Orthogonal Time Frequency Space (OTFS) Modulation

and Applications

Tutorial at SPCOM 2020, IISc, Bangalore, July, 2020

Yi Hong, Emanuele Viterbo, Raviteja Patchava

Department of Electrical and Computer Systems Engineering

Monash University, Clayton, Australia

Special thanks to
Tharaj Thaj, Khoa T.Phan

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 1 / 86

Overview I

1 Introduction
Evolution of wireless
High-Doppler wireless channels
Conventional modulation schemes (e.g., OFDM)
Effect of high Dopplers in conventional modulation

2 Wireless channel representation

Time–frequency representation
Time–delay representation
Delay–Doppler representation

3 OTFS modulation
Signaling in the delay–Doppler domain
Compatibility with OFDM architecture

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 2 / 86

Overview II

4 OTFS Input-Output Relation in Matrix Form

5 OTFS Signal Detection

Vectorized formulation of the input-output relation
Message passing based detection
Other detectors

6 OTFS channel estimation

Channel estimation in delay-Doppler domain
Multiuser OTFS

7 OTFS applications
SDR implementation of OTFS
OTFS with static multipath channels

Link to download Matlab code:

https://ecse.monash.edu/staff/eviterbo/OTFS-VTC18/OTFS_sample_code.zip

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 3 / 86

 Introduction

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 4 / 86

Evolution of wireless

Mobile 4G LTE
OFDMA
Mobile 3G
CDMA PS data, VOIP

Mobile 2G
TDMA Voice, SMS, PS data
transfer
Mobile 1G
Analog FDMA Voice, SMS, CS data
transfer
Voice, Analog traffic

1980s, N/A 1990s, 0.5 Mbps 2000s, 63 Mbps 2010s, 300 Mbps

Waveform design is the major change between the generations

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 5 / 86

High-Doppler wireless channels

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 6 / 86

Wireless Channels - delay spread

r2
Reflected path
r3

r1
LoS path

Delay of LoS path: τ1 = r1 /c

Delay of reflected path: τ2 = (r2 + r3 )/c
Delay spread: τ2 − τ1

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 7 / 86

Wireless Channels - Doppler spread

Reflected path

v cosθ
θ

LoS path
v
Doppler frequency of LoS path: ν1 = fc vc
Doppler frequency of reflected path: ν2 = fc v cos
c
θ

Doppler spread: ν2 − ν1

TX: s(t) RX: r (t) = h1 s(t − τ1 )e −j2πν1 t + h2 s(t − τ2 )e −j2πν2 t

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 8 / 86
Typical delay and Doppler spreads

Delay spread (c = 3 · 108 m/s)

∆rmax Indoor (3m) Outdoor (3km)

τmax 10ns 10µs

Doppler spread

νmax fc = 2GHz fc = 60GHz

v = 1.5m/s = 5.5km/h 10Hz 300Hz
v = 3m/s = 11km/h 20Hz 600Hz
v = 30m/s = 110km/h 200Hz 6KHz
v = 150m/s = 550km/h 1KHz 30KHz

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 9 / 86

Conventional modulation scheme – OFDM

OFDM - Orthogonal Frequency Division Multiplexing

Subcarriers

Frequency

OFDM divides the frequency selective channel into multiple parallel

sub-channels

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 10 / 86

OFDM system model

Figure: OFDM Tx

Figure: OFDM Rx
(*) From Wikipedia, the free encyclopedia
(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 11 / 86
OFDM time domain input-output relation

Received signal – channel is constant over OFDM symbol (no Doppler)

h = (h0 , h1 , · · · , hP−1 ) – Path gains over P taps
 
h0 0 ··· 0 hP−1 hP−2 ··· h1
 h1
 h0 ··· 0 0 hP−1 ··· h2 

 . .. .. .. .. .. .. ..
 .

 . . . . . . . .


 ..
 
.. .. .. .. .. .. 
 .
 . . . . . . hP−1  
r=h∗s=  .. .. .. .. .. .. ..  s
. . . . . .

hP−1 . 
 
 . .. .. .. .. .. .. .. 
 . . . . . . .
 . . 

 .. .. 
 
.. .. .. .. .. ..
 . . . . . . . . 
0 0 ··· hP−1 hP−2 ··· h1 h0
| {z }
M×M Circulant matrix (H)

Eigenvalue decomposition property H = FH DF where D = diag[DFTM (h)]

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 12 / 86

OFDM frequency domain input-output relation
At the receiver we have
P−1
X
r = Hs = FH DFs = hi Πi s
i=0
 
0 ··· 0 1
 .. 
1 . 0 0
where Π is the permutation matrix . . . . .. 

 .. . . . .
0 ··· 1 0
(notation used later as alternative representation of the channel)
At the receiver we have input-output relation in frequency domain

y = Fr = D x
|{z} where x = Fs and s| ={zFH x}
Diagonal matrix with subcarrier gains Tx IFFT

OFDM Pros
Simple detection (one tap equalizer)
Efficiently combat the multi-path effects
(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 13 / 86
Effect of high multiple Dopplers in OFDM

H matrix lost the circulant structure – decomposition becomes erroneous

Introduces inter carrier interference (ICI)

ICI

Frequency

OFDM Cons
multiple Dopplers are difficult to equalize
Sub-channel gains are not equal and lowest gain decides the performance

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 14 / 86

Effect of high Dopplers in OFDM

Orthogonal Time Frequency Space Modulation (OTFS)(∗)

Solves the two cons of OFDM
Works in Delay–Doppler domain rather than Time–Frequency domain

——————
(*) R. Hadani, S. Rakib, M. Tsatsanis, A. Monk, A. J. Goldsmith, A. F. Molisch, and R.
Calderbank, “Orthogonal time frequency space modulation,” in Proc. IEEE WCNC, San
Francisco, CA, USA, March 2017.

