every modern PC modem to optimise the gain and phase characteristics of the end to end connection at the start of each transmission and therefore achieve maximum reliable data transfer. In radio systems, where the transmission path may be changing during data transfer, the equalization process must be updated at regular intervals which adds complexity but can deliver more robust and higher data rate communication. This type of equalization is used in every GSM digital phone. Equalization of this type is only really practical with DSP implementation, however the processing overhead is small and well within the capability of a $5 DSP device.
There is no denying the fact that Digital Synthesis as the name suggests creates the source frequency from digitally generated samples of the output waveform which are then passed through a D/A converter. With sufficient processing speed and word-length in the device, it is practical to achieve very fine resolution of the output frequency (mill hertz accuracy) with incredibly fast switching times (s).
Current single chip DDS devices from Infineon, Analog Devices, QUALCOMM, etc costing only a few $ can generate frequencies up to 100MHz, using a clock rate of about 300MHz. Their main drawback with this clock rate is that they consume vast amounts of power, (approx.1W!). Further the stability is only as good as the stability of the clock source driving the DDS unit.
By using DDS in conjunction with conventional PLL based synthesis it is possible to get the benefits of each method, but with a cost, power consumption and size penalty that makes them currently much more suited to high power base-station applications than low power telemetry units.
Transmitter & Receiver
Architectures
Transmitter & Receiver Architectures:
Basic Building Blocks
Frequency Source (Carrier)
In order to make the information signal (e.g. data) pass through the air, it must be modulated on to a carrier signal whose frequency is well suited to the propagation environment, conforms to the licensed operating bands, and is sufficiently stable to allow detection by a tuned receiver in the presence of interference.
Modulation
The method of imposing the information signal onto the carrier signal is termed modulation and must be accomplished cost effectively and accurately for maximum range and minimum interference.
Amplification
The amplifier is a key part of the transceiver, and must be efficient (dc power in to RF power out), low cost, non-polluting, and possibly linear. Output power is dictated by regulation, range requirement, battery life, cost and linearity considerations.
Antenna
The antenna is often the most poorly engineered part of a radio system. Good design will ensure maximum range, high amplifier efficiency (good matching), good selectivity, minimal pollution, good interference rejection, good sensitivity, reduced design headaches.
Reception
Key to the sensitivity, dynamic range and strong signal handling properties of the radio is the receiver ‘front end’. The main task is to boost weak wanted signals, often in the presence of strong unwanted signals whilst introducing minimal noise and distortion. In many cases, some selective filtering is required to assist this task.
Demodulation
The process of removing the information signal from the carrier is termed demodulation. The challenge is to design a circuit (or algorithm) that will achieve this task optimally in the presence of noise, interference and varying signal strength, frequency and phase, whilst being compact, power efficient and cheap.
Data Processing
Pre and post processing of the information signal is often an after thought for low power radio applications, usually implying some form of microprocessor or DSP engine with the presumed complexity, cost, power consumption and size penalties. The benefits of matched filtering, error detection and correction (coding), channel equalization, etc, are however significant in terms of range, robust transmission, power conservation and data rate optimization.
The following sections cover each of these building blocks in greater detail.
Frequency Sources:
Requirements
The two most crucial factors affecting the design of a carrier frequency source are:
• Frequency Stability
• Phase Noise
Frequency Stability: The stability of the oscillator with temperature (and ageing) determines the channel spacing required to contain the modulated carrier signal. Conversely, for a given regulated channel spacing, the frequency stability determines the maximum data rate that can be supported without violating the channel boundary (spectral mask).
A UHF transmitter in the 868MHz band having to meet ETS300 220 for a 25kHz channel, may require a stability of better than +/- 3ppm, (better if maximum data rate is to be achieved), whilst for wideband operation at 433MHz, also under ETS300 200, the permissible error is up to +/- 800kHz or 1800ppm.
3ppm source accuracy at 868MHz gives a frequency uncertainty of:
868x106 x 3x10-6 = 2.604kHz
Phase Noise: The oscillator phase noise (phase jitter) results in a broadband component to the carrier signal which will extend into adjacent channels. If the phase noise is too high, this can corrupt the modulation source itself (if PSK or QAM based), and limit adjacent channel selectivity due to reciprocal mixing.
