Electronic Systems

Data converters.

First, we should define analog and digital signals. An analog signal is any continuous signal that
evolve in the time analogously to some physical variable. These variables can be presented in
the form of a current, a voltage or an electric charge. They vary continuously between a lower
limit and upper limit. When these limits coincide with the limits admitted by a given device,
the signal is normalized, which means that the signal/noise(S/N) relation is better.

On the other hand, we have digital signals. They represent data as a sequence of discrete
values. At any given time, it can only take on one of a finite number of values. Each level
represents 1 or 0. They are represented by two voltage bands, for example zero volts and five
volts.

There are two types of data conversion: Analog to digital conversion and digital to analog
conversion.

Analog-Digital Converter (ADC)

Analog-Digital Converter is an electronic integrated circuit which converts a continuous-time

and continuous-amplitude analog signal to a discrete-time and discrete-amplitude digital
signal. This conversion involves entail quantization of the input, approaching the input values
to a finite quantity of levels. The quantization process will be like the rounding or truncation of
a number of infinite decimal places, so it necessarily introduces a small amount of error.
Quantization error is the noise introduced by quantization in an ideal analog-digital converter.
It is a rounding error between the exact value of the analog signal and the nearest quantization
level. In the next picture we can see an example of quantization error, where in red is the
analog signal, in black is a 4-bit digital signal and in green we can see the quantization noise. As
conclusion, if we have more bits the quantization noise is smaller.

Many times signals are sampled at the minimum rate required, and the quantization noise
introduced is white noise spread over the whole pass band of the converter. If economy is not
a problem, is a good idea to consider to sample at a rate much higher than Nyquist rate,
because of anti-aliasing and reduction of noise, and this is oversampling.

The performance of an ADC is primarily characterized by its bandwidth and signal to noise
ratio. The signal to noise ratio (S/N) compares the level of a desired signal to the level of
background noise, so it’s the relation between the average power of the input signal and the
average power of the background noise.

The total harmonic distortion value (THD), is the RMS value of the harmonics produced by the
analog-digital converter relative to the RMS level of a sinusoidal input signal near full-scale.
The non-linearity in the converter will produce harmonics that were not present in the original
signal. These usually degrades the performance of the ADC. Therefore the THD should have a
minimum value for less distortion. As the input signal amplitude or frequency increase, the
distortion increases.

Digital-Analog Converter (DAC)

Digital-Analog Converter is an electronic integrated circuit which converts a discrete-time and

discrete-amplitude digital signal to a continuous-time and continuous-amplitude analog signal.
Digital-Analog Converters are the interface between abstract digital world and the real life
(analog). With just simple switches, a network of resistors, current sources or capacitors we
can implement this conversion.

Differential and integral nonlinearity

Differential nonlinearity is defined as the maximum and minimum difference in the step width
between the actual transfer function and the perfect transfer function. Suppose that we have
an ideal ADC. On this case, the step widths should be 1 LSB (Least significant byte). If we don’t
have an ideal ADC, steps won’t be 1 LSB. As we can see on this picture, the width of the steps
are not equals, and we have for example one 1.5LSB step, and one 0.5 LSB step. This means
that the differential nonlinearity of this case is ±0.5 LSB.

Integral nonlinearity is the maximum vertical difference between the actual and the ideal
curve. It indicates the amount of deviation of the actual curve from the ideal transfer curve.
Integral nonlinearity can be interpreted as a sum of differential nonlinearity. The INL can be
measured by connecting the midpoints of all output steps of actual ADC and finding the
maximum deviation from the ideal curve in terms of LSBs. For example, in the next figure, we
can see that the max INL is +0.75 LSB. This is the end-point method.
Types of ADC

Dual Slope A/D Converter

The fundamental components of this type of ADC are: Integrator, Electronically Controlled
Switches, Counter, Clock, Control Logic and Comparator.

A Dual-slope ADC integrates an unknown input voltage 𝑉𝐼𝑁 for a fixed amount of time 𝑇𝐼𝑁𝑇 ,
then de-integrates 𝑇𝐷𝐸𝐼𝑁𝑇 using a known reference voltage 𝑉𝑅𝐸𝐹 for a variable amount of time.
Pros: conversion result is insensitive to errors in the component values, there are fewer
adverse effects from noise and high accuracy.

Cons: this circuit is slow, accuracy is dependent on the use of precision external component
and is expensive.

Successive approximation A/D Converter

Flash A/D Converter

This circuit has 2𝑛 − 1 comparators, 2𝑛 resistors and control logic as main components.

Uses the 2𝑛 resistors to form a ladder voltage divider, which divides the reference voltage into
2𝑛 equal intervals. Therefore, uses the 2𝑛 − 1 comparators to determine in which of these
2𝑛 voltage intervals the input voltage 𝑉𝑖𝑛 lies. The combinational logic then translates the
information provided by the output of the comparators.

Pros: Very fast and very simple operational theory and speed is only limited by gate and
comparator propagation delay.

Cons: This circuit is expensive, prone to produce glitches in the output and each addicional bit
of resolution requires twice the comparators.
Delta-Sigma A/D Converter

This circuit has resistors, capacitor, comparators, control logic and DAC as main components

In this case, input is oversampled, and goes to the integrator. This integration is compared to
ground and iterates and produces a serial bit stream. The output is a serial bit stream with # of
1´s proportional to 𝑉𝑖𝑛 .

With this arrangement the sigma-delta modulator automatically adjust its output to ensure
that the average error at the quantizer output is zero. The integrator value is the sum of all
past values, so whenever there is a non-zero error value the integrator value just keeps
building until the error is once again forced to zero.

Pros: High Resolution and no need for precision components.

Cons: Slow due to oversampling and is only good for low bandwidth.
Uses a n-bit DAC to compare DAC and original analog results. Uses successive approximation
register supplies an approximate digital code to DAC of 𝑉𝑖𝑛 . Comparator changes digital output
to bring it closer to the input value.

The process is easy. First of all MSB initialized as 1, then, convert digital value to analog using
DAC and compares guess to analog input. If 𝑉𝑖𝑛 > 𝑉𝐷𝐴𝐶 then set bit 1, if not, set bit 0.

Pros: Capable of high speed and reliable, have a decent accuracy compared to other ADC
types, good tradeoff between speed and cost and capable of outputting the binary number in
serial format.

Cons: With high resolution this system will be slower, and the speed is limited to 5Msps.

Types of DCA

Kelvin Divider

This is the simplest structure. It consists of 2𝑛 equal resistors in series and 2𝑛 switches (usually
CMOS). The output is taken from the appropriate tap by closing just one of the switches.
Thanks to since only two switches operate during a transition, it is a low glitch architecture.

Thermometer (Fully Decoded)

There is a current-output DAC architecture which consists of 2𝑛 − 1 switchable current
sources connected to an output terminal. This output must be close to ground. In the case on
the picture, we have thermometer DAC which use resistors connected to a reference voltage
to generate the currents.

