1

COSTA’S RECEIVER
Shaun Job
210108043

I. R EVISITING THE DSB-AM DEMODULATION

In the previous class we looked into carrierless modulation, specifically the double sideband
supressed carrier (DSB-FC) AM. However we did run into some issues during it’s demodulation,
So let’s jog our memory a bit and see what happens !
Taking the diagram above as reference (neglecting ∆ initially), We have

s(t) = m(t)cos(2πfc t) (1)

Let us now define,

s′ (t) = s(t)cos(2πfc t)
(2)
m(t)
= (1 + cos(4πfc t))
2
After passing s’(t) through a low pass filter h(t) we will have,
ˆ = h(t) ∗ s′ (t)
m(t)
(3)
m(t)
=
2

August 26, 2023 DRAFT

 2

If the pass band gain of the low pass Filter is set as 2 (to get rid of the fraction) Then,

ˆ = m(t)
m(t) (4)

Therefore, we can recover the transmitted signal easily. However, there is just one glaring problem
with this approach that becomes immediately evident
The carrier signal, cos (2πfc t) ), must be generated using local oscillators. The modulator and
the receiver will each have their own local oscillators, and there is no way to ensure that these
oscillators are perfectly synchronized. This means it is imperative that there must be some phase
difference, ∆ between the carrier signals at the transmitter and the receiver.
So How does this change our demodulated signal ??
Following the signal flow from the earlier diagram, this time considering ∆ we get,

s(t) = m(t)cos(2πfc t) (5)

s′ (t) = s(t) cos (2πfc t + ∆)

m(t)
= (cos(∆) + cos(4πfc t + ∆)) (6)
2
ˆ = m(t) cos(∆)
m(t)
2
Looking at the result of the Equation set(6), we can solve the problem of phase difference by
adjusting the pass gain of the low pass filter to 1/cos(∆).
But clearly that’s not going to be easy !! as the Phase Difference ∆ is unknown and cannot be
ˆ = 0 and thus the Signal can never be Recovered.
controlled. Also if ∆ = π/2, m(t)
So how do we deal with such case? This is where the Costa’s Receiver comes into picture.

II. COSTA’S RECEIVER

So how do we deal with such signals? We define a signal s”(t) such that,

s”(t) = s(t) sin (2πfc t + ∆) (7)

On substitution for s(t) and applying trigonometric identities we get,

s”(t) = m(t)cos(2πfc t) sin (2πfc t + ∆)

(8)
m(t)
= (sin(∆) + sin (4πfc t + ∆))
2

August 26, 2023 DRAFT

 3

Now on passing this through a low pass filter the 4πfc t component gets filtered. Let’s call the
¯ .
residual signal m(t)
So we have, We define a signal s”(t) such that,

¯ = m(t) sin(∆)
m(t) (9)
2
Now the magic happens, Consider the ratio r(t),
¯
m(t)
r(t) = (10)
ˆ
m(t)
On further simplification using Equations (6) and (9),
¯
m(t)
m(t)
sin∆
2
r(t) = = = tan(∆) ≈ ∆(f or∆ ≈ 0) (11)
ˆ
m(t)
m(t)
cos∆
2

So, when ∆→ 0, r(t) ≈ ∆

So by using the magnitude of r(t), which is nothing but the phase error ∆ itself we can adjust
the phase of the Local oscillator generating the carrier signal, thus making both the transmitter
and receiver coherent.
The functional block diagram of the costa’s loop receiver is given below.

III. MAKING USE OF BANDWIDTH

So far our discussions have only been about real-valued message signals i.e m(t) is real.
Therefore the Laplace transform of the message signal i.e M(f) will be symmetric about the

August 26, 2023 DRAFT

 4

Y-axis. Basically the frequency spectrum would be an even function. For real signals, only one
sideband say for example the upper sideband is enough to transmit all the information carried
by the signal as shown in the figure below.
for the Upper side Band we have |f | ≥ fc

However, this use of bandwidth is not really efficient at all!! Practically the an entire half of
bandwidth sits untapped. So how do we use the entirety of bandwidth?
We can either Limit ourselves to just one sideband or we can make the signal complex.
If the signal is made complex, it’s spectrum would be assymetric and thus both the sidebands
can be used for transmitting information.
Let’s look into the second method, which is making the Signal Complex
This method is called Quadrature Carrier Multiplexing (QCM) or Quadrature Amplitude
Modulation (QAM) which enables two DSBSC modulated waves, resulting from two different
message signals to occupy the same transmission band width and two message signals can be
separated at the receiver.
Let m(t) be composed of 2 parts mI (t) and mQ (t) where mI (t) is called inphase component
and mQ (t) is the quadrature component.
Note : Refer the diagram below,m1 (t) and m2 (t) from this diagram shall henceforth be referred
to as mI (t) mQ (t)