Tuned Resonators – LC, Ceramic, SAW, Crystal
A frequency stable oscillator requires a temperature stable tuned element in the feedback circuit.
A simple LC stabilised oscillator is generally not stable enough and will also suffer badly from microphony.
Using a transmission line in place of the LC circuit will be much less susceptible to microphony and usually exhibit less drift with temperature.
A ceramic resonator, widely used as a clock source for microprocessor applications, has a frequency tolerance of only approx. 3000ppm and hence is not adequate for most radio applications.
Co-axial resonators (often ceramic based) can provide very high Q elements and hence have excellent phase noise. Their temperature stability is however poor and on their own they are not sufficient to act as the carrier source. (Locking a coaxial resonator to a stable (e.g. crystal) source is however a very viable and widely used synthesis process).
Surface Acoustic Wave (SAW) devices can provide the accuracy needed for wideband operation, giving typically +/- 300ppm stability over –40oC to +60oC for frequencies up to 1GHz or beyond. The usually have an initial tolerance of approx. +/- 50kHz which can be tuned on manufacture (at a price). Because they provide a low cost, on-frequency source, SAW based oscillators are very popular for wideband low power radio solutions.
A crystal oscillator is the most widely used stable frequency source, relying on the natural fundamental (or overtone) vibration frequency of the quartz crystal. Depending on the quality of the crystal and the type and tolerance of the cut, accurate, temperature stabilised devices can be fabricated. For a very cheap crystal, a temperature tolerance of +/- 15ppm is typical, whereas for a Temperature Compensated Crystal Oscillator (TCXO), a +/- 3ppm tolerance, or better, is possible. The major drawback with crystal based oscillators compared with SAW devices is that the oscillation frequency is typically below 30MHz for a fundamental mode design, and possible up to 100MHz using a fifth overtone design. (Overtone crystals are more expensive and difficult to control in a low cost circuit design).
Frequency Synthesis:
In order to make use of the stability of a crystal reference, whilst generating a carrier frequency of several hundred MHz, some form of frequency multiplication or synthesis is required. This next section considers a range of options for realising this function.
Harmonic Generation.
Passing a signal through a non-linear circuit will result in harmonics of the waveform being generated. Turning a sinewave carrier into a squarewave will generate a signal rich in odd order harmonics of the fundamental frequency. Using a tuned circuit to extract one of the harmonics can deliver the required output carrier frequency. As the harmonic amplitude falls with order, it is usual to cascade a number of low order harmonic generation components, each giving perhaps x2 or x3 frequency multiplication.
An alternative to using a non-linearity to generate harmonics is the use of linear mixing to create multiples of the fundamental frequency. A perfect balanced multiplier will act as a frequency doubler, and if cascaded can generate any integer value of the fundamental source frequency. This approach to frequency synthesis is both relatively expensive and in-flexible. Today it has been superseded by the fully integrated phase locked loop based synthesiser, (see below).
Injection Locking
Injection locking is a common technique for providing stable microwave oscillators but finds little use in lower frequency applications due to its extreme design/layout sensitivity. The technique is based on the fact that a free running oscillator will (usually) lock onto a nearby low level frequency source (or harmonic of the source) and thus combine the benefits of the free running source – perhaps high output power or good phase noise, with the stability of the low level injected reference.
Phase Locked Loop Synthesizers
The phase locked loop (PLL) forms the basis of most modern carrier generation solutions. Like injection locking, the technique aims to get a high quality (good phase noise), high frequency oscillator to be locked to a stabile, low level, low cost, (often crystal derived), source.
The most simple form of PLL synthesizer contains a voltage controlled oscillator (VCO) operating at the required carrier frequency, (e.g. co-axial resonator or SAW design), a frequency divider (a digital divider circuit), a phase comparison circuit and a loop filter. The circuit acts to lock the reference frequency (fREF) to fVCO/N, resulting in an output frequency of N.fREF .
By adding a further divider between the reference frequency source and the PLL, it is possible to synthesise a whole range of output frequencies with value: N.fREF/M. With the divider ratio N/M under microprocessor control, a flexible channel synthesiser is realised. Today it is possible to buy single or dual synthesisers on a single chip containing all the circuitry required (excluding VCO and loop filter).