R-2R DAC

The R-2R resistor ladder network directly converts a parallel digital symbol into an analog
voltage. Some of the advantages that this circuit have are: Easily scalable to any desired
number of bits, Uses only two values of resistors which make for easy and accurate fabrication
and integration and the output impedance is equal to R, regardless of the number of bits,
simplifying filtering and further analog signal processing circuit design

We can consider the R-2R DAC is as a binary weighted voltage divider. The 2R leg in parallel
with each R resistor in series creates this binary weight. As a result, we only need one switch
for each bit of resolution. The switch is either connected to ground or to the reference voltage.
 Oscillators
Basics of oscillators (Harmonic Oscillators & Relaxation oscillators)

An oscillator is a circuit that convert direct current into alternating current. The resultant signal
is usually a sine wave or a square wave. Depends of the oscillating frequency we have four
main types:

- Low frequency oscillator (LFO): Is usually below 20 Hz. Pulses with this frequency are
useful for music, for example, electronic musicians use LFO to add simple vibrato or
tremolo to a melody or triggering gate envelopes or controlling the rate of
arpeggiation.
- Audio oscillator: produces frequencies in audio range, from 16Hz to 20kHz.
- Radio frequency oscillators: produces signals in the radio frequency range from 100kHz
to 100 GHz.
- Oscillators designed to produce a high-power AC output from DC supply (inverters).

In the next pages we will talk about two main types of electronic oscillators, the linear and
nonlinear oscillators.

Harmonic oscillators (linear ones)

Harmonic oscillators are so called because the type of signal they discharge is a simple sine
wave. This is because the charge is transferred gradually and then removed gradually as
well which make this kind of oscillators ideal for machines that require the kind of steady
movement.

Feedback oscillator:

This kind of oscillator is made from an amplifier (could be a transistor or operational

amplifier) and a feedback network. The feedback network has the purpose of send some of
the system’s output back to be reamplified after a suitable time delay. When the power
supply is first switched on, it produces a noise that travels around the loop and is amplified
and filtered until it converges on a sine wave at a single frequency.

According to the type of frequency selective filter they use in the feedback loop we have:

RC Oscillator:
This linear oscillator circuit uses a RC network (a combination or resistors and capacitors).
This circuits have the advantages of their low cost, they are easier to integrate into
microelectronic devices and good to work with low frequencies where the coils are not
practical. On the other hand, we have that the coils are more stables, this is why the LC
oscillators are more useful for a bigger frequency.

The frequency oscillation is determined by the Barkhausen criterion, which says that the
circuit will only oscillate at frequencies for which the phase shift around the feedback loop
is equal to 360°, and the loop gain is equal to one. This is: 𝐴𝛽(𝑗𝜔0 ) = 1. In RC oscillator
circuits which use a single inverting amplifying device (operational amplifier or transistor)
with the feedback applied to the invertin input, the amplifier provides a 180° of the phase
shift, so the other 180° must be provided by the RC network. This is the reason that RC
oscillators require at least two frequency-determining capacitors in the circuit (two poles),
because an individual capacitor provide a maximum of 90° of phase shift.

So, in order to have a sinusoidal 𝑉𝑜 :

- The feedback amplifier must be unstable to only one frequency (𝜔0 )

- The loop gain 𝐴𝛽must be 1 for this frequency
- Could depend on 𝜔 𝐴 𝑜𝑟 𝛽, or both (in general)
- Both 𝐴 𝑎𝑛𝑑 𝛽 are gain values with loading effects.

To ensure that the oscillation begins, it is necessary to comply condition of excess gain, this
is the condition of start. This is why with 𝐴𝛽 = 1 the oscillations keep constant, if they
exist, but they do not grow or are created. With 𝐴𝛽 > 1 the oscillations will increase
indefinitely.

Common RC oscillator circuits are listed below:

Phase-shift oscillator

In this case, the feedback network is three identical cascaded RC sections. If we suppose
that R= R1=R2=R3, and C= C1=C2=C3, each RC section contributes a 60° phase shift, so the
feedback network (𝛽) fixes a phase shift or 180° in 𝜔0 . The oscillation frequency will be
 1
𝜔0 =
𝑅𝐶√6
And the gain condition will be:
𝑉
𝑘 = 29( )
𝑉
This means that the feedback network has 1/29 attenuation so:

𝑅𝑏 = 29𝑅
Sometimes amplitude limiters are needed, the function of the limiters are:

- Try to avoid the saturation of the active device.

- Non-linear devices are introduced (diodes, zeners, etc).
- The most efficient way is by reducing the gain depending on the amplitude of the output
wave.

Quadrature oscillator https://prezi.com/z4tqtkwdlktp/oscilador-de-cuadratura/

The quadrature oscillator uses two cascaded operational amplifier integrators in a feedback
loop. Supplies two sinusoidal signals with 90° offset, so we have a sine and cosine.

Twin-T Oscillator

The twin-T oscillator uses two “T” RC circuits operated in parallel. One circuit is an R-C-R
which acts as low-pass filter. The second circuit is a C-R-C which operates as a high-pass
filter. C-R-C part is advanced and R-C-R is delayed, so they may cancel one another for
1
𝑓=
2𝜋𝑅𝐶
Wien bridge oscillator

This is one of the most common gain-stabilized circuits. In this circuit we two RC circuits
where one of them is in series and the other one is in parallel. The frequency of the
oscillation is given by:
1
𝑓=
2𝜋𝑅𝐶

LC Oscillator:

This kind of oscillator is formed by a coil and a capacitor in parallel. Its operation is based
on the storage of energy in the form of electric charge in the condenser and in the form of
a magnetic field in the coil. The capacitor, in a time equal to zero, offers an impedance
close to zero ohms, so it allows a large current to flor through it, which decreases until its
plates are filled with as many positive and negative charges as the plates size allows. On
this instant, the capacitor works as insulator and do not permit more current flow and
create an electric field between both plates. On the other hand, in a time equal to zero the
coil has an almost infinite impedance, so it does not allow the current flow though it, and
when the time advances, the current begins to flow, creating a magnetic field.

Due to the capacitor and the coil are in parallel, this creates a swing of the current
between the capacitor and the coil. This oscillation constitutes an electromagnetic
oscillation.

