ECE 463: Digital Communications Lab.

Lecture 9: OFDM I
Haitham Hassanieh
Previous Lecture:
ü Maximum Likelihood Decoding
ü QAM & QPSK
ü BER vs. SNR
ü Quantization Noise & AGCs

This Lecture:
q Multi-Carrier Modulation
q Orthogonal Frequency Division Multiplexing
 Digital Communication System
Mixer TX
Bits-to-Symbols
Bits Mapper
PA

Modulation Pulse DAC LPF BPF

(Encoding) Shaping

RX PLL

Mixer
Symbols-to-
Δ&' ℎ"# $ Bits Mapper Bits
LNA BPF LPF AGC ADC Matched Symbol Frame Demodulation
CFO Channel
Filter Timing Sync. Correction (Decoding)
Equalization
Recovery

PLL

Single Carrier Modulation

Symbols modulated on a single carrier frequency
( ) cos 2.&' $
 Single Carrier Modulation
2.5

1.5

Symbols modulated on a single carrier frequency

2
|H|
1
w band, convolving with the wireless channel reduces to multiplying by a single
er h and we can now write the received signal y(t) as: 0.5
• Low Spectral
y(t) Efficiency: sinc & raised cosine leakage
= hx(t) + n(t).
0
0 10 20 30 40 50 60 70
Tap Index

Channel: For wide band we can approximate Figure

the 5:wireless
Time Domain Wide
channel h(t)Band
by aChannel h(t) (≈11 tap
nel i.e. multiple delayed impulses as shown in Figure 5. For a k tap channel
gnal y(t) can be written as: • Frequency Selective Fading: Convolution with h(t) in the time dom
i=k
! ) in the frequency domain. For narrow band, h(t) is an
cation with H(f
! flat. For wide band, H(f ) results in different attenuation for different f
y(t) = h(i)s(t − iτ 6.
in Figure ) The figure also shows that for narrow bands the channel can
−# #
i=0 flat.
• ISI: Inter-Symbol-Interference limitsy(t)
performance
= h(t) ∗ s(t) + n(t) ⇔ Y (f ) = H(f )S(f ) + N
2.5

1.5
2
|H|

0.5

0
−80 −60 −40 −20 0 20 40 60 80
0 10 20 30 40 50 60 70
Frequency in MHz
Tap Index
 Selective Fading: Convolution with h(t) in the time domain results multipli-
Multi-Carrier Modulation
H(f ) in the frequency domain. For narrow band, h(t) is an impulse and H(f ) is
de band, H(f ) results in different attenuation for different frequencies as shown
Symbols
The figure modulated
also shows on multiple
that for narrow Sub-carrier
bands the channel frequencies
can be approximated as

y(t) = h(t) ∗ s(t) + n(t) ⇔ Y (f ) = H(f )S(f ) + N

• Divide spectrum into many narrow bands

! " = $ %& ' e)*+,-./

&

−80 −60 −40 −20 0 20 40 60 80

Frequency in MHz

Figure•6: Transmit symbols

Frequency on Fading
Selective different
forcarriers
100 MHz in narrow
channel bands
• Channel is Flat à No need to worry about ISI
bol-Interference: Multi-path results in inter-symbol-interference i.e. delayed
rfere with the symbol being decoding. The effect is sever and results in decoding
de band since the symbol length is short and of the order of the delayed taps. The
 Selective Fading: Convolution with h(t) in the time domain results multipli-
Multi-Carrier Modulation
H(f ) in the frequency domain. For narrow band, h(t) is an impulse and H(f ) is
de band, H(f ) results in different attenuation for different frequencies as shown
Symbols
The figure modulated
also shows on multiple
that for narrow Sub-carrier
bands the channel frequencies
can be approximated as