Fractional-N Synthesis
A more recent extension of the basic PLL based synthesiser is the fractional-N device. This allows the generation of carrier frequencies that are no longer constrained to multiples of the integer divider ratio N/M without compromising on phase noise of the output waveform. These new devices allow the generation of pure high frequency sources with flexible channel spacing from a 100’s of hertz to 100’s of kilohertz.
The only current drawback with fractional N synthesis is the somewhat poor level of spurious components that can arise at some output frequencies. Careful design can play a big part in achieving a satisfactory solution. Again single or dual fractional N synthesisers are available on chip, (e.g. Motorola, Nat Semi), with a cost of around $3 and operating frequencies up to 3GHz.
Direct Digital Synthesis
Digital synthesis as the name suggests creates the source frequency from digitally generated samples of the output waveform which are then passed through a D/A converter. With sufficient processing speed and word-length in the device, it is practical to achieve very fine resolution of the output frequency (millihertz accuracy) with incredibly fast switching times (s).
Current single chip DDS devices from Infineon, Analog Devices, Qualcom, etc costing only a few $ can generate frequencies up to 100MHz, using a clock rate of about 300MHz. Their main drawback with this clock rate is that they consume vast amounts of power, (approx.1W!). Further the stability is only as good as the stability of the clock source driving the DDS unit.
By using DDS in conjunction with conventional PLL based synthesis it is possible to get the benefits of each method, but with a cost, power consumption and size penalty that makes them currently much more suited to high power base-station applications than low power telemetry units.
Modulation:
The process of imposing the information signal onto the carrier is termed modulation and can be achieved in a wide variety of ways depending on the type of modulation required.
Essentially there are three basic ways of changing the properties of the carrier signal to convey information: Amplitude Modulation, Frequency Modulation and Phase Modulation. Modern low power radio devices use all three of these methods and some of the more recent designs use combinations to achieve improved data rate and range.
A detailed discussion of digital modulation formats and their merits, performance and implementation is covered in ‘Digital Communications – Design for the Real World’ published by Addison Wesley Longman, and in the Digital Modulation section of this course. For now, a brief overview of methods for modulating a carrier, independent of the data type, is given.
FM Modulation
Varactor
The simplest method of varying the frequency of a carrier it to include a voltage controlled impedance element in the oscillator circuit. The Varactor (voltage controlled capacitor) is most widely used, costing only a few pence, and providing reasonably linear variation in capacitance (and hence oscillator frequency) with applied voltage.
PLL Based Modulation
To achieve stability of the modulated source, the PLL is again widely used, with the varactor based oscillator acting as the VCO. Frequency modulation can be achieved in one of three ways:
a) Modulating the reference source (typically using a varactor based crystal oscillator or DDS source), in which case the modulation must not exceed the tracking bandwidth of the loop otherwise the modulation will not be recreated at the output. (Note: the output frequency shift, (modulation index), is N/M times the input frequency shift).
b) Adding the modulation signal to the VCO control voltage within the loop. This only works if the modulation rate exceeds the tracking bandwidth of the loop such that the loop is unable to track it out, which it will attempt to do!
c) A combination of the above.
Vector Modulation
A vector modulator implements a complex mathematical sum, multiplying two input signals with in-phase (e.g. cosine) and quadrature (e.g. sine) versions of the carrier and summing the result. If the input signals are arranged to be cos(1t) and sin( 1t), where 1 is the desired frequency shift to be imposed on the carrier, then the output of the summing junction with be duly shifted by the required amount.
With single chip implementations of a vector (quadrature) modulator now available (RFMD, Analog Devices, HP, Motorola, etc). this approach is gaining support, making the design of the carrier source (synthesiser) much simpler. In addition, the vector modulator is not limited to frequency modulation but can impose any complex modulation format (amplitude, frequency, phase) onto a carrier.
AM Modulation
On-Off Keying
The simplest, (and most spectrally polluting), form of amplitude modulation for data transmission is to turn on and off (gate) the carrier source is sympathy with the data signal. Simply altering the bias or controlling the supply to the oscillator will achieve this. The reason that this approach is often frowned upon is that pulsing a carrier on and off creates a broad modulation spectrum as the harmonics of the square wave modulating signal extend many times beyond the data rate being conveyed.