The speed which the current flows from the capacitor to the coil, occurs with its own
frequency, called resonance frequency, which depends on the values of the capacitor and
the coil:
1
𝑓= 𝐻𝑧
2𝜋√𝐿𝐶
Colpitts Oscillator

Colpitts oscillators uses an LC circuit in the feedback loop to provide the necessary phase
shift and to act as a resonant filter that passes only the desired frequency of oscillation.
The main characteristic is that the feedback for the active device is taken from a voltage
divider made of two capacitors in series across the inductor.
𝐶1 𝐶2
The equivalent capacitor in the feedback circuit is 𝐶𝑇 = 𝐶1 +𝐶2

1
𝑓≅
2𝜋√𝐿𝐶𝑇
The Clapp oscillator

This circuit is a variation of the Colpitts one because adds a third capacitor in the feedback
circuit. Since C1 and C2 are both connected to ground at one end, the junction capacitance
of the transistor and other stray capacitances appear in parallel with C1 and C2 to ground,
altering their effective values. However, C3 is not affected, and thus provides a more
1
accurate and stable frequency of oscillation. In this case, 𝐶𝑇 = 1 1 1 and if C3 is much
+ +
𝐶1 𝐶2 𝐶3
1
smaller than C1 and C2, 𝑓 ≅
2𝜋√𝐿𝐶3
Hartley Oscillator

The Hartley oscillator is like the Colpitts except that the feedback circuit consists of two
series inductors and parallel capacitor. Where 𝐿 𝑇 = 𝐿1 + 𝐿2
1
𝑓≅
2𝜋√𝐿 𝑇 𝐶

As we can see, the feedback circuit has the same effect in the Hartley as in the Colpitts.

Crystal oscillator

An electronic oscillator that uses the mechanical resonance of a vibrating crystal of

piezoelectric material to create an electrical signal with a precise frequency. As we saw
before, a circuit will oscillate if it has positive feedback and a loop gain greater than one.
With crystal as a series element in the loop and no other frequency-sensitive elements in
the circuit, it will oscillate at the crystal’s fundamental series-resonant frequency. Its
properties are very stable over time and insensitive to changes in temperature or
humidity. However, when they are used for high-precision reference oscillators, they are
enclosed in a temperature-controlled box.

Here we can see the crystal symbol (a) and the equivalent circuit (b):
The capacitor 𝐶0 corresponds to a capacitor whose dielectric is the quartz crystal. The rest
of the elements do not have physical support, they only model the properties of the
crystal. Each RLC circuit resonates to a tone, the first is the fundamental and the rest its
harmonics. The value of the fundamental frequency depends on the physical dimensions
of the crystal and the orientation of its cut with respect to the crystal lattice. A tuning-fork
crystal is usually cut such that its frequency dependence on temperature is quadratic with
the maximum around 25 °C . A common parabolic coefficient for a 32kHz tuning-fork
crystal is -0.04ppm/°C ²

0.04𝑝𝑝𝑚2
𝑓 = 𝑓0 [1 − (T − 𝑇0 )2 ]
°C
Some factors affecting frequency stability are:

Oscillator tuning port noise: Tuning sensitivity is a system-level parameter that relates the
maximum available tuning voltage to the required tuning-frequency range, in units of
Hz/volt. If the tuning sensitivity of a voltage-controlled crystal oscillator varies dramatically
over the tuning band, the performance of oscillator in a phase locked loop will be worse as
the noise bandwidth of the control loop will vary as a function of voltage controlled crystal
oscillator frequency. The tuning sensitivity can change in response to noise at the tuning
port. So tuning-port noise must be minimized.

Power supply noise: Such noise typically appears as steps or impulses on the power supply
of the oscillator, and it affects both frequency and phase, causing cycle-to-cycle jitter.

Vibration-induced noise: It is caused by the sensitivity to acceleration of crystals. So the

random and periodic mechanical vibrations found in many types of equipment and
instruments can induce significant phase noise in high-performance crystal oscillators. It
can be classifies to sine vibration-induced phase noise and random vibration induced
phase noise.

There are two ways to use crystal to build an oscillator, in series and in parallel. In series
the circuit oscillates when the crystal behaves as a short circuit, at ωs. It takes an LC circuit
to determine the harmonic in which it will oscillate. In parallel mode the crystal replaces
the coil, in ωs <ω <ωa. (a) Serie mode and (b) Parallel mode examples.
 Relaxation oscillators (nonlinear ones)

In electronics a relaxation oscillator is a nonlinear oscillator circuit that produces a

nonsinusoidal repetitive signal, like a triangle wave or square wave. The circuit consists of a
feedback loop containing a switching device like a transistor, comparator, operational
amplifier, etc. Square-ware relaxation oscillators are used to provide the clock signal for
sequential logic circuit like timers and counters, however, crystal oscillators are often
preferred because of their greater stability. Triangle-wave oscillators are used in voltage-
controlled oscillators, inverters and switching power supplies and some analog to digital
converters. Its important to know that relaxation oscillators are used at lower frequencies and
have poorer frequency stability than linear oscillators.

We have two main types of relaxation oscillators. On the one hand we have flyback oscillator,
where the energy storage capacitor is charged slowly but discharged rapidly. On the other
hand we have astable multivibrator, where the capacitor is both charged and discharged
slowly through a resistor, so we have increasing and decreasing ramp as output.

The most common relaxation oscillator circuits are: Multivibrator, Pearson-Anson oscillator,
Ring oscillator, Delay-line oscillator and Royer oscillator.

Phase noise on oscillators

Maybe this parameter is the most important one in many oscillators and it deserves an in-
depth discussion on what it is, how it affects a system and how it can be minimized in an
oscillator design. That noise increases at frequencies close to the oscillation frequency or its
harmonics. With the noise being close to the oscillation frequency, its hard to remove by
filtering without also removing the oscillation signal.
 Signal synthesizers

As we saw in LC Oscillators and crystal oscillators, they have advantages and disadvantages. In
the first one the advantage is the possibility of variability in the output frequency, the
disadvantage is the poor stability in frequency; in the second the advantage is that is stable in
frequency and the disadvantage is the non-variability of the output frequency. One of the
purpose of this kinds of circuits is to develop the synthesized oscillators or frequency
synthesizers.

Frequency synthesis consist in the generation of a variable frequency signal, using two or more
interconnected oscillators in a convenient way. We obtain a discrete variation of frequency,
being able to make the steps as small as we desire, the synthesizer must be capable of
producing as many frequencies as possible. We have two kind of synthesizers: Direct
synthesizers and indirect synthesizers.

Direct synthesizers: It is a system that generates the output frequencies based on the four
fundamental arithmetic operations, using mixer circuits. Direct synthesizer has the advantage
that if the base oscillator is a very stable one, the output frequencies will be stables too.
However, is not used at present because of its great complexity, because need many
operations, including very complex filters, resulting in a high cost.

Indirect synthesizers (PLL):

This kind of circuits are so important in telecommunication systems in receiving and

transmitting stages, either for analog or digital modulation. Nowadays we have many
integrated circuits and modules with small size. High reliability and low cost. This circuit allow
us, with an internally generated signal (reference), to control a loop (PLL) and obtain in the
output a signal whose stability in frequency depends on the stability of the control or
reference signal. This reference signal use to be form crystal oscillators, because of their great
frequency stability. It also allows us to obtain a discrete variation of the output frequency
where the range and resolution depend on the divider network and the value of the reference
frequency that enters in the phase comparator.

As we saw PLLs are circuits that synchronize the signal of a VCO in frequency and in phase with
a reference signal. The simplest architecture of the PLL’s consist in two clocks, a phase
detector and a voltage-controlled oscillator (VCO). The phase detector is responsible of
comparing the frequency of the oscillator, giving as output signal a function that corresponds
to the difference of these phases, this is, the negative feedback; The VCO simply generates a
frequency signal from a voltage, that is, it acts as an integrator with a gain.