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 15 / 86

 Wireless channel representation

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 16 / 86

Wireless channel representation

Different representations of linear time variant (LTV) wireless channels

time-variant impulse response

g(t; τ )
F F

time-frequency SFFT delay-Doppler

response
(OFDM)
H(t; f ) h(τ; ν) response
(OTFS)

F F

B(ν; f )
Doppler-variant transfer response

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 17 / 86

Wireless channel representation

The received signal in linear time variant channel (LTV)

Z
r (t) = g (t, τ ) s(t − τ )dτ → generalization of LTI
| {z }
time-variant impulse response
Z Z
= h(τ, ν) s(t − τ )e j2πνt dτ dν → Delay–Doppler Channel
| {z }
Delay–Doppler spreading function
Z
= H(t, f ) S(f )e j2πft df → Time–Frequency Channel
| {z }
time-frequency response

Relation between h(τ, ν) and H(t, f )

Z Z 
h(τ, ν) = H(t, f )e −j2π(νt−f τ ) dtdf 


Z Z Pair of 2D symplectic FT
j2π(νt−f τ )
h(τ, ν)e

H(t, f ) = dτ dν 


(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 18 / 86

Wireless channel representation

2
1
Doppler 0
-1
-2
0 1 2 3 4
Delay

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 19 / 86

Wireless channel representation

2
1
Doppler 0
-1
-2
0 1 2 3 4
Delay

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 20 / 86

Time-variant impulse response g (t, τ )

—————
* G. Matz and F. Hlawatsch, Chapter 1, Wireless Communications Over Rapidly Time-Varying
Channels. New York, NY, USA: Academic, 2011

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 21 / 86

Time-frequency and delay-Doppler responses

SFFT
−−−→
←−−−
ISFFT

Channel in Time–frequency H(t, f ) and delay–Doppler h(τ, ν)

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 22 / 86

Time–Frequency and delay–Doppler grids

Assume ∆f = 1/T Delay

1
M ∆f M
Frequency
∆f M

2D SFFT

2 2D ISFFT
1
1 2 N Time
2
T
1
Channel h(τ, ν) 1 2 N Doppler
1
NT
P
X
h(τ, ν) = hi δ(τ − τi )δ(ν − νi )
i=1

1 1



Assume τi = lτi M∆f and νi = kνi NT

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 23 / 86

OTFS Parameters

Subcarrier M Bandwidth Symbol duration delay lτmax

spacing (∆f ) (W = M∆f ) (Ts = 1/W ) spread
15 KHz 1024 15 MHz 0.067 µs 4.7 µs 71 (≈ 7%)

Carrier N Latency Doppler UE speed Doppler kνmax

frequency (NMTs resolution (v ) frequency
(fc ) = NT ) (1/NT ) v
(fd = fc )
c
30 Kmph 111 Hz 1 (≈ 1.5%)
4 GHz 128 8.75 ms 114 Hz 120 Kmph 444 Hz 4 (≈ 6%)
500 Kmph 1850 Hz 16(≈ 25%)

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 24 / 86

 OTFS modulation

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 25 / 86

OTFS modulation

Time-Frequency Domain
x[k; l] X[n; m] Heisenberg s(t) Channel r(t) Wigner Y [n; m] y[k; l]
ISFFT h(τ; ν) SFFT
Transform Transform

Delay-Doppler Domain

Figure: OTFS mod/demod

Time–frequency domain is similar to an OFDM system with N symbols in a

frame (Pulse-Shaped OFDM)

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 26 / 86

Time–frequency domain

Modulator – Heisenberg transform

N−1
X M−1
X
s(t) = X [n, m]gtx (t − nT )e j2πm∆f (t−nT )
n=0 m=0

Simplifies to IFFT in the case of N = 1 and rectangular gtx

Channel
Z
r (t) = H(t, f )S(f )e j2πft df

Matched filter – Wigner transform

Z
0
Y (t, f ) = Agrx ,r (t, f ) , grx∗ (t 0 − t)r (t 0 )e −j2πf (t −t)
dt 0

Y [n, m] = Y (t, f )|t=nT ,f =m∆f

Simplifies to FFT in the case of N = 1 and rectangular grx

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 27 / 86

Time–frequency domain – ideal pulses
If gtx and grx are perfectly localized in time and frequency then they satisfy
the bi-orthogonality condition and

Y [n, m] = H[n, m]X [n, m]

where
Z Z
H[n, m] = h(τ, ν)e j2πνnT e −j2πm∆f
11
τ
dτ dν
Section 2.2: A discretized system model

Symbol
f

Subcarrier
···

2F
F
t
0

0 T 2T · · ·

Figure 2.1: Pulse-shaped OFDM interpretation of the signaling scheme (2.13). The
Figure: Time–frequency domain
shaded areas represent the approximate time-frequency support of the pulses gk,l (t).

————— the beginning of Section 2.2.2. Specifically, the channel coefficients h[k, l] inherit the
two-dimensional stationarity property of the underlying continuous-time system func-
* F. Hlawatsch and G. Matz, tion
Eds., Chapter
L (t, f ) [see 2, the
(2.2)]. Furthermore,
H Wireless
noise coefficientsCommunications
w[k, l] are i.i.d. CN (0, 1) as Over Rapidly
a consequence of the orthonormality of (g(t), T, F). These two properties are crucial for
Time-Varying Channels. NewtheYork,
ensuing analysis.
NY, USA: Academic, 2011 (PS-OFDM)
A drawback of (2.14) is the presence of (self-)interference [the second term in (2.14)],
which makes the derivation of capacity bounds involved, as will be seen in Section 2.4.
(Monash University, Australia) The signaling scheme (2.13)OTFS
can be interpreted
modulation as PS-OFDM [KM98], where the input SPCOM 2020, IISc 28 / 86
Signaling in the delay–Doppler domain
Time–frequency input-output relation

Y [n, m] = H[n, m]X [n, m]

where


nk
− ml
XX
H[n, m] = h [k, l] e j2π N M

k l

ISFFT
N−1 M−1
1 XX


nk
− ml
X [n, m] = √ x[k, l]e j2π N M

NM k=0 l=0

SFFT
N−1 M−1
1 XX


nk
− ml
y [k, l] = √ Y [n, m]e −j2π N M

NM n=0 m=0

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 29 / 86

Delay–Doppler domain input-output relation
Received signal in delay–Doppler domain
P
X
y [k, l] = hi x[[k − kνi ]N , [l − lτi ]]M
i=1
= h[k, l] ∗ x[k, l] (2D Circular Convolution)

1 1
0.8

0.8 0.8

0.6

0.6 0.6

0.4
0.4 0.4

0.2
0.2 0.2

0
0 0
0
0
1 5
2
5
3
20 10
4 10
10 5 9
15 6 8 15
7 7 15
15 6
10 8 20
5 10
9 4
20 10 25
5 3 5
2
25 1 30

(a) Input signal, x[k, l] (b) Channel, h[k, l] (c) Output signal, y [k, l]