Mixer Based
Controlling the turn on and turn off rate of the carrier so that the amplitude grows and decays smoothly significantly reduces spectral pollution without degrading the performance of the system. Whilst it is possible to design a circuit that will achieve this for the carrier source itself, it is often simpler and cheaper to use a mixer to impose a pre-shaped data signal onto the fixed amplitude carrier.
Mixer Types
Whilst on the subject of mixers, it is worth looking at the properties of various types of mixer implementation:-
Ideal Mixer: The ideal mixer will perform the mathematical multiplication of the two input signals, creating components positioned at frequencies equal to the sum and difference of the input signals and no additional components. For this to be the case, the mixing (multiplying) device must be perfectly linear and there must be no leakage of the input signals to the output port.
Double Balanced Mixer: The term balanced mixer is used to imply that neither of the input terms will appear at the mixer output. In practice, suppression of these input components is never perfect in an analogue mixer circuit, (it can be virtually perfect in a digital implementation). A balanced mixer can be implemented using a transformer coupled diode arrangement (termed a passive mixer) of the type supplied by Mini Circuits and others, or using an active transistor based design, the most well know of which was invented by Barry Gilbert of Analog Devices, and called the Gilbert cell mixer.
Both types of mixer produce signals at odd harmonics of the carrier frequency, particularly the diode ring mixer. In most cases these can be easily filtered out.
Single Balanced Mixer: The difference between a single and double balanced mixer in practical terms is that the output will contain both even and odd harmonics. With 2nd order harmonic being the most difficult to suppress, double balanced mixers are the favoured solution.
Phase Modulation
Continuous phase modulation can be achieved using a frequency modulator with an integrator imposed between the modulating signal and the modulation circuit. Where discrete phase changes are required for conveying digital information, it is often much simpler to switch phases of the carrier source or use vector modulation – see next section.
Complex Modulation
To realise a combination of amplitude, frequency or phase modulation it is possible to use combinations of the techniques described above. The most elegant solution, however, (and today arguably the smallest and cheapest), is the vector modulator.
Vector Modulator
By applying the correct input to the two arms of the vector modulator, it is possible to impose any amplitude, frequency or phase shift directly onto a carrier in real time. If the carrier feeding the vector modulator is at the final output frequency for the radio, then the modulation is imposed directly onto the output signal in what is termed a direct conversion transmitter.
The function of the vector modulator can be realised in analogue or digital form. The digital form, whilst nearly perfect, has an upper frequency limit today of approx. 100MHz. The analogue form can be used with any frequency, (single chip devices now extend beyond 3GHz), but have imperfections in the mixing and summing process. These imperfections in practice result in an image of the modulation signal appearing at the output of the modulating circuit and residual carrier leakage.
In some cases it is possible to compensate for these imperfections by pre-distorting the input signal.
Combined Modulator/Amplifier
We shall see in the amplifier section to follow that it may in fact be advantageous, in terms of power efficiency, to split a complex modulating signal into its separate amplitude and phase/frequency components and independently impose these variations onto the carrier as part of the amplification process.
Amplification:
Objectives
The key design objectives for an RF amplifier are to boost the output signal level into the antenna with minimum cost, size, distortion, power dissipation, and spectral pollution. Optimising all of these parameters is a major design challenge and one that becomes increasingly difficult as the complexity of the modulation format is raised. The following gives a brief discussion of amplifier solutions to meet these challenges for different modulation types.
Constant Envelope Amplifiers
The simplest modulation formats to amplify are frequency modulation (FSK) or On-off keying. In both cases, when the amplifier needs to deliver power, it is at a constant level and hence at a fixed operating point on the amplifier power transfer characteristic. This means that the amplifier only needs to be optimised to deliver power efficiently at that output level and for a fixed input level. The linearity of the transfer characteristic is thus unimportant. As a result, most simple FM units use Class C biased amplifier designs with efficiencies of between 50% and 70%.
The position is complicated a bit when the carrier needs to be pulsed on and off, (i.e. with On-off keying or Time Division Multiplexed Operation), and the rise and fall times of the pulse edges need to be carefully controlled to minimise spectral pollution. This requirement exists for some of the more stringent ETSI standards including ETS300-113 and the GSM digital cellular standard. In this case, more design effort is needed to control (linearise) the power transfer characteristic of the amplifier.