Simplest example of PLL

Voltage controlled oscillator (VCO)

The voltage-controlled oscillators, are specially designed oscillators to allow to modify its
output frequency, by applying a direct voltage of control, in this way the output of the VCO is
its frequency and the input is the control voltage. Furthermore, an alternating voltage can be
applied to obtain a frequency modulation.

Phase detector

Logic circuit that generates a voltage signal which represents the difference in phase between
two periodic signal inputs. One good type of phase detector is a XOR logic gate because of his
performance. This devise is usually a source of current, the current leaves or enters the device
depending on the result of the difference between the phases of the input, this means that the
current can be positive or negative, which is an indication off which signal is advanced or
delayed with respect to a given time reference.

Phase frequency detector

An asynchronous circuit originally made of two D-flip-flops. The logic determines which of the
two signals has a zero-crossing earlier or more often. This circuit is better that the common
phase detector because improves the pull-in range and lock time.

Multi-Phase loops

Sometimes we need high performance synthesizers, which incorporate mixers and digital
divider. Using these techniques, we can produce high performance wide range signal sources
with very small step sizer. In the next image we have an example of multi-loop synthesizers,
it’s just an example of use, where the first loop gives the smaller steps and the second one
provides larger steps.
Numerically controlled oscillator

Numerically controlled oscillator is a digital signal generator which creates a synchronous,

discrete-time, discrete-valued representation of a sinusoidal waveform. Some of the
advantages that this kind of oscillator provide are:

-The need to adjust the analog components due to the passage of time is eliminated.

-The digital interface facilitates the control of the system remotely.

-Better than other oscillators in terms of: agility, accuracy, stability and reliability.

This is a basic block diagram where:

 FCW is the frequency control word

 PR is the phase register
 PAC is the phase-to-amplitude converter

The phase accumulator adds to the value held at its output a frequency control value at each
clock sample and then the phase-to-amplitude converter uses the phase accumulator output
word usually as an index into a waveform LUT to provide a corresponding amplitude sample.

Depends of the number of bits carried in the phase accumulator, we have a bigger or smaller
frequency resolution. This concept is defined as the smallest possible incremental change in
𝐹𝑐𝑙𝑜𝑐𝑘
frequency, that is: 𝐹𝑟𝑒𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛 = 2𝑁
, where N is the number of bits carried in the PA.

This system introduces phase error noise that can be solved with interpolation method with
the look-up table.
Direct digital synthesizer

Direct digital synthesizer is a type of frequency synthesizer used to create arbitrary waveforms
from a single fixed frequency reference clock. The reference provides a stable base time for
the system and determines the accuracy of DSS frequency. This provides the clock to the NCO
that produces its discrete time output. A basic DSS consist of a reference frequency, usually a
crystal oscillator, a numerically controlled oscillator and a digital to analog converter.

Then the DAC converts the digital waveform to an analog waveform. The output
reconstruction filter rejects the spectral replicas produced by the Zero Order Hold inherent in
the analog conversion process.

A DSS has many advantages over its analog counterpart, the phase-locked loop, like much
better frequency agility, improved phase noise and precise control of the output phase across
frequency switching transitions. However, we have to be careful with the spurious products
that are the result of harmonic and non-harmonic distortion in the creation of the output
waveform due to non-linear effects in the signal processing chain in the NCO.
 Digital filtering

A filter can be defined as the process that alters the nature of a signal. It is a computational
process or algorithm by which a digital signal is transformed into a second sequence of
samples or digital output signal. Filters are widely used in all areas of signal processing,
musical, and are an essential component in any chain of communication. They form the basis
of signal processing, which can be applied to signs of all kinds (sounds, images, video, seismic
vibrations, etc).

The impulsive response is the reaction of a filter to an impulse input. The impulse response
characterizes a filter in the time domain. When we work in the digital domain, this impulse
response will be discretized in time and therefore defined by a series of sample, this is, ℎ[𝑛].

With de Fourier transform of a impulsive response of a filter corresponds to its transfer

function of frequency representation, which characterizes the filter in the frequency domain,
this is, 𝐻(𝑓). This characterization is carried out through its amplitude spectrum and the phase
spectrum. To describe the phase and the amplitude response we will work with the frequency
response in polar form, that is:

𝐻(𝑗𝑤) = |𝐻(𝑗𝑤)|𝑒 𝑗∠𝐻(𝑗𝑤) , frequency response

Therefore,

cos(𝑤𝑡) → |𝐻(𝑗𝑤)| · cos[𝑤𝑡 + ∠𝐻(𝑗𝑤)]

Where, |𝐻(𝑓)|is the amplitude response and + ∠𝐻(𝑗𝑤)is the phase response

An important note is that one filter can not be precise at the same time in the temporal and
frequency domain.

If we have an input 𝑥[𝑛] that we process with a filter to have 𝑦[𝑛] output signal, the spectrum
of the output signal 𝑌(𝑓) is obatained by multiplying the input spectrum 𝑋(𝑓) by the
frequency response of the 𝐻(𝑓) filter, that is 𝑌(𝑓) = 𝑋(𝑓) · H(f). If we work in the time
domain instead, the equivalent operation is 𝑦[𝑛] = 𝑥[𝑛] ∗ ℎ[𝑛].

The filters are described by a mathematical tool called z transform. It relates the effects os
sample delays to a two-dimensional image of the frequency representation 𝐻(𝑓) that deforms
the complex plane z. The poles in that plane represents the resonance peaks or points that
make the frequency response infinite. The zeros represent the points where the amplitude is
null in the frequency response. For example, in a 3-poles filter has 3 resonance peaks. The z
transform is an essential concept for the design of filters, because allows to relate the
characteristics of the filter that we want to design and the parameters of its implementation.
Basic operation of digital filters

The basic operation of a digital filter is relatively simple. We distinguish in fact two types of
operation: Finite impulse response filters (FIR) and infinitive impulse response filters (IIR).

FIR IIR

We can describe the filters by means of an equation that relates an input signal with an output
signal in the digital domain. In this way, the filter output is specified as the result of addition,
subtraction and multiplication of current and previous input samples. These equations are
technically called linear equations in differences. Linear means that if the input of a filter is the
sum of two scaled functions, the filter output is equal to the scaled sum of the filter outputs
for each of these functions.

In the FIR case we slightly delay a copy of the input signal (from one or more sampling periods)
and combine the delayed input signal with the new input signal. We can express this relation
with an equation like:

𝑦[𝑛] = 𝑎0 · 𝑥[𝑛] + 𝑎1 · 𝑥[𝑛 − 1] + 𝑎2 · 𝑥[𝑛 − 2]+ . . . + 𝑎𝑁 · 𝑥[𝑛 − 𝑁]

𝑁−1

= ∑ 𝑎𝑘 𝑥(𝑛 − 𝑘)
𝑘=0

Filter’s Z transform:

𝐻(𝑧) = ℎ(0)𝑧 0 + ℎ(1)−1 + ℎ(2)−2 + ⋯ + ℎ(𝑁 − 1)𝑁−1

This equation means that the current sample of the output 𝑦[𝑛] is equal to the sum of the
samples of the current input 𝑥[𝑛] multiplied by the factor a0 and all the previous samples x[n-
N] multiplied by their respective factor. This factor is the coefficients of the filter. We can
modify the characteristics of the filter if we change these values. N is the order of the filter.

