mywbut.

com

Chapter 3

Complex Baseband Representation of

Bandpass Signals

3.1 Introduction
Almost every communication system operates by modulating an information bearing waveform onto a
sinusoidal carrier. As examples, Table 3.1 lists the carrier frequencies of various methods of electronic
communication.

Type of Transmission Center Frequency of Transmission

Telephone Modems 1600-1800 Hz
AM radio 530-1600 KHz
CB radio 27 MHz
FM radio 88-108 MHz
VHF TV 178-216 MHz
Cellular radio 850 MHz
Indoor Wireless Networks 1.8GHz
Commercial Satellite Downlink 3.7-4.2 GHz
Commercial Satellite Uplink 5.9-6.4 GHz
Fiber Optics 2 × 1014 Hz

Table 3.1: Carrier frequency assignments for different methods of information transmission.

One can see by examining Table 3.1 that the carrier frequency of the transmitted signal is not the
component which contains the information. Instead it is the signal modulated on the carrier which con-
tains the information. Hence a method of characterizing a communication signal which is independent
of the carrier frequency is desired. This has led communication system engineers to use a complex
baseband representation of communication signals to simplify their job. All of the communication
systems mentioned in Table 3.1 can be and typically are analyzed with this complex baseband repre-
sentation. This handout develops the complex baseband representation for deterministic signals. Other
references which do a good job of developing these topics are [Pro89, PS94, Hay83, BB99]. One ad-
vantage of the complex baseband representation is simplicity. All signals are lowpass signals and the
fundamental ideas behind modulation and communication signal processing are easily developed. Also
any receiver that processes the received waveform digitally uses the complex baseband representation
to develop the baseband processing algorithms.

1
mywbut.com

3.2 Baseband Representation of Bandpass Signals

The first step in the development of a complex baseband representation is to define a bandpass signal.

Definition 3.1 A bandpass signal, xc (t), is a signal whose one-sided energy spectrum is both: 1) cen-
tered at a non-zero frequency, fC , and 2) does not extend in frequency to zero (DC).

The two sided transmission bandwidth of a signal is typically denoted by BT Hertz so that the one-sided
spectrum of the bandpass signal is zero except in [fC − BT /2, fC + BT /2]. This implies that a bandpass
signal satisfies the following constraint: BT /2 < fC . Fig. 3.1 shows a typical bandpass spectrum. Since
a bandpass signal, xc (t), is a physically realizable signal it is real valued and consequently the energy
spectrum will always be symmetric around f = 0. The relative sizes of BT and fC are not important,
only that the spectrum takes negligible values around DC. In telephone modem communications this
region of negligible spectral values is only about 300Hz while in satellite communications it can be many
Gigahertz.

Gxc ( f )

BT

− fC fC

Figure 3.1: Energy spectrum of a bandpass signal.

A bandpass signal has a representation of

√ √
xc (t) = xI (t) 2 cos(2πfc t) − xQ (t) 2 sin(2πfc t) (3.1)
√
= xA (t) 2 cos (2πfc t + xP (t)) (3.2)

where fc is denoted the carrier frequency with fC − BT /2 ≤ fc ≤ fC + BT /2. The signal xI (t) in (3.1)
is normally referred to as the in-phase (I) component of the signal and the signal xQ (t) is normally
referred to as the quadrature (Q) component of the bandpass signal. xI (t) and xQ (t) are real valued
lowpass signals with a one-sided non–negligible energy spectrum no larger than BT Hertz. It should be
noted that the center frequency of the bandpass signal, fC , (see Fig. 3.1) and the carrier frequency, fc
are not always the same. While fc can theoretically take a continuum of values in most applications
an obvious value of fc will give the simplest representation1 . The carrier signal is normally thought of
as the cosine term, hence the I component is in-phase with the carrier. Likewise the sine term is 90◦
out-of-phase (in quadrature) with the cosine or carrier term, hence the Q component is quadrature to
the carrier. Equation (3.1) is known as the canonical form of a bandpass signal. Equation (3.2) is the
amplitude and phase form of the bandpass signal, where xA (t) is the amplitude of the signal and xP (t)