Figure: OTFS signals

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 30 / 86

Fractional doppler effect

Actual Doppler may not be perfectly aligned with the grid

 
1
νi = (kνi + κνi ) , kνi ∈ Z, −1/2 < κνi < 1/2
NT

Induces interference from the neighbor points of kνi in the Doppler grid due
to non-orthogonality in channel relation – Inter Doppler Interference (IDI)
Received signal equation becomes
P Ni
e j2π(−q−κνi ) − 1
X X  
y (k, l) = hi 2π x [[k − kνi + q]N , [l − lτi ]M ]
i=1 q=−Ni Ne j N (−q−κνi ) −N

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 31 / 86

Compatibility with OFDM architecture

Time-Frequency Domain (N OFDM symbols)

x[k; l]
Precoder
X[n; m] OFDM s(t) Channel r(t) OFDM Y [n; m] Decoder
y[k; l]
(ISFFT) Modulator H(t; f ) Demodulator (SFFT)

Delay-Doppler Domain

Figure: OTFS mod/demod

OTFS is compatible with LTE system

OTFS can be easily implemented by applying a precoding and decoding
blocks on N consecutive OFDM symbols

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 32 / 86

OTFS with rectangular pulses – time–frequency domain

Assume gtx and grx to be rectangular pulses (same as OFDM) – don’t follow
bi-orthogonality condition

Time–frequency input-output relation

Y [n, m] = H[n, m]X [n, m] + ICI + ISI

ICI – loss of orthogonality in frequency domain due to Dopplers

ISI – loss of orthogonality in time domain due to delays

————
(*) P. Raviteja, K. T. Phan, Y. Hong, and E. Viterbo, “Interference cancellation and iterative
detection for orthogonal time frequency space modulation,” IEEE Trans. Wireless Commun., vol.
17, no. 10, pp. 6501-6515, Oct. 2018. Available on: https://arxiv.org/abs/1802.05242

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 33 / 86

 OTFS Input-Output Relation in Matrix Form

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 34 / 86

OTFS transmitter implementation: M = 2048, N = 128

time time
(128 symbols) (128 symbols)
IFFT IFFT
128 128
delay (M=2048)

delay (M=2048)
XMxN . XMxN . time
Q-QAM Q-QAM (128 symbols)

delay
delay
. … . …
MN*log2(Q) bits ISFFT MN*log2(Q) bits .
.
IFFT MxN IFFT
128 128
Doppler (N=128) Doppler (N=128)
FFT … FFT
2048 2048

time
(128 symbols)
(2048 subcarriers)
frequency

…
Heisenberg transform M>N TX complexity PAPR
P/S+CP
time-frequency -> time OTFS MN*log2(N) N
(N-symbol OFDM transmitter)
IFFT … IFFT OFDM MN*log2(M) M
2048 2048

P/S+CP Only
2048 samples one CP
…

Time domain signal (128 symbols, 2048 samples each)

OTFS transmitter implements inverse ZAK transform (2D→1D)

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 35 / 86

OTFS: Tx matrix representation

Transmit signal at 2D time domain: ISFFT+Heisenberg+pulse shaping on

delay–Doppler
S = Gtx FH H H
M (FM XFN ) = Gtx XFN
| {z }
ISFFT
In vector form:

s = vec(S) = (FH
N ⊗ Gtx )x

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 36 / 86

OTFS receiver implementation: M = 2048, N = 128
2048 samples
time varying …
channel
Time domain signal (128 symbols, 2048 samples each)

time FFT
(128 symbols) 128

delay (M=2048)
remove . YMxN
CP received

delay
… .
+ . Symbols
S/P
FFT
128
Doppler (N=128)

Received signal at delay–Doppler domain: pulse shaping+Wigner+SFFT on

time–frequency received signal
Y = FH
M (FM Grx R)FN = Grx RFN

In vector form:
y = (FN ⊗ Grx )r
OTFS receiver implements ZAK transform (1D→2D)
(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 37 / 86
OTFS: matrix representation – channel

Received signal in the time–frequency domain

Z Z
r (t) = h(τ, ν)s(t − τ )e j2πν(t−τ ) dτ dν + w (t)

Channel
P
X
h(τ, ν) = hi δ(τ − τi )δ(ν − νi )
i=1

Received signal in discrete form

P
X j2πki (n−li )
r (n) = hi e| {z
MN
}s([n − li ]MN ) + w (n), 0 ≤ n ≤ MN − 1
| {z }
i=1 Doppler Delay

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 38 / 86

OTFS: matrix representation – channel
Received signal in vector form

r = Hs + w

H is an MN × MN matrix of the following form

P
X
H= hi Πli ∆(ki ) ,
i=1

where, Π is the permutation matrix (forward cyclic shift), and ∆(ki ) is the
diagonal matrix
   j2πki (0) 
0 ··· 0 1 e MN 0 ··· 0
1 . . . 0 0
j2πki (1)
···
   
(ki )
 0 e MN 0 
Π = .

.
 , ∆ = 
. . .

 .. . . . . . . ..   .. .. ..
  

j2πki (MN−1)
0 · · · 1 0 MN×MN 0 0 ··· e MN
| {z } | {z }
Delay (similar to OFDM) Doppler

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 39 / 86

OTFS: matrix representation – channel
Received signal at delay–Doppler domain
y = (FN ⊗ Grx )H(FH
 
N ⊗ Gtx ) x + (FN ⊗ Grx )w
= Heff x + w
e
Effective channel for arbitrary pulses
Heff = (IN ⊗ Grx )(FN ⊗ IM )H(FHN ⊗ IM )(IN ⊗ Gtx )
rect
= (IN ⊗ Grx ) Heff (IN ⊗ Gtx )
| {z }
Channel for rectangular pulses (Gtx =Grx =IM )
Effective channel for rectangular pulses
P
X h i
Hrect hi (FN ⊗ IM )Πli (FH (FN ⊗ IM )∆(ki ) (FH

eff = N ⊗ IM ) N ⊗ IM )
i=1
| {z }| {z }
P(i) (delay) Q(i) (Doppler)
P
X P
X
= hi P(i) Q(i) = hi T(i)
i=1 i=1

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 40 / 86

OTFS: Example for computing Hrect
eff

M = 2, N = 2,MN = 4
1
li = 0 and ki = 0 (no delay and Doppler)
0

Πli =0 = I4 ⇒ P(i) = (F2 ⊗ I2 )(FH

2 ⊗ I2 ) = I4 0 1
(ki =0) (i)
∆ = I4 ⇒ Q = (F2 ⊗ I2 )(FH
2 ⊗ I2 ) = I4
(i) (i) (i)
T =P Q = I4 ⇒ Narrowband channel