In the IRR case we delay a copy of the output signal, which we combine with the new input
signal. This kind of filters are also called recursive filters or with feedback. This method allows
to implement filters with a more complex response and with less data. Since we constantly
inject energy into the circuit, the impulse response has an infinite potential duration, and
that’s where the name comes from. The typical equation for an IIR filter is:

𝑦[𝑛] = 𝑎0 · 𝑥[𝑛] + 𝑎1 · 𝑥[𝑛 − 1] + 𝑎2 · 𝑥[𝑛 − 2] + . . . + 𝑎𝑁 · 𝑥[𝑛 − 𝑁] − 𝑏1 · 𝑦[𝑛

− 1] − 𝑏2 · 𝑦[𝑛 − 2] − 𝑏3 · 𝑦[𝑛 − 3] − . . . − 𝑏𝑀 · 𝑦[𝑛 − 𝑀]

This equation express that the output is function of N+1 input samples, and M previous output
samples.

There are some differences between FIR and IIR filters. The FIR filters generally offer a more
lineal phase response and never go into oscillation because they don’t have feedback, which
means that they don’t become unstable. However, they require many terms in their equations
and that makes them more expensive in terms of calculations or computational load. On the
other hand, the IIR filters are very effective and can provide very steep cut slopes. Moreover,
by having characteristics of feedback, they have tendency to oscillate and resonate.

The most common filters are: low pass filter, high pass filter and band pass filter.

Low-pass filter Band-pass filter High-pass filter

Low-pass filters pass frequencies below a certain frequency. On the other hand, High-pass filters pass
frequencies above a certain frequency. Both types of filters are defined by their cutoff frequency, that is
the frequency when the max amplitude is reduced 3dB, as we can see in their respective figures.

Band-pass filter is a type of filter that passes a certain frequency range of a signal and attenuates the
passage of the rest.

Here we have an example of their circuits:

Low-pass filter Band-pass filter High-pass filter

Design of FIR filters

The advantage of FIR filters is that they can be designed to have linear phase. This means that certain
symmetry conditions are verified:

 A non-causal system with symmetric conjugate impulse response (ℎ(n)=h*(-n)) has a real
transfer function.
 A non-causal system with an antisymmetric conjugate impulse response (ℎ(𝑛) = −ℎ ∗
(−𝑛))has a pure imaginary transfer function.

If we consider FIR systems with real coefficients, a symmetric conjugate sequence is said to be a
EVEN sequence, and an anti-symmetric conjugate sequence is an ODD sequence. Depending on the
number of filter coefficients and the type of symmetry we have several possibilities.

Type Number of terms (N) Symmetry

I Odd Symmetrical ℎ(𝑘) = ℎ(𝑁 − 1 − 𝑘)
II Even Symmetrical ℎ(𝑘) = ℎ(𝑁 − 1 − 𝑘)
III Odd Anti-Sym ℎ(𝑘) = −ℎ(𝑁 − 1 − 𝑘)
IV Even Anti-Sym ℎ(𝑘) = −ℎ(𝑁 − 1 − 𝑘)

Type I

Type II

Type III
 Type IV

We have three main design method of FIR digital filters: Window method and frequency sampling.

The windows method is based on delimiting the infinite impulse response of an ideal filter, the
frequency sampling method proposes that a series of points of the frequency response of the system be
fixed and, from the inverse DFT, to obtain the filter coefficients.

Window method

If we want to implement a low pass filter with an ideal response (abrupt transition from the pass band
to the attenuated band), the impulse response is infinite and not causal. To obtain a feasible FIR filter, it
can be proposed to truncate h (n) and delay it until it becomes causal.

The process is the next one, we will suppose that we have the next samples:

 Obtain the impulse response of the ideal filter that we want to design ℎ𝑖 (𝑛), like low-pass filter,
high-pass filter, etc.
 Truncated this impulse response ℎ(𝑛) = ℎ𝑖 (𝑛) · 𝑤(𝑛), 𝑤(𝑛) is the impulse response of the
window.
 Move the impulse response in an appropriate number of samples to make it causal (the
impulse response of the ideal filter can also be displaced previously, so that the sequenced
sequence is causal)
Design of IIR

Now we will study the design of the IIR filters using analog prototype. There are two method that don’t
resolve the analog to digital domain transform correctly, those method are Impulse Invariant
Approximation and Derivative Approximation. This are the conditions that the transform must satisfy to
be optimal:

 Given a real and rational function in S, the resulting function in Z, will also be rational and with
real coefficients.
 Given a stable analog filter, the resulting digital filter will also be (the left half of the Laplace
domain must be transformed within the circle of unit radius in the Z domain).
 In particular, a reactance function that corresponds to the imaginary axis of the Laplace plane,
the circumference of the unit radius, is of interest.
 Given an order of the analog filter, the resulting digital filter will be of the same order.

There is a relationship that meets all these conditions, the so-called bilinear transformation:
2 1 − 𝑧 −1 2 𝑧−1
𝑠= ·( −1
)= ·
𝑇 1+𝑧 𝑇 𝑧+1

The Z transform can be written has:

𝑇 1 + 𝑧−1
𝐻(𝑧) = ·
2 1 − 𝑧−1
 Digital Converters
https://www.slideshare.net/FernandoOjeda5/decimation-and-interpolation-49640390

In the previous sections we talked about digital converters. Now let’s go deeper into more
specific aspects. Sometimes the sampling rate of a system needs to be changed for a lower or
higher sampling rate for an appropriate processing signal. This need is covered by the
decimation and interpolation.