2
mywbut.com

is the phase of the signal. A bandpass signal has two degrees of freedom and the I/Q or the amplitude
and phase representations are equivalent. The transformations between the two representations are
given by 
xA (t) = xI (t)2 + xQ (t)2 xP (t) = tan−1 [xQ (t), xI (t)] (3.3)
and
xI (t) = xA (t) cos (xP (t)) xQ (t) = xA (t) sin (xP (t)) . (3.4)
Note that the tangent function in (3.3) has a range of [−π, π] (i.e., both the sign xI (t) and xQ (t) and
the ratio of xI (t) and xQ (t) are needed to evaluate the function). The particulars of the communication
design analysis determine which form for the bandpass signal is most applicable.
A complex valued signal, denoted the complex envelope, is defined as

xz (t) = xI (t) + jxQ (t) = xA (t) exp [jxP (t)] .

The original bandpass signal can be obtained from the complex envelope by
√
xc (t) = 2
[xz (t) exp [j2πfc t]] .

Since the complex exponential only determines the center frequency, the complex signal xz (t) contains
all the information in xc (t). Using this complex baseband representation of bandpass signals greatly
simplifies the notation for communication system analysis. As the quarter goes along hopefully the
additional simplicity will become very evident.
Example 3.1: Consider the bandpass signal
√ √
xc (t) = 2 cos(2πfm t) 2 cos(2πfc t) − sin(2πfm t) 2 sin(2πfc t)

where fm < fc . A plot of this bandpass signal is seen in Fig. 3.2 with fc = 10fm . Obviously we have

xI (t) = 2 cos(2πfm t) xQ (t) = sin(2πfm t)

and
xz (t) = 2 cos(2πfm t) + j sin(2πfm t).
The amplitude and phase can be computed as

xA (t) = 1 + 3 cos2 (2πfm t) xP (t) = tan−1 [sin(2πfm t), 2 cos(2πfm t)] .

A plot of the amplitude and phase of this signal is seen in Fig. 3.3.

The next item to consider is methods to translate between a bandpass signal and a complex envelope
signal. Basically a bandpass signal is generated from its I and Q components in a straightforward fashion
corresponding to (3.1). Likewise a complex envelope signal is generated from the bandpass signal with
a similar architecture. Using the results
√
xc (t) 2 cos(2πfc t) = xI (t) + xI (t) cos(4πfc t) − xQ (t) sin(4πfc t)
√
xc (t) 2 sin(2πfc t) = −xQ (t) + xQ (t) cos(4πfc t) + xI (t) sin(4πfc t) (3.5)

Fig. 3.4 shows these transformations where the lowpass filters remove the 2fc terms in (3.5). Note in
the Fig. 3.4 the boxes with π/2 are phase shifters (i.e., cos (θ − π/2) = sin(θ)) typically implemented
with delay elements. The structure in Fig. 3.4 is fundamental to the study of all modulation techniques.

3
mywbut.com

1
x (t)

0
c

-1

-2

-3
0 0.5 1 1.5 2
Normalized time, f mt

Figure 3.2: Plot of the bandpass signal for Example 3.1.

Amplitude Phase
2
150

1.5 100
Phase, degrees

50
Amplitude

1 0

-50
0.5
-100

-150
0
0 0.5 1 1.5 2
Normalized Time

Figure 3.3: Plot of the amplitude and phase for Example 3.1.

4
mywbut.com

Complex Baseband to Bandpass to Complex

Bandpass Conversion Baseband Conversion

x I (t ) LPF x I (t )

+ xc (t )
2 cos(2πfc t ) ∑ 2 cos(2πfc t )

-
π 2 π 2

xQ (t ) LPF − xQ (t )

Figure 3.4: Schemes for converting between complex baseband and bandpass representations. Note that
the LPF simply removes the double frequency term associated with the down conversion.