An additional problem that can complicate constant envelope amplifier design is the need for power control, either to ensure correct operation of a multi user system such as direct sequence spread-spectrum, or to again minimise spectral pollution or extend battery life.
Linear Amplifiers
As the modulation formats become more complex, including multiple amplitude and phase states such as QAM, then the amplifier must exhibit a more linear gain and phase response in order to preserve the integrity of the modulating signals and minimise spectral pollution caused by intermodulation products.
The traditional method of achieving linearity is to bias a transistor into a linear portion of it operating characteristic and not try to extract power outside of this linear region. Class A designs achieve this result, but are hopelessly inefficient and expensive due to the high bias currents needed and the fact that the transistor cannot be driven to its maximum output rating. Efficiency for a class A amplifier is about 25%. In some cases class A/B amplifiers will give sufficient linearity to meet a given specification with efficiencies approaching 40%. Of course 60% off the power is still being dissipated as heat!
Composite (Linearised) Amplifiers
In order to improve amplifier efficiency and reduce cost whilst still achieving the required level of linearity, a range of amplifier linearisation techniques have been developed, some more practical than others. In this section we will look at just one method, Cartesian Loop, which is proving popular in the market place.
Cartesian Loop
One of the most widely used linearisation techniques in the market-place is based on Cartesian feedback which controls the linearity of an entire transmit chain, including the up-converter and amplifier stages. This technique is well suited to integration with a few Cartesian ICs already available in the market-place.
The nature of the feedback process means that the amount of correction afforded by the control system decreases with increasing modulation bandwidth. Typically, intermodulation improvement in the order of 30dB is possible over a 25kHz modulation bandwidth, and more than 10dB improvement has been demonstrated for CDMA-type bandwidths in excess of 1MHz. Efficiencies of Cartesian loop transmitters can be as high as 70%.
Receivers:
Objectives:
The three most important attributes of a good receiver design are:
• Sensitivity
• Selectivity
• Low Current Consumption
Sensitivity: The ability of the receiver up weak signals close to the noise floor. In a good design, the sensitivity should be dominated by the noise figure of the front end.
Selectivity: Rejection of unwanted signals either in band (co-channel interference), close to the wanted band (adjacent channel interference) or well away from the wanted band. The out of band signals will, in general, be much stronger than the wanted signal. Selectivity is influence by many factors including:
• Synthesiser phase noise
• Channel filtering
• Receiver dynamic range
• Receiver linearity
• Demodulation method.
Low Current Consumption: Achieving good dynamic range, linearity and phase noise is often at odds with a low current receiver/synthesiser solutions.
The following considers a number of radio receiver architectures that meet some or all of these objectives.
Tuned Radio Frequency
The simplest of all receivers is the tuned radio frequency (TRF) receiver. Selectivity occurs at the incoming frequency using a tuned circuit (filter) followed by the detector. This method works well for AM with strong received signal strength at low carrier frequencies (e.g. below 1MHz), but sruggles to give adequate selectivity above a few 10’s MHz. It is also difficult to realise sufficient gain at the input frequency without introducing instability. There are however some receiver of this type on the market which use SAW based filtering to achieve selectivity at RF. An example is the Amplifier Sequenced Technique used by RFM. In this implementation, the problem of gain instability is overcome by applying gain at different in different time slots as the signal propagates through the RF chain. To assist with this, the receiver incorporates a SAW delay line between successive amplifier stages, which are pulsed on or off in sequence to achieve the overall system gain.
Super-Regenerative Receivers
The super-regenerative receiver actually makes use of the fact that when an amplifier is close to instability, the gain can be very large giving the potential for a highly sensitive receiver design. The trick with regenerative receivers is to maintain the receive chain on the point of instability. Typically this is achieved by repeatedly quenching the instability as it occurs and then allowing it to ‘grow’ back.
Whilst this approach yields a very low cost and simple receiver solution, it has poor selectivity as for other TRF designs, and is easily blocked by other signals. Further, radiation from the high gain oscillations on which the technique relies can often violate regulatory limits unless very careful design and screening is used.