In the case of decimation, the sampling rate is reduced by M factor. This means that when a
signal is downsampled we reduce the amount of data by taking only every M-th sample of the
signal and discarding all others. When the signal is downsampled, the replicas generated by this
𝜋
first stem can overlap, this occurs if the original signal is not bandlimited to . This effect is
𝑀
called aliasing, and the way to avoid this is using a low-pass digital filter that should precede
the downsampler. Here we have an example of a decimator:

Downsampling process can be expressed in time domain as:

𝑦[𝑛] = 𝑥[𝑀𝑛]

And in the frequency domain as:

𝑀−1
𝑤 − 2𝜋𝑘
𝑌(𝑤) = ∑ 𝑋( )
𝑀
𝑘=0

For the interpolation process first, we have to upsampling the input signal by L factor, by
inserting L-1 equally spaced zeros between each pair of samples. Unlike the decimation, a low-pass filter
is added after the upsampling. The process of upsampling introduces the replicas of the main spectra ar
2𝜋
every . This is called imaging, since there are L-1 replicas in 2𝜋.
𝐿

Upsampling process can be expressed in time domain as:

𝑛 𝑛
𝑥 [ ] 𝑖𝑓 ( 𝜖𝐼)
𝑦[𝑛] = { 𝐿 𝐿
0 𝑂𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
And in frequency domain as:

𝑌(𝑤) = 𝑋(𝐿𝑤)
 https://www.scribd.com/document/251757376/Filtros-CIC-Cascaded-Integrator-Comb
Cascaded integrator-comb filter
CIC filters are used as an antialiasing filter in the decimation processes (reduction in the
sampling frequency) in the previous stage, and as an anti-imaging. Among applications are
associated with demodulation in modern wireless systems and in A/D D/A delta-sigma
converters. CIC filter consists of one or more integrator and comb filter pair. In the case of a
desimating CIC, the input signal is fed throught one or more cascaded integrators, then a
downsampler, followed by one or more comb sections. On the other hand, an interpolating CIC
is simply the reverse of this architecture, with the downsampler replaced with an upsampler
Unlike most FIR filters, it has a decimator or interpolator built into the architecture.
The basic structure of CIC in decimation and interpolation is showed below:

CIC decimator
When decimation operation is applied in a CIC filter, and the filter is decimated by R factor, the
transfer function is the Z domain is:
𝑁
1 − 𝑍 −𝑀
𝐻(𝑧) = ( )
1 − 𝑍 −1
The comb filters and the integrator do not depend of the frequency variations. Here we have
the blocks diagram of this implementation is showed below:

In this figure it is observed that the integrating section, with N integration stages, processes the
data input at a sampling rate of 𝑓𝑠 , while the comb filter sextion, with N comb filters with
𝑓
differential M delay each one, operates at a lower sampling rate, which is from 𝑠 .
𝑅
A fundamental characteristic in the decimator CIC filters is given around the nulls in multiples
of 1/𝑀, because then the band is folded in that area, aliasing occurs.

CIC interpolator
The structure of an interpolator CIC filter is showed below:
This structure is similar to the structure of the decimator CIC filter, with the order exchanged
between the integrating and com filter sections. In this case there is an increase in the
frequency of sampling with an R factor between said sections.
The fact of exchanging positions between the sections of the filter does not modify the
operation essential, since integration and filtering perform linear operations, this allows the
difference.
1
As in the decimator CIC, around multiples of , by folding the band, produce images.
𝑀

Digital Up Conversion
The function of the DUP is to convert digital baseband samples to digital intermediate signals.
Interpolation filter is used to increase sampling rate. Radio frequency (RF) up converter follows
D/A converter which converts intermediate frequency signal to RF signal for transmission.
Hence IF translation frequency is determined by a Local Oscillator. Finally, RF power amplifier is
used which amplifies the signal to take care of path loss from transmit end to receive end. The
digital up converter has these modules: Interpolation filter, digital mixer and digital local
oscillator. Hare we can see a representation of the DUC:

Digital Down Conversion

The function of the DDC is to convert digital intermediate frequencies samples into digital
baseband samples with a lower sampling rate in order to simplify the subsequent radio stages.
The process preserves all the information in the original signal less that which is lost to
rounding errors in the mathematical processes. The modules of the DDC are: Digital mixer,
digital local oscillator and low-pass filter. Here we can see the representation of the DDC:
 Digital to digital conversion
The digital to digital conversion consists of convert digital data into digital signals and vice
versa. It can be done in two ways, line coding and block coding. Binary data is transmitted by
coding each data bit in each digital signal element. If all the elements of the signal have the
same sign, this signal is unipolar. On the other hand, we have the polar signal where a logical
state will be represented with a positive voltage level and the other a negative level.

Some types of digital transmission are:

Non-return to Zero and Non-return to Zero Inverted (Line coding)

In the case of Non-return to Zero (NRZ) uses two different voltage levels to represent binary
values. Generally, positive voltage represents 1 and negative value represents 0. This type can
be unipolar and polar. In the polar case we have non-return to zero level and a variation called
non-return to zero inverted.Below we can see the two types of encoding:

Unipolar one is represented by a DC bias on the transmission line, while zero is represented by
the absence of bias. Bipolar Non-return to zero level one is represented by one physical level
(usually positive one), while zero is represented by another level (usually a negative one).
Bipolar non-return to zero inverted is a method of mapping a binary signal to physical signal.
The two-level NRZI are represented by the presence or absence of a transition at a clock
boundary. Usually one is transmitted as a transition and zero is transmitted as no transition.

Manchester coding (Line coding)

This encoding scheme is a combination or RZ and NRZ-L. Bit time is divided into two halves. It
transits in the middle of the bit and changes phase when a different bit is encountered. When
there are equal and consecutive bits there is a transition to the start of the second bit, which is
not considered by the receiver when decoding, only the transitions separated evenly in time
are those that are considered by the receiver. There are some transitions that do not happen in
the middle of a bit. These transitions do not carry useful information and are only used to place
the signal in the next state where the next transition will take place. Although this allows the
signal to auto-synchronize, what it does is double the bandwidth requirement, in comparison
with other codes such as the NRZ Codes. Here we can see the representation of the
Manchester code in compare with original data and the clock signal:
Bipolar coding
Bipolar encoding uses three voltages level, this is the main difference with the previous ones,
which uses just two voltages level. The possibilities are positive, negative and zero. Zero
represent binary 0 and the positives and negatives voltages level represent the binary 1. The
main benefit of this type is that have a better error detection compared with the previous ones.
Below we can see the representation of bipolar coding:

Block coding
To ensure accuracy of the received data frame redundant bits are used. For example, in even-
parity, we can add another bit to make the count of 1s in the frame even. This technique
decreases the error. We can represent block coding easily with the number of bit block ratio,
𝑖𝑛𝑖𝑡𝑖𝑎𝑙 𝑏𝑖𝑡 𝑏𝑙𝑜𝑐𝑘 𝑚𝐵
this mean, = , where m-bit block is substituted with n-bit block (n>m). The
𝑏𝑖𝑡𝑠 𝑎𝑓𝑡𝑒𝑟 𝑎𝑑𝑑𝑖𝑡𝑖𝑜𝑛 𝑛𝐵
block coding involves three steps: Division, substitution and combination.
 Modulators and demodulators
Modulation encompasses the set of techniques used to carry information about a carrier wave,
typically a sine wave. These techniques allow a better use of the communication channel,
which makes it possible to transmit more information simultaneously, as well as improving the
resistance against possible noise and interference. According to the American National
Standard for Telecommunications, modulation is the process, or the result of the process, of
varying a characteristic of a carrier wave according to a signal that carries information. The
purpose of modulation is to superimpose signals on the carrier waves. 1

Basically, modulation consists in making a parameter of the carrier wave change in value
according to the variations of the modulating signal, which is the information that we want to
transmit. We have three basic types of digital modulations: amplitude-shift keying, frequency-
shift keying and phase-shift keying depending of which value we change in the carrier wave
(Amplitude, frequency or phase, respectively).