3.3 Spectral Characteristics of the Complex Envelope

3.3.1 Basics
It is of interest to derive the spectral representation of the complex baseband signal, xz (t), and compare
it to the spectral representation of the bandpass signal, xc (t). Assuming xz (t) is an energy signal, the
Fourier transform of xz (t) is given by
Xz (f ) = XI (f ) + jXQ (f ) (3.6)
where XI (f ) and XQ (f ) are the Fourier transform of xI (t) and xQ (t), respectively, and the energy
spectrum is given by  
∗
Gxz (f ) = GxI (f ) + GxQ (f ) + 2 XI (f )XQ (f ) (3.7)
where GxI (f ) and GxQ (f ) are the energy spectrum of xI (t) and xQ (t), respectively. The signals xI (t)
and xQ (t) are lowpass signals with a one-sided bandwidth of less than BT so consequently Xz (f ) and
Gxz (f ) can only take nonzero values for |f | < BT .
Example 3.2: Consider the case when xI (t) is set to be the message signal from Example 1.15(computer
voice saying “bingo”) and xQ (t) = cos (2000πt). XI (f ) will be a lowpass spectrum with a bandwidth of
2500Hz while XQ (f ) will have two impulses located at ±1000Hz. Fig. 3.5 show the measured complex
envelope energy spectrum for these lowpass signals. The complex envelope energy spectrum has a
relation to the voice spectrum and the sinusoidal spectrum exactly as predicted in (3.6).

Eq. (3.6) gives a simple way to transform between the lowpass signal spectrums to the complex
envelope spectrum. A similar simple formula exists for the opposite transformation. Note that xI (t) and
xQ (t) are both real signals so that XI (f ) and XQ (f ) are Hermitian symmetric functions of frequency
and it is straightforward to show
Xz (−f ) = XI∗ (f ) + jXQ
∗
(f )

5
mywbut.com

Energy spectrum of the complex envelope

20

10

-10

-20

-30

-40

-50

-60
-5000 -4000 -3000 -2000 -1000 0 1000 2000 3000 4000 5000
Frequency, Hertz

Figure 3.5: The complex envelope resulting from xI (t) being a computer generated voice signal and
xQ (t) being a sinusoid.

Xz∗ (−f ) = XI (f ) − jXQ (f ). (3.8)

This leads directly to

Xz (f ) + Xz∗ (−f )
XI (f ) =
2
Xz (f ) − Xz∗ (−f )
XQ (f ) = . (3.9)
j2

Since xz (t) is a complex signal, in general, the energy spectrum, Gxz (f ), has none of the usual properties
of real signal spectra (i.e., spectral magnitude is even and the spectral phase is odd).
An analogous derivation produces the spectral characteristics of the bandpass signal. Examining
(3.1) and using the Modulation Theorem of the Fourier transform, the Fourier transform of the bandpass
signal, xc (t), is expressed as
  
1 1 1 1
Xc (f ) = √ XI (f − fc ) + √ XI (f + fc ) − √ XQ (f − fc ) − √ XQ (f + fc ) .
2 2 2j 2j

This can be rearranged to give

 
XI (f − fc ) + jXQ (f − fc ) XI (f + fc ) − jXQ (f + fc )
Xc (f ) = √ + √ (3.10)
2 2
Using (3.8) in (3.10) gives
1 1
Xc (f ) = √ Xz (f − fc ) + √ Xz∗ (−f − fc ). (3.11)
2 2
This is a very fundamental result. Equation (3.11) states that the Fourier transform of a bandpass
signal is simply derived from the spectrum of the complex envelope.√ For positive values of f , Xc (f ) is
obtained by translating Xz (f ) to fc and scaling the amplitude by 1/ 2. For negative values of f , Xc (f )
is obtained by flipping Xz (f ) around the origin, taking the complex conjugate, translating the result to

6
mywbut.com

Gxz ( f )

BT

Figure 3.6: The complex envelope energy spectrum of the bandpass signal in Fig. 3.1 with fc = fC .

√
−fc , and scaling the amplitude by 1/ 2 . This also demonstrates that if Xc (f ) only takes values when
the absolute value of f is in [fc − BT , fc + BT ], then Xz (f ) only takes values in [−BT , BT ]. The energy
spectrum of xc (t) can also be expressed in terms of the energy spectrum of xz (t) as