y(t) = h(t) ∗ s(t) + n(t) ⇔ Y (f ) = H(f )S(f ) + N

• Divide spectrum into many narrow bands

! " = $ %& ' e)*+,-./

&

0 " = $ ℎ& %& ' e)*+,-./

&

−80 −60 −40 −20 0 20 40 60 80

Frequency in MHz

Figure•6: Transmit symbols

Frequency on Fading
Selective different
forcarriers
100 MHz in narrow
channel bands
• Channel is Flat à No need to worry about ISI
bol-Interference: Multi-path results in inter-symbol-interference i.e. delayed
rfere with the symbol being decoding. The effect is sever and results in decoding
de band since the symbol length is short and of the order of the delayed taps. The
 Selective Fading: Convolution with h(t) in the time domain results multipli-
Multi-Carrier Modulation
H(f ) in the frequency domain. For narrow band, h(t) is an impulse and H(f ) is
de band, H(f ) results in different attenuation for different frequencies as shown
Symbols
The figure modulated
also shows on multiple
that for narrow Sub-carrier
bands the channel frequencies
can be approximated as

y(t) = h(t) ∗ s(t) + n(t) ⇔ Y (f ) = H(f )S(f ) + N

• Divide spectrum into many narrow bands

! " = $ %& ' e)*+,-./

&

0 " = $ ℎ& %& ' e)*+,-./

&

−80 −60 −40 −20 0 20 40 60 80

Frequency in MHz

Figure•6: Transmit symbols

Frequency on Fading
Selective different
forcarriers
100 MHz in narrow
channel bands
• Channel is Flat à No need to worry about ISI
bol-Interference: Multi-path results in inter-symbol-interference i.e. delayed
Not That Simple!
rfere with the symbol being decoding. The effect is sever and results in decoding
de band since the symbol length is short and of the order of the delayed taps. The
 Multi-Carrier Modulation
Symbols modulated on multiple Sub-carrier frequencies
• Divide spectrum into many narrow bands Guard Bands

! !

• Significant Leakage between adjacent subcarriers

• Need Guard Bands à Very inefficient!

Solution: Make the Sub-Carriers Orthogonal

 Multi-Carrier Modulation
Symbols modulated on multiple Sub-carrier frequencies

Make the Sub-Carriers Orthogonal

OFDM: Orthogonal Frequency Division Multiplexing

• Subcarriers are orthogonal: At the sub-carrier frequency, the

sampled value has zero leakage from other subcarriers.

• Subcarrier separation can be very small, for N subcarriers and

bandwidth B:
$
Δ" =
%
OFDM: Orthogonal Frequency Division Multiplexing

• Subcarriers are orthogonal: At the sub-carrier frequency, the

sampled value has zero leakage from other subcarriers.

• Subcarrier separation can be very small, for N subcarriers and

bandwidth B:
$
Δ" =
%

How to Achieve This?

OFDM: Orthogonal Frequency Division Multiplexing

Use DFT: Discrete Fourier Transform

2+3
1 -./ 1
+, 2 0
N-Point DFT: ! "# = ' ( ) *
&
145

2+3
-./0 1
N-Point IDFT: (()) = ' ! "# * ,
2
/0 45

Send symbols in Frequency Domain

! "# = 8 9 → Compute and transmit ( ) using IDFT
OFDM: Orthogonal Frequency Division Multiplexing

Send symbols in Frequency Domain

! "# = % & → Compute and transmit ( ) using IDFT

• *subcarrier à IDFT of length *

• Symbols % & can come from any modulation: BPSK, QPSK, QAM…

• ( ) is complex à need + & . à No point using PAM or ASK …

• OFDM Symbol: * samples of ( ) generated from the same

modulated symbols using IDFT.

• OFDM Symbol Time: / = */1 where 1 is the bandwidth.