Phase-shift keying
In this case of digital modulation, we just change the phase of the carrier wave. The modulation
is accomplished by varying the sine or cosine input at a precise time. This variation has specific
number of phase options, each one of them assigned to a unique bit pattern. This means that in
the case of a system with just 1 or 0, we will have to different phase values. Each patter of bits
forms the symbol that is represented by the particular phase. To represent PSK schemes is
useful to use a constellation diagram that show the points in the complex plane. Due to the fact
that the amplitude in the different points do not change in FSK, the only difference is because
of the angle (positioned with uniform angular spacing around a circle). Below we can see an
example of FSK constellation.

PSK with 2 possibilities (BPSK)

In the BPSK example the two phases are separated by 180° (also known as 2-PSK). There are
not particular positions for the points, for example in this case they are in the real axis:
 Therefore, it handles the highest noise level or distortion before the demodulator reaches an
incorrect decision, so is the best PSK, the problem is that only can modulate 1bit/symbol.

There are essential definitions to determine the error-rates

 𝐸𝑏 , 𝑒𝑛𝑒𝑟𝑔𝑦 𝑝𝑒𝑟 𝑏𝑖𝑡
 𝐸𝑠 = 𝑛𝐸𝑏 , 𝑒𝑛𝑒𝑟𝑔𝑦 𝑝𝑒𝑟 𝑠𝑦𝑚𝑏𝑜𝑙
 𝑇𝑏 , 𝑏𝑖𝑡 𝑑𝑢𝑟𝑎𝑡𝑖𝑜𝑛
 𝑇𝑠 , 𝑠𝑦𝑚𝑏𝑜𝑙 𝑑𝑢𝑟𝑎𝑡𝑖𝑜𝑛
1
 𝑁𝑜 , 𝑛𝑜𝑖𝑠𝑒 𝑝𝑜𝑤𝑒𝑟 𝑠𝑝𝑒𝑐𝑡𝑟𝑎𝑙 𝑑𝑒𝑛𝑠𝑖𝑡𝑦
2
 𝑃𝑏 , 𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑏𝑖𝑡 𝑒𝑟𝑟𝑜𝑟
 𝑃𝑠 , 𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑠𝑦𝑚𝑏𝑜𝑙 𝑒𝑟𝑟𝑜𝑟

The bit error rate can be found as:

2𝐸𝑏
𝑃𝑏 = 𝑄 (√ )
𝑁0

Quadrature phase-shift keying (QPSK)

Another important example of PSK modulation. Now, the points are separated a phase of 90°:

The basis schema of the QPSK modulator and demodulator are showed below:

Modulator Demodulator

The QPSK modulator uses a bit-splitter, two multipliers with local oscillator, a 2-bit serial to parallel
converter, and a summer circuit. Following is the block diagram for the same. At the modulator’s input,
the message signal’s even bits and odd bits are separated by the bits splitter and are multipled with the
same carrier to generate odd BPSK (𝑃𝑆𝐾𝐼 ) and even BPSK (𝑃𝑆𝐾𝑄 ). The 𝑃𝑆𝐾𝑄 signal is anyhow phase
shifted by 90° before being modulated.
The QPSK demodulator uses two product demodulator circuits with local oscillator, two band pass
filters, two integrator circuits, and a 2-bit parallel to serial converter. Following is the diagram for the
same. The two product detectors at the input of demodulator simultaneously demodulate the two BPSK
signals. The pair of bits are recovered here from the original data. These signals after processing, are
passed to the parallel to serial converter.

Amplitude-shift keying

This technique of modulation varies the amplitude of the carrier wave. Depends of the number of bit
patterns we have a more or less discrete amplitude levels. For a binary message sequence there are two
levels, and one of them is usually zero. This is the easiest situation and is easy to represent respect the
original signal.

One of the disadvantages of ASK, compared with FSK and PSK is that it has not got a constant envelope.
This makes its processing more difficult, since linearity becomes an important factor. However, it does
make for easy of demodulation with an envelope factor.

More sophisticated encoding schemes have been developed which represent data in groups using
additional amplitude levels. For example, we can represent 8 levels with 3 bits. ASK system can be
divided into three groups. The first one is the modulator, the second one is a linear model of the effects
of the channel and the last one is the demodulator, as we can see in this figure.

Different symbols are represented with different voltages. For example, if the maximum allowed value
2𝐴
for the voltage is A, the possibilities will go from A to -A following the next expression: 𝑣𝑖 = 𝑖−𝐴
𝐿−1
where L is the number of levels that are used for transmission and i a integer number that goes from 0
2𝐴
to L-1. The difference between one voltage and other is Δ = .
𝐿−1

The way this system operates starts in the source S, which produces an array of symbols, then the
impulse generator creates impulses with an area of v[n]. These impulses go through the filter and are
sent to the channel. This means that for each symbol a different carrier wave is sent with the relative
amplitude. When the signal is going through the channel, the signal is affected by noise (atmospheric
noise, distortions, etc). Finally, the signal reaches the receiver and is converted to a digital signal with
the A/D converter. Due to the fact that the ASK is a particular case of AM modulation, the AM
demodulators are valid too for the detection of ASK signals. This demodulation can be synchronous
detection or envelope detection.
Synchronous detection

Synchronous ASK detector consists of a Square law detector, low pass filter, a comparator, and a voltage
limiter. The ASK modulated input signal is given to the Square law detector. A square law detector is one
whose output voltage is proportional to the square of the amplitude modulated input voltage. The low
pass filter minimizes the higher frequencies. The comparator and the voltage limiter help to get a clean
digital output.

Envelope detection

With this method we do not have the problems of phase and frequency that could appear with the
synchronous method. In this case, the ASK signal that arrive the demodulator pass through a non-linear
element like a diode before the low-pass filter. This detector consists of a half-wave rectifier, a low-pass
filter, and a comparator. The modulated ASK signal is given to the half-wave rectifier, which delivers a
positive half output. The low-pass filter suppresses the higher frequencies and gives an envelope
detected output from which the comparator delivers a digital output.
 Frequency-shift keying

This technique of modulation varies the frequency of the carrier wave. Depends of the number of bit
patterns we have a more or less discrete frequencies levels. The easiest situation is with just one bit
pattern (BFSK). With this scheme, the 1 is called the mark frequency and the 0 is called the space
frequency, because of their spacings. Below we can see an example of FSK modulation.

In FSK, the bit-rate(bps) is equal to the baud rate (the rate of change at the output, in symbols/s). the
2𝑓𝑑
general modulation rate is 𝑀 − 𝐹𝑆𝐾𝑛 = where 𝑓𝑑 is the maximum deviation in frequency
(𝑀−1)𝑅𝑠𝑦𝑚𝑏
and 𝑅𝑠𝑦𝑚𝑏 is the symbol rate per second.