1 1
Gxc (f ) = Gxz (f − fc ) + Gxz (−f − fc ). (3.12)
2 2

Equation (3.12) guarantees that the energy spectrum of the bandpass signal is an even function and that
the energy of the complex envelope is identical to the energy of the bandpass signal. Considering these
results, the spectrum of the complex envelope of the signal shown in Fig. 3.1 will have a form shown in
Fig. 3.6 when fc = fC . Other values of fc would produce a different but equivalent complex envelope
representation. This discussion of the spectral characteristics of xc (t) and xz (t) should reinforce the
idea that the complex envelope contains all the information in a bandpass waveform.
Example 3.3: (Example 3.1 continued)

xI (t) = 2 cos(2πfm t) xQ (t) = sin(2πfm t)

1 1
XI (f ) = δ(f − fm ) + δ(f + fm ) XQ (f ) = δ(f − fm ) − δ(f + fm )
2j 2j
Xz (f ) = 1.5δ(f − fm ) + 0.5δ(f + fm )
Note in this example BT = 2fm .
1.5 1 1.5 1
Xc (f ) = √ δ(f − fc − fm ) + √ δ(f − fc + fm ) + √ δ(f + fc + fm ) + √ δ(f + fc − fm )
2 2 2 2 2 2

Example 3.4: For the complex envelope derived in Example 3.2 the measured bandpass energy spectrum
for fc =7000Hz is shown in Fig. 3.7. Again the measured output is exactly predicted by (3.12).

7
mywbut.com

Energy spectrum of the bandpass signal

20

10

-10

-20

-30

-40

-50

-60
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Frequency, Hertz x 10 4

Figure 3.7: The bandpass spectrum corresponding to Fig. 3.5.

3.3.2 Bandwidth of Bandpass Signals

The ideas of bandwidth of a signal extend in an obvious way to bandpass signals. For bandpass energy
signals we have the following two definitions
Definition 3.2 If a signal xc (t) has an energy spectrum Gxc (f ) then BX is determined as

10 log max Gxc (f ) = X + 10 log (Gxc (f1 )) (3.13)

f

where Gxc (f1 ) > Gxc (f ) for 0 < f < f1 and

10 log max Gxc (f ) = X + 10 log (Gxc (f2 )) (3.14)

f

where Gxc (f2 ) > Gxc (f ) for f > f2 where f2 − f1 = Bx .

Definition 3.3 If a signal xc (t) has an energy spectrum Gxc (f ) then BP is determined as
 f2
2 f1 Gxc (f )df
P = (3.15)
Exc
where BP = f2 − f1 .
Note the reason for the factor of 2 in (3.15) is that half of the energy of the bandpass signal is associated
with positive frequencies and half of the energy is associated with negative signals
Again for bandpass power signals similar ideas hold with Gxc (f ) being replaced with Sxc (f, T )., i.e.,
Definition 3.4 If a signal xc (t) has a sampled power spectral density Sxc (f, T ) then BX is determined
as

10 log max Sxc (f, T ) = X + 10 log (Sxc (f1 , T )) (3.16)

f

where Sxc (f1 , T ) > Sxc (f, T ) for 0 < f < f1 and

10 log max Sxc (f, T ) = X + 10 log (Sxc (f2 , T )) (3.17)

f

where Sxc (f2 , T ) > Sxc (f, T ) for f > f2 where f2 − f1 = Bx .

8
mywbut.com

Definition 3.5 If a signal xc (t) has an power spectrum Sxc (f, T ) then BP is determined as
 f2
2 f1 Sxc (f )df
P = (3.18)
Pxc (T )
where BP = f2 − f1 .

3.4 Linear Systems and Bandpass Signals

This section discusses methods for calculating the output of a linear, time-invariant (LTI) filter with
a bandpass input signal using complex envelopes. Linear system outputs are characterized by the
convolution integral given as  ∞
yc (t) = xc (τ )h(t − τ )dτ (3.19)
−∞
where h(t) is the impulse response of the filter. Since the input signal is bandpass, the effects of the
filter in (3.19) can be modeled with an equivalent bandpass filter. This bandpass LTI system also has
a canonical representation given as

hc (t) = 2hI (t) cos(2πfc t) − 2hQ (t) sin(2πfc t). (3.20)

The complex envelope for this bandpass impulse response is given by

hz (t) = hI (t) + jhQ (t)

where the bandpass system impulse response is

hc (t) = 2
[hz (t) exp [j2πfc t]] .