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 41 / 86

OTFS: Example for computing Hrect
eff

1
li = 1 and ki = 0 (delay but no Doppler)
0
2 3
0 1 0 0 0 1
61 0 7 0 0
Πli =1 =4
6 7 ⇒ block circulant matrix with 2 × 2 (M × M) block size
0 05 1 0
0 0 0 1
2 3
0 1 0 0
61 0 0 0 7
P(i) = (F2 ⊗ I2 )Π(FH2 ⊗ I2 ) = 6
40 0 0 e−j2π 12 5
7

0 0 1 0
(using the block circulant matrix decomposition → generalization of circulant
matrix decomposition in OFDM)
∆(ki =0) = I4 ⇒ Q(i) = (F2 ⊗ I2 )(FH
2 ⊗ I2 ) = I4

T(i) = P(i) ⇒ T(i) s → circularly shifts the elements in each block (size M) of s
by 1 (delay shift)

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 42 / 86

OTFS: Example for computing Hrect
eff

1
li = 0 and ki = 1 (Doppler but no delay)
0

Πli =0 = I4 ⇒ P(i) = (F2 ⊗ I2 )(FH

2 ⊗ I2 ) = I4 0 1

1 0 0 0
2 3

(ki =1) 60 ej2π 41 0 0 7

∆ =4
6 2 7 ⇒ block diagonal matrix with 2 × 2 (M × M)
0 0 ej2π 4 0 5
3
0 0 0 ej2π 4
block size
2 3
0 0 1 0
60 j2π 41 7
0 0 e
Q(i) (1) H
= (F2 ⊗ I2 )∆ (F2 ⊗ I2 ) = 4
6 7
1 0 0 0 5
1
0 ej2π 4 0 0
(using the block circulant matrix decomposition in reverse direction)
T(i) = Q(i) ⇒ T(i) s → circularly shifts the blocks (size M) of s by 1 (Doppler
shift)

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 43 / 86

OTFS: Example for computing Hrect
eff

1
li = 1 and ki = 1 (both delay and Doppler)
0
2 3
0 1 0 0 0 1

(i)
61 0 0 0 7
P =6
40 −j2π 12 5
7
0 0 e
0 0 1 0
2 3
0 0 1 0
60 j2π 41 7
0 0 e
Q(i) =6
41
7
0 0 0 5
1
0 ej2π 4 0 0

T(i) = P(i) Q(i) ⇒ T(i) s → circularly shifts both the blocks (size M) and the
elements in each block of s by 1 (delay and Doppler shifts)

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 44 / 86

OTFS: channel for rectangular pulses

T(i) has only one non-zero element in each row and the position and value of
the non-zero element depends on the delay and Doppler values.
 k ([m−l ] )
−j2π Nn j2π i MNi M
e
 e , if q = [m − li ]M + M[n − ki ]N and m < li
k ([m−l ] )
T(i) (p, q) = e j2π i MNi M , if q = [m − li ]M + M[n − ki ]N and m ≥ li


0, otherwise.

Example: li = 1 and ki = 1
1
0 e j2π 4

0 0
0 0 1 0 
T(i) =
 1

0 e −j2π 4 0 0 
1 0 0 0

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 45 / 86

 OTFS Signal Detection

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 46 / 86

Vectorized formulation of the input-output relation

The input-output relation in the delay–Doppler domain is a 2D convolution

(with i.i.d. additive noise w [k, l])
XP
y [k, l] = hi x[[k − kνi ]N , [l − lτi ]M ] + w [k, l] k = 1 . . . N, l = 1 . . . M (1)
i=1

Detection of information symbols x[k, l] requires a deconvolution operation

i.e., the solution of the linear system of NM equations

y = Hx + w (2)

where x, y, w are x[k, l], y [k, l], w [k, l] in vectorized form and H is the
NM × NM coefficient matrix of (1).
Given the sparse nature of H we can solve (2) by using a message passing
algorithm similar to (*)
————
(*) P. Som, T. Datta, N. Srinidhi, A. Chockalingam, and B. S. Rajan, “Low-complexity
detection in large-dimension MIMO-ISI channels using graphical models,” IEEE J. Sel. Topics in
Signal Processing, vol. 5, no. 8, pp. 1497-1511, December 2011.

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 47 / 86

Message passing based detection

Symbol-by-symbol MAP detection

xb[c] = arg max Pr x[c] = aj y, H
aj ∈A
1

= arg max Pr y x[c] = aj , H
aj ∈A Q
Y

≈ arg max Pr y [d] x[c] = aj , H
aj ∈A
d∈Jc

Received signal y [d]

X
y [d] = x[c]H[d, c] + x[e]H[d, e] + z[d]
e∈Id ,e6=c
| {z }
(i)
ζd,c → assumed to be Gaussian

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 48 / 86

Messages in factor graph

Algorithm 1 MP algorithm for OTFS symbol detection

Input: Received signal y, channel matrix H
(0)
Initialization: pmf pc,d = 1/Q repeat
- Observation nodes send the mean and variance to variable nodes
- Variable nodes send the pmf to the observation nodes
- Update the decision
until Stopping criteria;
Output: The decision on transmitted symbols xb[c]

y[d] y[e1 ] y[eS ]

2 2
(µd;e1 ; σd;e 1
) (µd;eS ; σd;e S
)
pc;e1 pc;eS

x[e1 ] x[eS ] x[c]

fe1 ; e2 ; · · · ; eS g = Id fe1 ; e2 ; · · · ; eS g = Jc
Observation node messages Variable node messages

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 49 / 86

Messages in factor graph – observation node messages

y[d]

2 2
(µd;e1 ; σd;e 1
) (µd;eS ; σd;e S
)

Received signal
X x[e1 ] x[eS ]
y [d] = x[c]H[d, c] + x[e]H[d, e] + z[d]
e∈I(d),e6=c fe1 ; e2 ; · · · ; eS g = Id
| {z }
(i)
ζd,c → assumed to be Gaussian

Mean and Variance

Q
(i) (i−1)
X X
µd,c = pe,d (aj )aj H[d, e]
e∈I(d),e6=c j=1

 
2
Q Q
(i) (i−1) (i−1)
X X X
(σd,c )2 = pe,d (aj )|aj |2 |H[d, e]|2 − pe,d (aj )aj H[d, e]  + σ 2
 

e∈I(d),e6=c j=1 j=1

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 50 / 86

Messages in factor graph – variable node messages
y[e1 ] y[eS ]

pc;e1 pc;eS
Probability update with damping
factor ∆ x[c]
(i)
pc,d (aj ) =∆·
(i)
p̃c,d (aj ) + (1 − ∆) ·
(i−1)
pc,d (aj ), aj ∈A fe1 ; e2 ; · · · ; eS g = Jc

where
 
(i)
Y
p̃c,d (aj ) ∝ Pr y [e] x[c] = aj , H
e∈J (c),e6=d
Y ξ (i) (e, c, j)
= PQ (i)
e∈J (c),e6=d k=1 ξ (e, c, k)
2
 