The previous figure shows a modulated FSK signal that responds to the function 𝐴 · sin(2𝜋(𝑓 ± Δ𝑓)𝑡).
When the modulator is binary. The sign ± depends of wheter the bit to be transmitted is zero
or one, this is: 𝑓1 = 𝑓 + Δ𝑓, 𝑓0 = 𝑓 − Δ𝑓. Note that Δ𝑓 is the frequency deviation and is a
constant value which depends of the bandwidth of the modulated signal.
The general expression of an FSK signal when the modulator X(t) uses a multilevel code is:

𝐴 · sin(2𝜋(𝑓 ± 𝑋(𝑡)Δ𝑓)𝑡)

The fact that the amplitude of the modulated signal is constant and that the information is encoded with
frequency values, makes the FSK signal almost immune to the additive noise of the channel, given that it
mainly affects the amplitude. This is an important different with the ASK, that is very sensitive to the noise
that accumulates along the channel, so that the signal-to-noise ratio(S/N) at the receiver input can be so
low that the probability of error is not tolerable. This is the reason why ASK modulation is not used to
transmit data at high speed unless the transmission medium guarantees an adequate S/N, as in the case
of fiber optics. However, the FSK modulation is not affected by the additive noise of the channel,
since the modulated signal encodes the information with the frequency changes, that is, the
receiver only has to count the number of crosses per zero of the signals received. Therefore, it
suppresses the noise simply by cutting the amplitude of the FSK signal, without affecting the
information.

For a binary message, the FSK transmitter has this form

The FSK demodulator use the same two method as for ASK signals: can be synchronous
detection or envelope detection.

Synchronous detection

The FSK signal input is given to the two mixers with local oscillator circuits. These two are
connected to two band pass filters. These combinations act as demodulators and the decision
circuit chooses which output is more likely and selects it from any one of the detectors. The
two signals have a minimum frequency separation.

Envelope detection

The block diagram consists of two band pass filters, two envelope detectors, and a decision
circuit. The FSK signal passed through the two band-pass filters, tuned to space and mark
frequencies. The output form these two band-pass filters look like ASK signal, which is given to
the envelope detector. The signal in each envelope detector is modulated asynchronously.
 Quadrature Amplitude Modulation (QAM)

This modulation technique carries two independent signals by modulating both amplitude and
phase of a carrier signal. This is achieved by modulating same carrier, offset by 90°. The signal
modulated in QAM is composed of the linear sum of two signals previously modulated in
double side band (DSB). The digital QAM has a flow of binary data as input, which is divided
into groups of as many bits as required to generate N modulation states, hence speak of N-
QAM. For example, 8-QAM has three bits, which provide 8 different values, altering the phase
and amplitude of the carrier to have unique states. In this case, the constellation diagram is so
useful, in the case of 4-QAM, his constellation diagram is the same as QPSK. The points of the
constellation are uniformly arranged in a square grid with equal vertical and horizontal
separation. If we have high N, the points must be closer and therefore more susceptible to
noise and distortion. Another important fact is that the average energy of the constellation
remains the same.

8-QAM 16-QAM

QAM Modulator

The digital flow of data is divided into two parts, by means of a series-parallel converter, which
pass through two digital-analog converters. The signals are then passed through low-pass
filters and then multiplied by the same frequency carrier. This step generates two signals that
are sent as their summation.
QAM Demodulator

The received signal is divided in two ways by the power divider, and each way has a product
demodulator fed with a Local Oscillator signal, synchronized by a carrier recovery circuit, whose signal
suffers a 90° lag in one of these tracks. The signals at the output of the demodulators are processes with
low-pass filter to eliminate the multiples of the harmonics of the local oscillator signal and they are
converted in digital form to finally mix the digital data flows in a parallel-serial converter(P/S).
 Orthogonal Frequency Division Multiplex (OFDM)

OFDM is a modulation which consist in sending the information modulating in QAM or in PSK a set of
carrier waves in different frequency. Normally the OFDM modulation is used after passing the signal
through a channel encoder to correct the error produced in the transmission, then this modulation is
called COFDM (Coded OFDM).

The OFDM modulation is very robust against multipath, which is very usual in broadcasting channels, as
opposed to fading due to weather conditions and against RF interference.

Due to the characteristics of this modulation, the different signals with different delays and amplitudes
that reach the receiver contribute positively to the reception, so there is the possibility of creating
frequency broadcasting networks unique without interference problems.

What differentiates OFDM from other multiplexing procedures in frequency is orthogonality, since the
"adequate spacing" between carriers is an optimal spacing. This spacing is that the spectral separation
between consecutive carriers is always the same and equal to the inverse of the symbol period, so that
the OFDM signal can be expressed, in complex notation, as:
𝑁 𝑖
𝑠(𝑡) = ∑2−1 𝑁 𝑑𝑖 𝑒 𝑗2𝜋(𝑓𝑐+𝑇)𝑡
𝑖= −
2

𝑓𝑐 is the central frequency

T is the period
𝑑𝑖 is the symbol

In the next figure we have the representation of three orthogonal carriers. In the time domain we can
see that in the period of the carrier with the bigger period there are several periods of the other
carriers, all aligned in phase, while in spectral representation the maximum of each carrier coincides
with a null of the rest.

Time domain Frequency domain

The action of modulate or demodulate all the carriers at the same time of an OFSM signal basically
consist of apply the Fast Fourier Transform (FFT), which is well known and easy to implement in digital
processors.

About the modulation of the carriers, the symbol 𝑑𝑖 in OFDM multiplex on each carrier is modulated
with different information, though, for ease implementation, the modulation system is usually the same
for all of them, like we said before with QPSK or N-QAM. In addition, some carriers are usually reserved
to transmit sync and spectral equalization information, or to establish service channels.

OFDM has a high spectrum efficiency, resistance to the RF interface and lower multipath distortion.
Currently OFDM is not only used in wireless networks LAN 802.11a, but in the 802.11g, in high-speed
communications via as ADSLs and in the broadcasting of digital terrestrial television signals in Europe,
Japan and Australia.
Modulator

The modulation starts with s[n] which is a serial stream of binary digits (discrete). This signal pass
through a Serial to Parallel (S/N) and is demultiplexed into N parallel streams, and each one mapped to a
symbol stream using some modulation constellation as we said before. Then eh IFFT is computed on
each set of symbols, gibing a set of complex time domain samples. These samples are then quadrature-
mixed to passband in the standard way. The real and imaginary components are first converted to the
analogue domain using DAC; The analogue signals are then used to modulate cosine or sine waves at the
carrier frequency. These signals are then summed to give the s(t) signal which is transmitted.

Demodulator

Demodulator receives the signal s(t), which is then quadrature-mixed down to baseband using cosine
and sine waves at the carrier frequency. This also creates signals centered, so low-pass filters are used
to reject this. These signals are sampled and digitalized using ADC, and the pass through the FFT to back
to frequency domain. And finally this parallel streams are converted to a binary stream using an
appropriate symbol detector and the system return approximately the original binary stream at the
transmitter.