Superhetrodyne
The main problem with selectivity in the above receiver types can be overcome using a super-heterodyne design. Heterodyne in this context simply refers to the inclusion of a mixer in the chain to convert the incoming high frequency RF signal to a low frequency (usually a fixed intermediate frequency (IF)), where much improved selectivity is possible.
To achieve this mixing function, a local oscillator (LO) running at or near the incoming carrier frequency is needed. The difference between the LO and input signal frequency results in the intermediate frequency (IF) desired. By making the IF a fixed frequency and tuning the LO to select a given channel, the IF selection filter and additional amplification can be carefully optimally for good selectivity with small size and low cost. Typical IF’s used are 455kHz, 10.7MHz and 45MHz. Filters implemented using ceramics, SAW technology or resonant crystals are all available at these frequencies suiting a range of channel spacings.
A further advantage of using an IF is that more gain can be introduced into the receiver chain without instability occurring due to feedback into the input circuits.
Because the input of the receiver is exposed to strong adjacent and out-of-band signals, linearity and intermodulation performance of the front-end amplifier and mixer stages must be carefully considered.
A drawback with the simple superhet design is that the receiver cannot distinguish between signals appearing either above the receiver LO or below the receiver LO. In practice, the unwanted ‘image’ signal on one side of the LO must be filtered in the front end before reaching the mixer.
To alleviate the image problem and aid selectivity, multiple frequency conversion stages are often employed, with a high IF chosen as the first stage to make image rejection easy, and a low IF chosen for the second (and third) stages to make channel selection more precise. Commercial FM receiver IC’s typically incorporate two IF’s at 10.7MHz (45MHz) and 455kHz for this very reason.
In some cases it may be possible to use a frequency band where there are no signals on the image frequency and so do away with the requirement for image reject filtering. There will however be noise in the image band which will be mixed to IF doubling the noise power and reducing the sensitivity by 3dB.
Image Reject Mixing: One further solution to the image problem is an image reject mixer. This device uses careful phase and gain balancing with a pair of mixers to remove the image by vector cancellation rather than filtering. Active image reject mixers are now appearing on the market and are used in some modern cellular handsets. Typical image rejection is however only about 20dB, which is sufficient to eliminate the noise doubling effect, or relax the specification on image reject filters, but not sufficient to suppress a strong signal if present in the image band.
Digital IF Sampling: With recent advances in technology, it is now possible to digitise the waveform at an IF up to 70MHz and then implement further channel filtering and data detection in the digital domain. This brings enormous benefits in flexibility of the receiver unit, but also brings a power consumption penalty with the high sampling clocks required.
Zero IF (Direct Conversion)
A special case of the superhet receiver is the zero IF solution where the LO coincides with the incoming carrier frequency giving an IF of zero Hz. This is also called a direct conversion receiver for obvious reasons.
Because the signal in a zero IF design is mixed down about zero Hz, quadrature versions of the signal must be generated in order to allow the detector to differentiate between in-band components above the LO frequency and those below the LO frequency.
A big attraction of direct conversion is that there are no image problems, and that the IF selection filter becomes a pair of low pass filters at baseband rather than a bandpass filter at say 455kHz or 10.7MHz. This allows the filter to have even greater selectivity with better gain and phase response.
Direct conversion receivers are often used in conjunction with a digital signal processor to implement the channel filtering and data detection. In this case, the two baseband analogue outputs from the mixers are digitized using A/D converters. The advantage of using DSP to implement channel filtering is that near perfect gain and phase response can be realized with very high order (highly selective) filters with variable pass band. This in turn means that the channel spacing can be changed under software control.
There are unfortunately two design challenges with the direct conversion receiver solution. The first is the problem of local oscillator re-radiation from the antenna. Very careful design is needed to ensure that the LO, which is now at the frequency of the incoming signal, does not leak back through the front end mixer/amplifier/filter chain and violate spurious emission regulations. (Most ETSI specifications call for LO radiation levels to be below 2nW (-57dBm).
The second design challenge is the problem of dc-offset within the two baseband signals which, if present, can corrupt wanted information that has been mixed down around zero Hz. Causes of dc offset are either drift in the baseband components (e.g. op amps, filters, A/D converters), or dc from the mixer output caused by the LO mixing with itself or with the mixers acting as square law detectors for strong input signals . Again careful design can minimise this problem, but it can be the reason why direct conversion will not work for every application.