August 26, 2023 DRAFT

 5

s(t) = ℜ(mI (t)ej2πfc t )

(12)
= mI (t)cos(2πfc t) + mQ (t)sin(2πfc t)
Furthermore,
mI (t)
s(t)cos(2πfc t) = + [mI (t)cos(4πfc t) + mQ (t)sin(4πfc t)] (13)
2
And,
mQ (t)
s(t)sin(2πfc t) = + [mI (t)sin(4πfc t) − mQ (t)cos(4πfc t)] (14)
2
When we start Demultiplexing the signals (13) and (14) by using 2 low pass filters with a
passband gain of 2 (to cancel out the 1/2 generated in trignometric simplifications) we get back
our inphase and quadrature signals individually as shown in the diagram above.
There are numerous practical applications of QCM, some of them being,
Cable modems use QAM to transmit digital data over cable television infrastructure. This enables
high-speed internet access for homes and businesses.
Cellular technologies like LTE and 5G use QAM to encode and decode data for wireless
communication between cell towers and mobile devices.
So this is all about the second method, what about the first?? Where we just use a single sideband
and consider only 1 half of the frequencies...
Let’s recollect by having a look at the frequency spectrum..

August 26, 2023 DRAFT

 6

The frequency spectrum of our message signal m(t) can be expressed in terms of it’s Hilbert
transform. Let’s have a look at it, As it is evident from the diagram we define,
M+ (f ) = M (f )U (f )
(15)
M+ (f ) = M (f )U (−f )
Where U(f) is the unit step function. U(f) can be expressed in terms of sgn(f) [Signum function]
as U (f ) = (1 + sgn(f ))/2. We know that in frequency domain the Hilbert transform of M(f) is
Mh (f ) = −jsgn(f )M (f ) Substituting for sgn(f ) from Mh (f ) we get,
1
M+ (f ) = (M (f ) + jMh (f ))
2 (16)
1
M− (f ) = (M (f ) − jMh (f ))
2
Let’s consider only the Upper side band for this kind of modulation.

Clearly this upper sideband signal can be written as combination of the shifted versions of
M+ (f ) and M− (f ).
ϕusb (t) = M+ (f )(f − fc ) + M− (f )(f + fc ) (17)

August 26, 2023 DRAFT

 7

Now taking the inverse fourier transform and using the relations sin(θ) = (ejθ − e−jθ )/(2j) and
cos(θ) = (ejθ + e−jθ )/(2) we get,
ˆ
ϕusb (t) = m(t)cos(2πfC t) − m(t)sin(2πfc t) (18)
ˆ is the Hilbert transform of m(t)
where m(t)

This will only occupy half the bandwidth required by m(t). Also,
m(t) m(t) ˆ
m(t)
ϕusb (t)cos(2πfC t) = [ sin(4πfc t) − cos(4πfC t)] (19)
2 2 2
ˆ
m(t)
Passing this through a Low Pass Filter, we will get 2
(neglecting the factor of 2 in the carrier
signal shown in the above image).
In the presence of perfect phase synchronization, we can transmit the amplitude-modulated signal
using only half the bandwidth, and this technique is known as Single Sideband Modulation
(SSB). SSB modulation eliminates one of the sidebands (upper or lower) along with the carrier,
resulting in a significant reduction in bandwidth usage compared to Double Sideband (DSB)
modulation. This efficient use of bandwidth allows more channels to be accommodated within
the same frequency spectrum.
Because SSB modulation uses the available power more efficiently, it provides a better SNR(signal
to noise ratio) at the receiver.
However, the Costa’s loop is not suitable for phase recovery in single sideband suppressed
carrier (SSB-SC) systems or vestigial sideband suppressed carrier (VSB-SC) systems having
very small vestige. In general, this is due to the dependence of the quadrature components
on the in-phase components of the baseband signal.
So instead we use Pilot Signals(low power carrier) with the quadrature carrier multiplexing for
SSB signals to allow phase recovery at the receiver. But more on that later !

August 26, 2023 DRAFT