Direct and Coherent Detection

SIGNAL

E I
ph

DIRECT
beamsplitter
SIGNAL

E I ph
E
0

LOCAL
COHE RENT
OSCILLATOR

direct detection is linear in power Iph = σP

and quadratic in field amplitude: Iph = σA 〈 E 2〉 /2Z0

1
from:”Photodetectors”, by S.Donati, Prentice Hall 2000
 Coherent gain

In coherent detection, signal and local oscillator fields are:

E = E exp i(ωt+ϕ), E0 = E0 exp i(ω0t+ϕ0)

thus Iph= (σA/2Z0) [〈 E 2〉 +〈  E0 2〉 + 2 Re 〈 E0 E*〉 ] =

= (σA/2Z0) {E2 + E02+ 2 E E0 〈 cos [(ω -ω0)t+ϕ-ϕ0]〉 }
or, Iph = I + I 0 + 2µ √(I I0)
compared to I of direct detection, we find a
coherent gain Gcoh= Iph/I= 1+2µ √(I0 /I)

µ is the coherence factor. When ω=ω0 detection is called

homodyne, while if ω≠ω0 we have heterodyne detection.

2
from:”Photodetectors”, by S.Donati, Prentice Hall 2000
 Coherence factor

µ=〈 cos (ϕ-ϕ0)〉 ranges from µ= 0 (uncorrelated phases of signal and

local oscill ator) , to µ= 1 (complete correlation).

Now consider homodyne detection (ω=ω0) and write ϕ as the sum

of a mean 〈 ϕ〉 and a random part ϕr: ϕ = 〈 ϕ〉 +ϕr
Developing µ, µ = 〈 cos(〈 ϕ〉 +ϕr -ϕ0)〉
= cos(〈 ϕ〉 -ϕ0) 〈 cos ϕr〉 − sin(〈 ϕ〉 -ϕ0) 〈 sin ϕr〉 .
As 〈 ϕr〉 =0, also 〈 sin ϕr〉 =0 and if ϕr has a regular statistics. So:
µ = cos [〈 ϕ〉 -ϕ0] 〈 cos ϕr 〉 = cos ∆ϕ µΦ
Beating signal is multiplied by factor cos∆ϕ, that is, homodyne
detection is sensitive to the in-phase component with 〈 ϕ〉 =ϕ0; the
in-quadrature component with 〈 ϕ〉 = ϕ0+π/2 gives a zero output.

3
from:”Photodetectors”, by S.Donati, Prentice Hall 2000
 Phase fluctuations

The random part 〈 cosϕr〉 =µΦ describes relative phase fluctuations.

For ϕr small (<<1 rad), cosine is close to unity and its mean is ≈ 1;
for large ϕr (over 2π), cosine spans from -1 to + 1 and mean will be
≈ 0. For small ϕr <<1, developing cosine in series of ϕr:
µΦ = 〈 cos ϕ r〉 = 〈 1 - ϕr2/2! + ϕr4/4! +...〉 ≈ 1− σΦ 2/2
we see that µ is connected to phase variance σΦ .
+π +π

0 t 0 t

−π −π
small ϕ r large ϕr

1 1

0 t 0 t

-1 -1

<cos ϕr > ≈ 1 <cos ϕr > ≈0

4
from:”Photodetectors”, by S.Donati, Prentice Hall 2000
 Coherent S/N ratio
In direct detection, (S/N)2dir = I 2/ [2e(I+ Id)B + 4kTB/R]
and quantum limit is (S/N) 2dir/q = I /2eB for I>>Id+4kT/R.
In homodyne, signal is (σA/2Z0)2µEE0 , noise is sum of local
oscillator and signal shot-noises, plus Johnson noise of load:
[(σA/2Z )2µEE ] 2
______________________________
0 0
(S/N) 2hom =
2e[(σA/2Z0)(E2+E02)+Id] B+4kTB/R
4µ2I I 0
____________________
=
2e(I+I0+Id)B +4kTB/R
dividing by I0 and letting IR=2kT/eR,
4µ2 I
_________________
(S/N) 2hom =
[2e(1+(I+Id+I R)/I 0] B

5
from:”Photodetectors”, by S.Donati, Prentice Hall 2000
 Quantum limited S/N
for I0>>I0q=I+ Id+I R, the quantum limit is always reached:
(S/N)2hom /q = 4µ2I /2eB
In coherent detection the Q-L condition is on local oscillator
amplitude, not on signal amplitude as in direct detection. Making
local oscillator I0>>I+Id+IR large enough, Q-L is reached, even at
weak signal levels.
Heterodyne detection follows the same arguments, but beating
signal is now at the frequency ω-ω0, so that
Iph = 2√(I I 0 ) cos [(ω-ω0)t+〈 ϕ〉 -ϕ0] 〈 cosϕr〉
(S/N) 2het = 2µφ2I I 0 / [2e(Ι+Ι0+Id)B + 4kTB/R]
i.e., it has a modest penalty - a factor of 2 (or 3dB) respect to
homodyne but does not require the phase adjustment.
6
from:”Photodetectors”, by S.Donati, Prentice Hall 2000
 Condition for coherent detection
Requirements:
H phase matching of signal and local oscillator (for homodyne),
or beating will be reduced by a factor: cos(〈 ϕ〉 -ϕ0)
H phase coherence, or signal will be reduced by: 〈 cosϕr〉 =µφ.
H superposition of E and E0 on the PD with spatial coherence
or beating will be reduced by a factor:
µsp =∫AE(x, y) . E0*(x,y)dxdy /[∫AE(x, y)2dxdy ∫AE0(x,y)2dxdy ]1/2
H superposition of E and E0 with polarization matching or signal
is reduced by:
E.E0/EE0=µpol (E, E0 = Jones matrixes)
All previous expressions are generalized by using µφ →µφ µsp µpol