The representation of the bandpass system in (3.20) has a constant factor difference from the bandpass
signal representation of (3.1). This factor is a notational convenience that permits a simpler expression
for the system output (as is shown shortly). Using similar techniques as in Section 3.3, the transfer
function is expressed as
Hc (f ) = Hz (f − fc ) + Hz∗ (−f − fc ).
This result and (3.11) combined with the convolution theorem of the Fourier transform produces an
expression for the Fourier transform of y(t) given as
1
Yc (f ) = Xc (f )Hc (f ) = √ [Xz (f − fc ) + Xz∗ (−f − fc )] [Hz (f − fc ) + Hz∗ (−f − fc )] .
2
Since both Xz (f ) and Hz (f ) only take values in [−BT , BT ], the cross terms in this expression will be
zero and Yc (f ) is given by
1
Yc (f ) = √ [Xz (f − fc )Hz (f − fc ) + Xz∗ (−f − fc )Hz∗ (−f − fc )] . (3.21)
2
Since yc (t) will also be a bandpass signal, it will also have a complex baseband representation. A
comparison of (3.21) with (3.11) demonstrates the Fourier transform of the complex envelope of y(t),
yz (t), is given as
Yz (f ) = Xz (f )Hz (f ).
Linear system theory produces the desired form
 ∞
yz (t) = xz (τ )hz (t − τ )dτ = xz (t) ⊗ hz (t). (3.22)
−∞

9
mywbut.com

In other words, convolving the complex envelope of the input signal with the complex envelope of
the filter response produces the complex envelope of the output signal. The different scale factor was
introduced in (3.20) so that (3.22) would have a familiar form. This result is significant since yc (t) can
be derived by computing a convolution of baseband (complex) signals which is generally much simpler
than computing the bandpass convolution. Since xz (t) and hz (t) are complex, yz (t) is given in terms of
the I/Q components as

yz (t) = yI (t) + jyQ (t) = [xI (t) ⊗ hI (t) − xQ (t) ⊗ hQ (t)] + j [xI (t) ⊗ hQ (t) + xQ (t) ⊗ hI (t)] .

Fig 3.8 shows the lowpass equivalent model of a bandpass system. The two biggest advantages of using
the complex baseband representation are that it simplifies the analysis of communication systems and
permits accurate digital computer simulation of filters and the effects on communication systems per-
formance.
Example 3.5: (Example 3.1 continued) The input signal and Fourier transform are

xz (t) = 2 cos(2πfm t) + j sin(2πfm t) Xz (f ) = 1.5δ(f − fm ) + 0.5δ(f + fm ).

Assume a bandpass filter with

HI (f ) = 2 −2fm ≤ f ≤ 2fm HQ (f ) = j ffm −fm ≤ f ≤ fm

= 0 elsewhere = j fm ≤ f ≤ 2fm
= −j −2fm ≤ f ≤ −fm
= 0 elsewhere.

This produces
f
Hz (f ) = 2− fm −fm ≤ f ≤ fm
= 1 fm ≤ f ≤ 2fm
= 3 −2fm ≤ f ≤ −fm
= 0 elsewhere.
and now the Fourier transform of the complex envelope of the filter output is

Yz (f ) = Hz (f )Xz (f ) = 1.5δ(f − fm ) + 1.5δ(f + fm ).

The complex envelope and bandpass signal are given as

√
yz (t) = 3 cos(2πfm t) yc (t) = 3 cos(2πfm t) 2 cos(2πfc t)

3.5 Conclusions

The complex baseband representation of bandpass signals permits accurate characterization and analysis
of communication signals independent of the carrier frequency. This greatly simplifies the job of the
communication system engineer. A linear system is often an accurate model for a communication system,
even with the associated transmitter filtering, channel distortion, and receiver filtering. As demonstrated
in Fig 3.9, the complex baseband methodology truly simplifies the models for a communication system
performance analysis.

10
mywbut.com

x I (t ) hI (t )
+
Σ yI (t )
-
hQ (t )

xQ (t )
hQ (t )
+
Σ yQ (t )
+
hI (t )

Figure 3.8: Block diagram illustrating the relation between the input and output complex envelope of
bandpass signals for a linear time invariant system.