(i)
 − y [e] − − He,c ak µe,c
ξ (i) (e, c, k) = exp  (i) 2 
(σe,c )

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 51 / 86

Final update and stopping criterion

Final update
Y ξ (i) (e, c, j)
pc(i) (aj ) = PQ (i)
e∈J (c) k=1 ξ (e, c, k)

xb[c] = arg max pc(i) (aj ), c = 1, · · · , NM.

aj ∈A

Stopping Criterion
Convergence Indicator η (i) = 1
NM  
1 X
η (i) = I max pc(i) (aj ) ≥ 0.99
NM c=1 aj ∈A

Maximum number of Iterations

Complexity (linear) – O(niter SQ) per symbol which is much less even
compared to a linear MMSE detector O((NM)2 )

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 52 / 86

Simulation results – damping factor ∆

100 45
OTFS, 120 Kmph OTFS, 120 Kmph

Average no. of iterations

40
10-1
35
10-2 4-QAM, SNR = 18 dB 4-QAM, SNR = 18 dB
30
BER

25
10-3
20
10-4
15
-5 10
10
0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1
∆ ∆
Figure: Variation of BER and average iterations no. with ∆. Optimal for ∆ = 0.7

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 53 / 86

Simulation results – OTFS vs OFDM with ideal pulses

100
OTFS, Ideal, 30 Kmph
OTFS, Ideal, 120 Kmph
4-QAM OTFS, Ideal, 500 Kmph
10-1 OFDM, 30 kmph
OFDM, 120 kmph
OFDM, 500 kmph
10-2
BER

10-3

10-4

10-5
5 10 15 20 25 30
SNR in dB

Figure: The BER performance comparison between OTFS with ideal pulses and OFDM
systems at different Doppler frequencies.

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 54 / 86

Simulation results – IDI effect

100
OTFS, 18 dB, 120 Kmph
OTFS, 18 dB, 500 Kmph
OTFS, 15 dB, 120 Kmph
10-1

4-QAM
10-2
BER

10-3

10-4

10-5
0 5 10 15 20
Ni

Figure: The BER performance of OTFS for different number of interference terms Ni
with 4-QAM.

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 55 / 86

Simulation results – Ideal and Rectangular pulses

100

10-1

10-2
BER

10-3 ×10-4
3.8
OTFS, Rect., WC, 30 Kmph
OTFS, Rect., WC, 120 Kmph
OTFS, Rect., WC, 500 Kmph
10-4 OTFS, Rect., WO, 30 Kmph
OTFS, Rect., WO, 120 Kmph
3.795 OTFS, Rect., WO, 500 Kmph
OTFS, Ideal
14.2 14.3 14.4 OFDM, 500 kmph
10-5
5 10 15 20 25 30
SNR in dB

Figure: The BER performance of OTFS with rectangular and ideal pulses at different
Doppler frequencies for 4-QAM.

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 56 / 86

Simulation results – Ideal and Rect. pulses - 16-QAM

100
OTFS, Rect., WC, 30 Kmph
OTFS, Rect., WC, 120 Kmph
16-QAM OTFS, Rect., WC, 500 Kmph
OTFS, Ideal
-1
10 OTFS, Rect., WO, 120 Kmph
OFDM
BER

10-2

10-3

10-4
10 15 20 25 30 35
SNR in dB
Figure: The BER performance of OTFS with rectangular and ideal pulses at different
Doppler frequencies for 16-QAM.

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 57 / 86

Simulation results – Low latency

10-1
OTFS, Rect., WC, 30 Kmph, N = 16, M = 128
OTFS, Rect., WC, 120 Kmph, N = 16, M = 128
OTFS, Ideal, N = 16, M = 128
OTFS, Ideal, N = 128, M = 512
OFDM, N = 16, M = 128
10-2
BER

16-QAM
-3
10

10-4
20 25 30 35 40
SNR in dB

Figure: The BER performance of OTFS with rectangular pulses and low latency
(N = 16, Tf ≈ 1.1 ms).

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 58 / 86

Matlab code
OTFS sample code.m

→ OTFS modulation – 1. ISFFT, 2. Heisenberg transform

X = fft(ifft(x).’).’/sqrt(M/N); % ISFFT
s mat = ifft(X.’)*sqrt(M); % Heisenberg transform
s = s mat(:);
→ OTFS channel gen – generates wireless channel
output: (delay taps,Doppler taps,chan coef)
→ OTFS channel output – wireless channel and noise
L = max(delay taps);
s = [s(N*M-L+1:N*M);s];% add one cp
s chan = 0;
for itao = 1:taps
s chan = s chan+chan coef(itao)*circshift([s.*exp(1j*2*pi/M...
*(-L:-L+length(s)-1)*Doppler taps(itao)/N).’;zeros(L,1)],delay taps(itao));
end
noise = sqrt(sigma 2/2)*(randn(size(s chan)) + 1i*randn(size(s chan)));
r = s chan + noise;
r = r(L+1:L+(N*M));% discard cp

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 59 / 86

Matlab code

→ OTFS demodulation – 1. Wiegner transform, 2. SFFT

r mat = reshape(r,M,N);
Y = fft(r mat)/sqrt(M); % Wigner transform
Y = Y.’;
y = ifft(fft(Y).’).’/sqrt(N/M); % SFFT

→ OTFS mp detector – message passing detector

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 60 / 86

Other detection methods

We will present a new low-complexity detection method at WCNC2020

on Tuesday in the Session T1-S7: Waveform and modulation
Output OTFS signal: y = Hx + w
1 MMSE detection:
−1
x̂ = (HH H + λI) HH y

Provides diversity but high complex O((NM)3 )

2 OTFS FDE (frequency domain equalization) in [1]
Equalization in time–frequency domain (one-tap) and apply the SFFT
Low complexity equalizer
Phase shifts can’t be applied and bad performance at high Dopplers
Small improvement on OFDM
——————–
[1]. Li Li, H. Wei, Y. Huang, Y. Yao, W. Ling, G. Chen, P. Li, and Y. Cai, “A simple two-stage
equalizer With simplified orthogonal time frequency space modulation over rapidly time-varying
channels,” available online: https://arxiv.org/abs/1709.02505.