• OFDM Frequency Bin Width: Δ" = 1// = 1/*

OFDM: Orthogonal Frequency Division Multiplexing

• OFDM invented long time ago but did not become viable until recently!
• DFT requires ! " # multiplications à very power hungry.
• FFT invented in 1965 requires ! " log " multiplications
• OFDM did not become viable until low power FFT chips came to market
 OFDM: Orthogonal Frequency Division Multiplexing
Mixer

Transmitter ℜ*{ }

Parallel to Serial
Serial to Parallel
DAC LPF BPF

IFFT
PLL
Bits Modulation
90 $ +
PA

ℑ.{ }
DAC LPF Mixer BPF

Receiver
Mixer

Parallel to Serial
Serial to Parallel
BPF LPF ADC %

FFT
PLL

90 $ + Demodulation Bits
LNA
&'

BPF Mixer LPF ADC

 y(t) = h(t) ∗ s(t) + n(t) ⇔ Y (f ) = H (f )S(f )
OFDM Symbol in Frequency Domain

−80
# −60 −40 −20 0
0
20 40 60
#
80

− Frequency in MHz
−1
2 2
6: Frequency Selective Fading for 100 MHz c
• FFT can be represented 0 to # − 1 or #/2 to #/2 − 1.
• OFDM Symbol
rference: created in digital baseband
Multi-path resultsà 0 bin
incorresponds to DC
inter-symbo
h the symbol being 4.5
decoding. 4.5
The effect is sev
ince the symbol 1 length./ 4 is 1
0123 short and of the ord
( 0 = *+ , - = * + , = 78
uss how we deal # with this #problem using OFD
362 362
g).
 y(t) = h(t) ∗ s(t) + n(t) ⇔ Y (f ) = H (f )S(f )
OFDM Symbol in Frequency Domain

−80
# −60 −40 −20 0
0
20 40 60
#
80

− Frequency in MHz
−1
2 2
6: Frequency Selective Fading for 100 MHz c
• FFT can be represented 0 to # − 1 or #/2 to #/2 − 1.
rference: Multi-path results in inter-symbo
• OFDM Symbol created in digital baseband à 0 bin corresponds to DC
h the symbol being decoding. The effect is sev
ince the
• DC of thesymbol length
circuits corrupts bits sentis short
on the 0 bin àand
Do notof
usethe
0 bin ord
uss how we deal with this problem using OFD
g).
 y(t) = h(t) ∗ s(t) + n(t) ⇔ Y (f ) = H (f )S(f )
OFDM Symbol in Frequency Domain
DC Bin

−80
" −60 −40 −20 0
0
20 40 60
"
80

− Frequency in MHz
−1
2 2
6: Frequency Selective Fading for 100 MHz c
• FFT can be represented 0 to " − 1 or "/2 to "/2 − 1.
rference: Multi-path results in inter-symbo
• OFDM Symbol created in digital baseband à 0 bin corresponds to DC
h the symbol being decoding. The effect is sev
ince the
• DC of thesymbol length
circuits corrupts bits sentis short
on the 0 bin àand
Do notof
usethe
0 bin ord
uss how we deal with this problem using OFD
g).
 y(t) = h(t) ∗ s(t) + n(t) ⇔ Y (f ) = H (f )S(f )
OFDM Symbol in Frequency Domain
DC Bin

−80
# −60 −40 −20 0
0
20 40 60
#
80

− Frequency in MHz
−1
2 2
6: Frequency Selective Fading for 100 MHz c
• Subcarriers orthogonal to each other but not to near by channels.
rference: Multi-path results in inter-symbo
• Need Guard Bins at sides of the channel à Transmit nothing there
h the symbol being decoding. The effect is sev
ince the symbol length is short and of the ord
uss how we deal with this problem using OFD
g).
 y(t) = h(t) ∗ s(t) + n(t) ⇔ Y (f ) = H (f )S(f )
OFDM Symbol in Frequency Domain
Guard Bins DC Bin Guard Bins

−80
# −60 −40 −20 0
0
20 40 60
#
80

− Frequency in MHz
−1
2 2
6: Frequency Selective Fading for 100 MHz c
• Subcarriers orthogonal to each other but not to near by channels.
rference: Multi-path results in inter-symbo
• Need
h the Guard Binsbeing
symbol at sides of the channel à Transmit
decoding. The nothing
effectthere is sev
ince the symbol length is short and of the ord
• Reduce Number of Guard band from # to 2 à Very Spectrally Efficient
uss how we deal with this problem using OFD
g).