3.6 Homework Problems

Problem 3.1. Many integrated circuit implementations of the quadrature upconverters produce a
bandpass signal having a form

xc (t) = xI (t) cos (2πfc t) + xQ (t) sin (2πfc t) (3.23)

from the lowpass signals xI (t) and xQ (t) as opposed to (3.1). How does this sign difference affect the
transmitted spectrum? Specifically for the complex envelope energy spectrum given in Fig. 3.6 plot the
transmitted bandpass energy spectrum.
Problem 3.2. Find the form of xI (t) and xQ (t) for the following xc (t)

a) xc (t) = sin (2π(fc − fm )t).

b) xc (t) = cos (2π(fc + fm )t).

c) xc (t) = cos (2πfc t + φp )

Problem 3.3. If the lowpass components for a bandpass signal are of the form

xI (t) = 12 cos(6πt) + 3 cos(10πt)

and
xQ (t) = 2 sin(6πt) + 3 sin(10πt)

a) Calculate the Fourier series of xI (t) and xQ (t).

b) Calculate the Fourier series of xz (t).

c) Assuming fc =20Hz calculate the Fourier series of xc (t)

d) Calculate and plot xA (t). Computer might be useful.

e) Calculate and plot xP (t). Computer might be useful.

11
mywbut.com

ransmitter Channel eceiver

x I (t ) 1 f (f ) (t)

2 cos( 2πf ) os( π )

∑ f H3 ( ) ( f)
π
π
2
xQ (t ) (f ) H5 ( ) (t

x z (t ) H z
(

Figure 3.9: A comparison between a) the actual communication system model and b) the complex
baseband equivalent model.

Problem 3.4. A bandpass filter has the following complex envelope representation for the impulse
response



1 t 1 t
hz (t) = 2 exp − + j2 exp − t≥0
2 2 4 4
= 0 elsewhere

a) Calculate Hz (f ). Hint: The transforms you need are in a table somewhere.

b) With xz (t) from Problem 3.3 as the input, calculate the Fourier series for the filter output, yz (t).

c) Plot the output amplitude, yA (t), and phase, yP (t).

d) Plot the resulting bandpass signal, yc (t) using fc =20Hz.

Problem 3.5. The picture of a color television set proposed by the National Television System Com-
mittee (NTSC) is composed by scanning in a grid pattern across the screen. The scan is made up
of three independent beams (red, green, blue). These independent beams can be combined to make
any color at a particular position. In order to make the original color transmission compatible with
black and white televisions the three color signals (xr (t), xg (t), xb (t)) are transformed into a luminance
signal (black and white level), xL (t), and two independent chrominance signals, xI (t) and xQ (t). These
chrominance signals are modulated onto a carrier of 3.58MHz to produce a bandpass signal for transmis-
sion. A commonly used tool for video engineers to understand this coloring patterns is the vectorscope
representation shown in Figure 3.10.

a) If the video picture is making a smooth transition from a blue color (at t=0) to green color (at
t=1), make a plot of the waveforms xI (t) and xQ (t).

12
mywbut.com

xQ (t )

0.45 exp[ j 77o ]

Yellow
0.59 exp[ j151o ]
Green 0.64 exp[ j13.7o ]
Red

0.64 exp[ j193o ] x I (t )

Cyan
0.59 exp[ j331o ]
Magenta

0.45 exp[ j 257o ]

Blue

Figure 3.10: Vector scope representation of the complex envelope of the 3.58MHz chrominance carrier.

b) What form would xI (t) and xQ (t) have to represents a scan across a red and green striped area.
For consistency in the answers assume the red starts and t=0 and extends to t = 1, the green
starts at t = 1+ and extends to t = 2, · · ·.
Problem 3.6. Consider two lowpass spectrum, XI (f ) and XQ (f ) in Figure 3.11 and sketch the energy
spectrum of the complex envelope, Gxz (f ).

Re[ X I ( f )] [ ]
Im XQ ( f )

− f1
− f1 f1 f1 f

Im[ X I ( f )] = 0 [
Re XQ ( f ) = 0 ]

Figure 3.11: Two lowpass Fourier transforms.