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 61 / 86

Other detection methods
3 OTFS MMSE-PIC (parallel ISI cancellation) in [2]
First applies the equalization in time–frequency domain (one-tap) and then
applies successive cancellation with coding
Successive cancellation

ŷ(i)j+1 = y − Hx̂j + H(:, i)x̂(i)j

 
x̂(i)j+1 = arg min ŷ(i)j+1 − H(:, i)a
a∈A

Moderate complexity
Better performance than [1] but still struggles with the high Doppler
4 MCMC sampling [3]
Approximate ML solution using Gibbs sampling based MCMC technique
High complexity O(niter NM) compared to message passing (O(niter SQ))
(Does not take advantage of sparsity of the channel matrix)
——————–
[2]. T. Zemen, M. Hofer, and D. Loeschenbrand, “Low-complexity equalization for orthogonal
time and frequency signaling (OTFS),” available online: https://arxiv.org/pdf/1710.09916.pdf.
[3]. K. R. Murali and A. Chockalingam, “On OTFS modulation for high-Doppler fading
channels,” in Proc. ITA’2018, San Diego, Feb. 2018.
(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 62 / 86
 OTFS channel estimation

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 63 / 86

Channel estimation in time–frequency domain

(lτi , kνi ) ((delay,Doppler)) values are obtained from the baseband time
domain signal equation
P
X
y (t) = hi x(t − τi )e j2πνi (t−τi )
i=1

PN based pilots and 2D matched filter matrix is used to determine (lτi , kνi )
Highly complex

—————
1 A. Fish, S. Gurevich, R. Hadani, A. M. Sayeed, and O. Schwartz, “Delay-Doppler channel
estimation in almost linear complexity,” IEEE Trans. Inf. Theory, vol. 59, no. 11, pp.
7632-7644, Nov. 2013.
2 K. R. Murali, and A. Chockalingam, “On OTFS modulation for high-Doppler fading
channels,” in Proc. ITA’2018, San Diego, Feb. 2018.

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 64 / 86

Channel estimation using impulses in the delay-Doppler
domain
Each transmit and receive antenna pair sees a different channel having a
finite support in the delay-Doppler domain
The support is determined by the delay and Doppler spread of the channel
The OTFS input-output relation for pth transmit antenna and qth receive
antenna pair can be written as
M−1
X N−1  
X 1 k −n l −m
x̂q [k, l] = xp [n, m] hw , + vq [k, l].
m=0 n=0
MN qp NT M∆f

—————
1 P. Raviteja, K.T. Phan, and Y. Hong, “Embedded Pilot-Aided Channel
Estimation for OTFS in Delay-Doppler Channels”, IEEE Trans. on Veh. Tech.,
March 2019 (Early Access).
2 M. K. Ramachandran and A. Chockalingam, “MIMO-OTFS in high-Doppler
fading channels: Signal detection and channel estimation,” available online:
https://arxiv.org/abs/1805.02209.
3 R. Hadani and S. Rakib, “OTFS methods of data channel characterization and
uses thereof.” U.S. Patent 9 444 514 B2, Sept. 13, 2016.

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 65 / 86

If we transmit

xp [n, m] = 1 if (n, m) = (np , mp )

= 0 ∀ (n, m) 6= (np , mp ),

as pilot from the pth antenna, the received signal at the qth antenna will be

k − np l − m p
 
1
x̂q [k, l] = hw , + vq [k, l].
MN qp NT M∆f
1 k l


MN hwqp NT , M∆f and thus Ĥqp can be estimated , since np and mp are
known at the receiver a priori
Impulse at (n, m) = (np , mp ) spreads only to the extent of the support of the
channel in the delay-Doppler domain (2D convolution)
If the pilot impulses have sufficient spacing in the delay-Doppler domain, they
will be received without overlap

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 66 / 86

Figure: Illustration of pilots and channel response in delay-Doppler domain in a 2×1
MIMO-OTFS system

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 67 / 86

SISO OTFS system with integer Doppler

N −1 N −1 5 5 5 5 5 5 5 5 5 5 5

5 5 5 5 5 5 5 5 5 5 5

kp + 2kν 5 5 5 5 5 5 5 5 5 5 5

kp + kν 5 5 5 5 5 ⊞ ⊞ ⊞ 5 5 5
kp kp 5 5 5 5 5 ⊞ ⊞ ⊞ 5 5 5
kp − kν 5 5 5 5 5 ⊞ ⊞ ⊞ 5 5 5
kp − 2kν 5 5 5 5 5 5 5 5 5 5 5

1 1 5 5 5 5 5 5 5 5 5 5 5

0 0 5 5 5 5 5 5 5 5 5 5 5

0 1 lp − lτ lp lp + lτ M −1 0 1 lp − lτ lp lp + lτ M −1

(a) Tx symbol arrangement (: pilot; ◦: (b) Rx symbol pattern (O: data detection,
guard symbols; ×: data symbols)
: channel estimation)

Figure: Tx pilot, guard, and data symbols and Rx received symbols

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 68 / 86

SISO OTFS system with fractional Doppler

N −1 N −1 5 5 5 5 5 5 5 5 5 5 5

kp + 2kν + 2k^ 5 5 5 5 5 5 5 5 5 5 5

kp + 2kν kp + kν + k^ 5 5 5 5 5 ⊞ ⊞ ⊞ 5 5 5
kp + kν 5 5 5 5 5 ⊞ ⊞ ⊞ 5 5 5
kp kp 5 5 ⊞ ⊞ ⊞ 5 5 5
5 5 5

kp − kν 5 5 5 5 5 ⊞ ⊞ ⊞ 5 5 5
kp − 2kν kp − kν − k^ 5 5 5 5 5 ⊞ ⊞ ⊞ 5 5 5
kp − 2kν − 2k^ 5 5 5 5 5 5 5 5 5 5 5

0 0 5 5 5 5 5 5 5 5 5 5 5

0 1 lp − lτ lp lp + lτ M −1 0 1 lp − lτ lp lp + lτ M −1

(a) Tx symbol arrangement (: pilot; ◦: (b) Rx symbol pattern (O: data detection,
guard symbols; ×: data symbols)
: channel estimation)