A number of pager designs already use the direct conversion approach, as do some proprietary radio modem products. Most recently, Analog Devices have released a single chip GSM receiver using a direct conversion architecture.
Demodulation:
Having selected the wanted signal from all the other signals in the ether, the next task is to demodulate the information signal from the carrier. This process is often more complex that the corresponding modulation process, partly because the signal level is variable and partly because of high levels of noise or in-band interference. Again, a range of detection methods exist for the various modulation types.
FM Detection
Many FM demodulators (or discriminators) have been developed, with perhaps the most common being the quadrature mixer technique and the phase locked loop detector.
Quadrature Mixer Technique: Here, the signal is limited to remove any amplitude variations due to the transmission path and then split into two paths. One path is fed directly to a mixer port, the other passed through a resonant element to impose a 90 degree phase shift and into the second mixer port. The resultant mixer output contains a dc component that is proportional to the rate of change of phase of the input signal and hence its instantaneous frequency. This type of detector can give a good linear response with frequency making it attractive for high quality reception.
PLL Discriminator: A second widely used technique is the PLL discriminator. Here a PLL is fed with the incoming FM waveform and attempts to force the VCO to track the frequency, (actually phase), changes in the input. Assuming that the VCO has a linear transfer characteristic, the control voltage to the VCO should be a true representation of the modulation signal used to impose the frequency modulation onto the source.
A benefit of the PLL detector is that the filtering action of the feedback control loop can be used to suppress unwanted out of band interference.
Other detectors based on filter slopes and zero-crossing counters can also be used, as well as more sophisticated DSP based detection algorithms.
AM Detection,
Almost any non-linearity will act as an AM detector, and for hearing aid users with GSM phones this is a real problem as they pick up the 200Hz frame pulse. In many cases, a simple diode based detector will suffice, although this has limited dynamic range. Logarithmic envelope detectors are now widely available in IC form, (used extensively for Received Signal Strength Indication (RSSI) purposes) which can extend the dynamic range to 70dB or 80dB.
For optimum sensitivity, coherent (or synchronous) AM detection is needed where the input signal is mixed with a local oscillator that has the exact frequency and phase of the source carrier. The known phase relationship between TX and RX carriers allows the detection process to eliminate half of the noise which in not phase aligned with the carrier, giving a minimum 3dB improvement in sensitivity.
Complex Detection.
More complex detection methods, best suited to DSP implementation, can be used and are discussed in more detail in the context of digital modulation techniques.
Data Processing.
In the majority of low power radio modems on the market there is little or no pre or post processing of the data signal to improve performance. In some cases, e.g. key fobs, it is not needed or desirable as the range requirement is small and cost, size and battery consumption dominate. In other applications such as control of plant and machinery or remote monitoring where improved sensitivity and tolerance to multipath is highly beneficial, then the processing techniques outlined below can be of great help:
Matched Filtering: A matched filter is the term used to define a transmit and receive data modulation and demodulation process that give optimum detection in the presence of noise. For example, with square (unshaped) data pulses used to modulate the carrier, the optimum filter to extract the data from the noise is an ‘integrate and dump’ filter. If however the data is shaped to reduce spectral pollution, perhaps using a Gaussian or Root Raised Cosine Filter, then the integrate and dump detector will not be optimum and a different filter type should be used. Implementing a matched filter system can improve sensitivity by up to 3dB.
Coding: Forward Error Correction (FEC) coding can give even greater improvements in sensitivity of up to 6dB for some modulation formats and coding methods. Although mathematically complex on the surface, the hardware/software implementation of a good coding algorithm is not too difficult and once in place will add very little cost or increase in battery drain.
Channel Equalization: Equalizers’ are used in every modern PC modem to optimise the gain and phase characteristics of the end to end connection at the start of each transmission and therefore achieve maximum reliable data transfer. In radio systems, where the transmission path may be changing during data transfer, the equalization process must be updated at regular intervals which adds complexity but can deliver more robust and higher data rate communication. This type of equalization is used in every GSM digital phone. Equalization of this type is only really practical with DSP implementation, however the processing overhead is small and well within the capability of a $5 DSP device.
Comments
Post a Comment