Problem 3.7. The lowpass signals, xI (t) and xQ (t), which comprise a bandpass signal are given in
Figure 3.12.
a) Give the form of xc (t), the bandpass signal with a carrier frequency fc , using xI (t) and xQ (t).
b) Find the amplitude, xA (t), and the phase, xP (t), of the bandpass signal.
c) Give the simplest form for the bandpass signal over [2T, 3T ].
Problem 3.8. The amplitude and phase of a bandpass signal is plotted in Figure 3.13. Compute the
in-phase and quadrature signals of this baseband representation of a bandpass signal.

13
mywbut.com

x I (t )
1

T 2T 3T 4T
-1

xQ (t )
1

T 2T 3T
4T
-1

Figure 3.12: xI (t) and xQ (t).

x A (t )

T 2T 3T T
x P (t )
π

π /2

T 2T 3T T
−π / 2

−π

Figure 3.13: The amplitude and phase of a bandpass signal.

Problem 3.9. (Design Problem) A key component in the quadrature up/down converter is the
generator of the sine and cosine functions. This processing is represented in Figure 3.14 as a shift in
the phase by 90◦ of a carrier signal. This function is done in digital processing in a trivial way but if
the carrier is generated by an analog source the implementation is more tricky. Show that this phase
shift can be generated with a time delay as in Figure 3.15. If the carrier frequency is 100MHz find the
value of the delay to achieve the 90◦ shift.
Problem 3.10. The block diagram in Fig. 3.16 shows a cascade of a quadrature upconverter and a
quadrature downconverter where the phases of the two (transmit and receive) carriers are not the same.

14
mywbut.com

2 cos(2πfc t ) 2 cos(2πfc t )

π 2 2 cos(2πfc t − π 2) = 2 sin(2πfc t )

Figure 3.14: Sine and cosine generator.

2 cos(2πfc t ) 2 cos(2πfc t )

τ 2 cos(2πfc (t − τ ))

Figure 3.15: Sine and cosine generator implementation for analog signals.

Show that yz (t) = xz (t) exp[−jθ(t)]. Specifically consider the case when the frequencies of the two
carriers are not the same and compute the resulting output energy spectrum GYz (f ).

yI (t )
x I (t ) LPF

+
xc (t )
2 cos(2πfc t ) ∑ 2 cos(2πfc t + θ (t ))

-
π 2 π 2

xQ (t ) LPF
− yQ (t )

Figure 3.16: A downconverter with a phase offset.

Problem 3.11. A periodic real signal of bandwidth W and period T is xI (t) and xQ (t) = 0 for a
bandpass signal of carrier frequency fc > W .

a) Can the resulting bandpass signal, xc (t), be periodic with a period of Tc < T ? If yes give an
example.

b) Can the resulting bandpass signal, xc (t), be periodic with a period of Tc > T ? If yes give an
example.

c) Can the resulting bandpass signal, xc (t), be periodic with a period of Tc = T ? If yes give an
example.

d) Can the resulting bandpass signal, xc (t), be aperiodic? If yes give an example.

15
mywbut.com

Problem 3.12. In communication systems bandpass signals are often processed in digital processors.
To accomplish the processing the bandpass signal must first be converted from an analog signal to a
digital signal. For this problem assume this is done by ideal sampling. Assume the sampling frequency,
fs , is set at four times the carrier frequency.
a) Under what conditions on the complex envelope will this sampling rate be greater than the Nyquist
sampling rate (see Section 1.4.1) for the bandpass signal?
       
1 2 3 4
b) Give the values for the bandpass signal samples for xc (0), xc 4fc , xc 4fc , xc 4fc , and xc 4fc .

c) By examining the results in b) can you postulate a simple way to downconvert the analog signal
when fs = 4fc and produce xI (t) and xQ (t)? This simple idea is frequently used in engineering
practice and is known as fs /4 downconversion.
Problem 3.13. A common implementation problem that occurs in an I/Q upconverter is that the sine
carrier is not exactly 90◦ out of phase with the cosine carrier. This situation is depicted in Fig. 3.17.
a) What is the actual complex envelope, xz (t), produced by this implementation as a function of
x̃I (t), x̃Q (t), and θ?
b) Often in communication systems it is possible to correct this implementation error by preprocessing
the baseband signals. If the desired output complex envelope was xz (t) = xI (t) + jxQ (t) what
should x̃I (t) and x̃Q (t) be set to as a function of xI (t), xQ (t), and θ to achieve the desired complex
envelope with this implementation?

x̃ I (t )

+ xc (t )
2 cos(2πf ct ) ∑
-
π 2 +θ

x̃Q (t )

Figure 3.17: The block diagram for Problem 3.13.