7
from:”Photodetectors”, by S.Donati, Prentice Hall 2000
 S/N, BER and photons /bit

N "1"

N
N/2 threshold S

1 threshold S
0 p(n) 0 "0"
p(01)

ideal direct detection: ideal coherent detection:

BER=e -N threshold at N/2
for BER= 10-9: N=10 p/b and P(01)=P(10)
(S/N)2=10

8
from:”Photodetectors”, by S.Donati, Prentice Hall 2000
 Photons per bit and modulation

• homodyne detection of amplitude modulated (ASK) signal:

BER = erfc N/2σN (N=number of photons per bit)
N=2(I .I0)1/2T/e=2(NsN0)1/2 σN=(2eI 0/2T)1/2T/e=N01/2
then, BER = erfc √Ns
and for BER=10-9 we get Ns=36 p/b
•homodyne detection of a phase-modulated PSK signal:
BER = erfc 2√Ns, and Ns=9 p/b
•heterodyne detection of a PSK-modulated signal:
BER= erfc√ 2Ns, and Ns=18
homodyne detection of a 4Φ−PSK modulated signal
BER = erfc 2√2Ns, and Ns=4.5

9
from:”Photodetectors”, by S.Donati, Prentice Hall 2000
 State-of-the-art receiver sensitivity

-20 x NEC88
x DIRECT 10 µ
AT&T 88 x NEC 89
O COHERENT x
FUJ 87x
RECEIVER SENSITIVITY (dBm)
x

RECEIVER SENSITIVITY (W)

x
-30 x SEL87 NTT 97 1µ

85
EC
x
er
mpl ifi

N
lpr ea bi t
-40 opt ic
a x NTT 92 36 ph / t 0.1µ
i
m s O ph/b
t e 0
t sy s O O NTT 89 1 / bi t
dir ec AT&T 85 4p
h
-50 SEL 88 O. NTT 85 10n
.O O.
AT&T 85 ts
BT 85 O l imi
n t
-60 er e 1n
c oh

-70 0.1n

0.1 0.4 1.0 2 4 10 20 40

TRANSMISSION RATE 1/T (Gbit/s)
OR TRANSMISSION BANDWIDTH 2B (GHz)

10
from:”Photodetectors”, by S.Donati, Prentice Hall 2000
 Balanced detector
I1
+
- S
E1

E I2

SIGNAL E
2

E
0
LOCAL
OSCILLATOR

I1 = (1/2)(I 0+I)+√(I 0I) sin(ϕ-ϕ0), I 2 = (1/2)(I 0+ I)-√(I 0I) sin(ϕ-ϕ0)

S = I 1-I2 = 2√(I 0I) sin (ϕ-ϕ0)

11
from:”Photodetectors”, by S.Donati, Prentice Hall 2000
 Beamsplitter phaseshift

At a beamsplitter, the continuity condition of electric fields at the separation boundary requires that
the incident Ei is always the sum of reflected Er and transmitted Et fields:
Ei = Et + Er (I)

where the underlines indicate rotating vectors. Also, in a l ossless beamsplitter power P is
unchanged upon splitting and, as P is proportional to E2, we have:

Ei2 = Et2 + Er2 (II)

To have both equations satisfied, the three vectors must lie on a right-angle triangle, as shown in
the fi gure below. Then, the angle - or phaseshift - be tween reflected Er and tr ansmitted Et vectors
is π/2 irrespective of the actual splitting ratio, while the angle ψ between incident and transmitted
fields increases from 0 to π/2 as Et decreases from Ei to 0 (or, R goes from 0 to 1). We can then
write, for the lossless beamsplitter:

_Er

_Er
_E i
_Et
E i
P E
_t
-
ψ

Et = √(1-R) Ei eiψ, Er = √R Ei ei(ψ−π/2)

For a lossy beamsplitter, Eq.(I) still applies, while (II) holds with the ≥ sign; then point P in
the figure shifts internal to the circle and the Er Et phaseshift becomes larger than π/2 (of an angle
p/2√[R(1-R)] where p is the loss).

12
from:”Photodetectors”, by S.Donati, Prentice Hall 2000
Balanced detectors with input subtraction

+V
bb
E
1

E
- S
2 +

- V bb

13
from:”Photodetectors”, by S.Donati, Prentice Hall 2000
 Coherent receiver with polarization diversity

+
GLAN -
cube BS
BS
balanced
SIGNAL detector

BS
GLAN
cube BS
∆f ≈
OSCILLATOR

+ DEMOD OUT
LOCAL

+ - balanced freq
contr f/V
detector CONV

14
from:”Photodetectors”, by S.Donati, Prentice Hall 2000
 Two -frequency heterodyne receiver

beamsplitter
or circulator
frequency-
ω stab. laser
two-frequency
source

transmitter ω+∆ω acoustooptical ω

and receiver modulator
telescope

+ -
balanced
detector
OUT

15
from:”Photodetectors”, by S.Donati, Prentice Hall 2000