Figure: Tx pilot, guard, and data symbols and Rx received symbols

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 69 / 86

MIMO OTFS system with fractional Doppler
N −1 ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕
⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕
^
kp + 2kν + 2k ⋄ ⋄ ⋄ ⋄ ⊕ ⊕ ⊕ ⊕
⋄ ⋄ ⋄ ⋄ ⊕ ⊕ ⊕ ⊕
kp ⋄ ⋄ ⋄ ⋄ ⊕ ⊕ ⊕ ⊕
⋄ ⋄ ⋄ ⋄ ⊕ ⊕ ⊕ ⊕
^
kp − 2kν − 2k ⋄ ⋄ ⋄ ⋄ ⊕ ⊕ ⊕ ⊕
⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕
0 ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⋄ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕
0 lp−lτ lp M −1 0 lp−lτ lp +lτ +1 M −1 0 lp−lτ lp +2lτ +2 M −1

(a) Antenna 1 (×: antenna 1 (b) Antenna 2 (♦: antenna 2 (c) Antenna 3 (⊕: antenna
data symbol) data symbol) 3 data symbol)

Figure: Tx pilot, guard, and data symbols for MIMO OTFS system (: pilot; ◦: guard)
N −1 5 5 5 5 5 5 5 5 5 5 5
5 5 5 5 5 5 5 5 5 5 5
5 5 5 5 5 5 5 5 5 5 5
^ 5
kp + kν + k 5 5 ⊞ ⊞ ⊠ ⊠ ⊗ ⊗ 5 5

kp 5 5 5 ⊞ ⊞ ⊠ ⊠ ⊗ ⊗ 5 5
^ 5
kp − kν − k 5 5 ⊞ ⊞ ⊠ ⊠ ⊗ ⊗ 5 5
5 5 5 5 5 5 5 5 5 5 5
5 5 5 5 5 5 5 5 5 5 5
0 5 5 5 5 5 5 5 5 5 5 5
0 lp lp +lτ +1 M −1

Figure: Rx symbol pattern at antenna 1 of MIMO OTFS system (O: data detection,

, , ⊗: channel estimation for Tx antenna 1, 2, and 3, respectively)

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 70 / 86

Multiuser OTFS system – uplink

N −1 ⋄ ⋄ ⋄ ⋄ ⋄
⋄ ⋄ ⋄ ⋄ ⋄
kp + 2kν + 2k^ ⋄ ⋄
⋄ ⋄
kp ⋄ ⋄
⊕ ⊕ ⊕ ⊕
kp − 2kν − 2k^ ⊕ ⊕ ⊕ ⊕
⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕
0 ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕ ⊕
0 lp−lτ lp M −1 0 lp−lτ lp +lτ +1 M −1 0 lp−lτ lp +2lτ +2 M −1

(a) User 1 (×: user 1 data (b) User 2 (♦: user 2 data (c) User 3 (⊕: user 3 data
symbol) symbol) symbol)

Figure: Tx pilot, guard, and data symbols for multiuser uplink OTFS system (: pilot; ◦:
guard symbols)

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 71 / 86

SISO-OTFS performance with the estimated channel
Simulation parameters: Carrier frequency of 4GHz, sub-carrier spacing of
15KHz, M = 512, N = 128, 4-QAM signaling, LTE EVA channel model
Let SNR p and SNR d denote the average pilot and data SNRs
Channel estimation threshold is 3σp , where σp2 = 1/SNR p is effective noise
power of the pilot signal
10-1
k̂ = 2
10-1
30 Kmph k̂ = 5
120 Kmph Full Guard
10-2
500 Kmph Ideal
10-2 Ideal

BER
10-3
BER

10-3 N = 128, M = 512, l τ = 20,

N = 128, M = 512, l τ = 20, SNRp= 50 dB, 4-QAM
10-4
SNRp= 40 dB, 4-QAM
10-4

10-5
10 12 14 16 18
10-5 SNRd in dB
10 12 14 16 18
SNRd in dB
(a) BER for estimated channels of different (b) BER for estimated channels of
Integer Dopplers Fractional Doppler

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 72 / 86

 OTFS applications

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 73 / 86

OTFS modem SDR implementation block diagram

Information
bits OTFS Modulation

x[k,l] X[n,m] s[n]

QAM Heisenberg Add cyclic Pulse Add

ISFFT
modulation Transform prefix Shaping Preamble
Channel
OTFS Demodulation

y[k,l] Y[n,m] r[n]

Preamble
Reciever
Channel Wigner Remove Detection and
SFFT Matched
Estimation Transform cyclic prefix Frame
Filter
Synchronization

Message
QAM Information
passing
demodulation bits
detector

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 74 / 86

Experiment setup and parameters
The wireless propagation channel can be observed in real time using LabView
GUI at the RX while receiving the OTFS frames.

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 75 / 86

OTFS received pilot in a real indoor wireless channel
DC Offset manifests itself as a constant signal in the delay-Doppler plane
shifted by Doppler equal to CFO.

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 76 / 86

OTFS received pilot in a partially emulated indoor mobile
channel
Doppler paths were added to the TX OTFS waveform and transmitted it into
a real indoor wireless channel for a time selective channel.

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 77 / 86

Error performance

OTFS Modem Performance OTFS vs OFDM (4-QAM)

100 100

10-1 10-1

10-2 10-2
BER / FER

BER
10-3 10-3

10-4 10-4
frame error rate(16-QAM) OTFS-static channel
bit error rate(16-QAM) OFDM-static channel
frame error rate(4-QAM) OTFS-mobile channel
bit error rate(4-QAM) OFDM-mobile channel
10-5 10-5
5 10 15 20 25 30 35 0 5 10 15 20 25 30 35
Tx gain in dB Tx gain in dB

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 78 / 86

OTFS with static multipath channels (zero Doppler)
Received signal

(size MN × 1) y = (FN ⊗ IM )H(FH

N ⊗ IM )x + w
e

↓ zero Doppler
(size M × 1) yn = H̆n xn + w
e n , for n = 0, · · · , N − 1

Equivalent to A-OFDM (asymmetric OFDM) in (*)

H̆n structure for M ≥ L 
h0 0 ···
n 
h1 e −j2π N
n
−j2π N
h1 h0 ··· h2 e 
H̆n =  .
 
.. .. ..
 ..