Problem 3.14. A commercial airliner is flying 15,000 feet above the ground and pointing its radar
down to aid traffic control. A second plane is just leaving the √runway as shown in Fig. 3.18. The
transmitted waveform is just a carrier tone, xz (t) = 1 or xc (t) = 2 cos (2πfc t)
The received signal return at the radar receiver input has the form
√ √
yc (t) = AP 2 cos (2π(fc + fP )t + θP ) + AG 2 cos (2π(fc + fG )t + θG ) (3.24)
where the P subscript refers to the signal returns from the plane taking off and the G subscript refers to
the signal returns from the ground. The frequency shift is due to the Doppler effect you learned about
in your physics classes.

16
mywbut.com

Figure 3.18: An airborne air traffic control radar example.

a) Why does the radar signal bouncing off the ground (obviously stationary) produce a Doppler
frequency shift?

b) Give the complex baseband form of this received signal.

c) Assume the radar receiver has a complex baseband impulse response of

h(t) = δ(t) + βδ(t − T ) (3.25)

where β is a possibly complex constant, find the value of β which eliminates the returns from the
ground at the output of the receiver. This system was a common feature in early radar systems
and has the common name Moving Target Indicator as stationary target responses will be canceled
in the filter given in (3.25).

Modern air traffic control radars are more sophisticated than this problem suggests. An important point
of this problem is that radar and communication systems are similar in many ways and use the same
analytical techniques for design.
Problem 3.15. A baseband signal (complex exponential) and linear system are shown in Fig. 3.19.
The linear system has an impulse response of
1
hz (t) = √ 0≤t≤T
T
= 0 elsewhere (3.26)

a) Describe the kind of filter that hz (t) represents at bandpass.

b) What is the input power? Compute yz (t).

c) Select a delay, τd , such that arg [yz (t)] = 2πf0 (t − τd ) for all f0 .

d) How large can f0 be before the output power is reduced by 10dB compared to the input power?

Problem 3.16. The following bandpass filter has been implemented in a communication system that
you have been tasked to simulate:


 1 fc + 7500 ≤ |f | ≤ fc + 10000



 2 fc + 2500 ≤ |f | < fc + 7500
Hc (f ) = 4
fc ≤ |f | < fc + 2500


3


3
fc − 2500 ≤ |f | < fc

 4
0 elsewhere

17
mywbut.com

exp[ j 2πf0 t ] hz (t ) yz (t )

τd [
exp j 2πf0 (t − τ d )]

Figure 3.19: The block diagram for Problem 3.15.

You know because of your great engineering education that it will be much easier to simulate the system
using complex envelope representation.

a) Find Hz (f ).

b) Find HI (f ) and HQ (f ).

c) If xz (t) = exp(j2πfm t) compute yz (t) for 2000 ≤ fm < 9000.

3.7 Example Solutions

Problem 3.2.

a) Using sin(a − b) = sin(a) cos(b) − cos(a) sin(b) gives

xc (t) = sin(2πfc t) cos(2πfm t) − cos(2πfc t) sin(2πfm t). (3.27)

By inspection we have
−1 −1
xI (t) = √ sin(2πfm t) xQ (t) = √ cos(2πfm t). (3.28)
2 2
√
b) Recall xc (t) = xA (t) 2 cos(2πfc t + xP (t)) so by inspection we have
1 1 1
xz (t) = √ exp(j2πfm t) xI (t) = √ cos(2πfm t) xQ (t) = √ sin(2πfm t). (3.29)
2 2 2
√
b) Recall xc (t) = xA (t) 2 cos(2πfc t + xP (t)) so by inspection we have
1 1 1
xz (t) = √ exp(jφp ) xI (t) = √ cos(φp ) xQ (t) = √ sin(φp ). (3.30)
2 2 2

18