. . . 
0 ··· h1 h0 M×M

Achieves maximum diversity when M ≥ L (max. delay)

⇐⇒ N parallel CPSC transmissions each of length M
————
(*) J. Zhang, A. D. S. Jayalath, and Y. Chen, “Asymmetric OFDM systems based on layered
FFT structure,” IEEE Signal Proces. Lett., vol. 14, no. 11, pp. 812-815, Nov. 2007.
(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 79 / 86
OTFS with static multipath channels (zero Doppler)

100 M = 1, OFDM
N c = 1024, L = 72, 16-QAM M=2
M=4
M = 128, 256, 1024; MP
10-1 CPSC, MMSE
M = 128; A-OFDM, ZF
M = 128; A-OFDM, MMSE
AWGN
BER

10-2

10-3

10-4
15 20 25 30 35 40
SNR in dB

Figure: BER of OTFS for different M with MN = Nc = 1024, L = 72, and 16-QAM

————
(*) P. Raviteja, Y. Hong, and E. Viterbo, “OTFS performance on static multipath channels,”
IEEE Wireless Commun. Lett., Jan. 2019, doi: 10.1109/LWC.2018.2890643.
(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 80 / 86
References I

1 R. Hadani, S. Rakib, M. Tsatsanis, A. Monk, A. J. Goldsmith, A. F. Molisch,

and R. Calderbank, “Orthogonal time frequency space modulation,” in Proc.
IEEE WCNC, San Francisco, CA, USA, March 2017.
2 R. Hadani, S. Rakib, S. Kons, M. Tsatsanis, A. Monk, C. Ibars, J. Delfeld, Y.
Hebron, A. J. Goldsmith, A.F. Molisch, and R. Calderbank, “Orthogonal time
frequency space modulation,” Available online:
https://arxiv.org/pdf/1808.00519.pdf.
3 R. Hadani, and A. Monk, “OTFS: A new generation of modulation addressing
the challenges of 5G,” OTFS Physics White Paper, Cohere Technologies, 7
Feb. 2018. Available online: https://arxiv.org/pdf/1802.02623.pdf.
4 R. Hadani et al., “Orthogonal Time Frequency Space (OTFS) modulation for
millimeter-wave communications systems,” 2017 IEEE MTT-S International
Microwave Symposium (IMS), Honololu, HI, 2017, pp. 681-683.
5 A. Fish, S. Gurevich, R. Hadani, A. M. Sayeed, and O. Schwartz,
“Delay-Doppler channel estimation in almost linear complexity,” IEEE Trans.
Inf. Theory, vol. 59, no. 11, pp. 7632–7644, Nov 2013.

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 81 / 86

References II

6 A. Monk, R. Hadani, M. Tsatsanis, and S. Rakib, “OTFS - Orthogonal time

frequency space: A novel modulation technique meeting 5G high mobility and
massive MIMO challenges.” Technical report. Available online:
https://arxiv.org/ftp/arxiv/papers/1608/1608.02993.pdf.
7 R. Hadani and S. Rakib. “OTFS methods of data channel characterization
and uses thereof.” U.S. Patent 9 444 514 B2, Sept. 13, 2016.
8 P. Raviteja, K. T. Phan, Q. Jin, Y. Hong, and E. Viterbo, “Low-complexity
iterative detection for orthogonal time frequency space modulation,” in Proc.
IEEE WCNC, Barcelona, April 2018.
9 P. Raviteja, K. T. Phan, Y. Hong, and E. Viterbo, “Interference cancellation
and iterative detection for orthogonal time frequency space modulation,”
IEEE Trans. Wireless Commun., vol. 17, no. 10, pp. 6501-6515, Oct. 2018.
10 P. Raviteja, K. T. Phan, Y. Hong, and E. Viterbo, “Embedded delay-Doppler
channel estimation for orthogonal time frequency space modulation,” in Proc.
IEEE VTC2018-fall, Chicago, USA, August 2018.

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 82 / 86

References III

11 P. Raviteja, K. T. Phan, and Y. Hong, “Embedded pilot-aided channel

estimation for OTFS in delay-Doppler channels,” IEEE Transactions on
Vehicular Technology, May 2019.
12 P. Raviteja, Y. Hong, E. Viterbo, and E. Biglieri, “Practical pulse-shaping
waveforms for reduced-cyclic-prefix OTFS,” IEEE Trans. Veh. Technol., vol.
68, no. 1, pp. 957-961, Jan. 2019.
13 P. Raviteja, Y. Hong, and E. Viterbo, “OTFS performance on static
multipath channels,” IEEE Wireless Commun. Lett., Jan. 2019, doi:
10.1109/LWC.2018.2890643.
14 Li Li, H. Wei, Y. Huang, Y. Yao, W. Ling, G. Chen, P. Li, and Y. Cai, “A
simple two-stage equalizer with simplified orthogonal time frequency space
modulation over rapidly time-varying channels,” available online:
https://arxiv.org/abs/1709.02505.
15 T. Zemen, M. Hofer, and D. Loeschenbrand, “Low-complexity equalization
for orthogonal time and frequency signaling (OTFS),” available online:
https://arxiv.org/pdf/1710.09916.pdf.

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 83 / 86

References IV
16 T. Zemen, M. Hofer, D. Loeschenbrand, and C. Pacher, “Iterative detection
for orthogonal precoding in doubly selective channels,” available online:
https://arxiv.org/pdf/1710.09912.pdf.
17 K. R. Murali and A. Chockalingam, “On OTFS modulation for high-Doppler
fading channels,” in Proc. ITA’2018, San Diego, Feb. 2018.
18 M. K. Ramachandran and A. Chockalingam, “MIMO-OTFS in high-Doppler
fading channels: Signal detection and channel estimation,” available online:
https://arxiv.org/abs/1805.02209.
19 A. Farhang, A. RezazadehReyhani, L. E. Doyle, and B. Farhang-Boroujeny,
“Low complexity modem structure for OFDM-based orthogonal time
frequency space modulation,” in IEEE Wireless Communications Letters, vol.
7, no. 3, pp. 344-347, June 2018.
20 A. RezazadehReyhani, A. Farhang, M. Ji, R. R. Chen, and B.
Farhang-Boroujeny, “Analysis of discrete-time MIMO OFDM-based
orthogonal time frequency space modulation,” in Proc. 2018 IEEE
International Conference on Communications (ICC), Kansas City, MO, USA,
pp. 1-6, 2018.
(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 84 / 86
References V

21 P. Raviteja, Y. Hong, E. Viterbo, E. Biglieri, “Effective diversity of OTFS

modulation,” IEEE Wireless Communications Letters, Nov. 2019.
22 Tharaj Thaj, Emanuele Viterbo, “OTFS Modem SDR Implementation and
Experimental Study of Receiver Impairment Effects,” 2019 IEEE International
Conference on Communications Workshops (ICC 2019), Shanghai.
23 Tharaj Thaj and Emanuele Viterbo, “Low Complexity Iterative Rake Detector
for Orthogonal Time Frequency Space Modulation” in Proceedings of WCNC
2020, Seoul.

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 85 / 86

 Thank you!!!

Questions?

(Monash University, Australia) OTFS modulation SPCOM 2020, IISc 86 / 86