Noise Properties of Optical Receiversusing Distributed Amplification

JOURNAL ^ON

VOIUlVI E XLVIII AUGUST 1997

MICROWAVE

OPTOELE,CTRONICS

Editorial ^{.. T Berceli}

^L

Optical millimeter-wave generation techniques

for broadband radio

access

networks ... R. A. Griffin,

P.

M. Lane and J. J. O'Reilly

²

Generation of high repetition rate optical pulse trains

using frequency quadrupling in a harmonically modeJocked fiber ring laser ... K. K. Gupta and D. Novak

⁹

Nonlinear travelling wave photodetector for millimeter-wave

harmonic frequency generatlon ^{L V} Ryjenkova, M. Alles and D. Jager

74

Optically controlled semiconductor coplanar-strip ^waveguide

attenuator/modulator: desigii considerations .. . ^{.. S.} Gevorgian and E. Kollberg

18

Applications of analog fiber-optic links . ^{C. H. Cox} il, E. L Ackerman, R. Helkey and G. E. Betts

²²

Frequency conversion methods by interferometer

and photodiode in microwave optical links . .. A. Hilt, G. Maury, B. Cabon, A. Vilcot and T Berceli

²⁶

Noise properties of optical receivers

using distributed ampliflcation ^{G. J6r6,} A. Hilt, A. ZOlomy and T Berceli

³²

Single mode fiber dis ersion in microwave optical

^system

using direct detection

^.

... ^T Marozsdk and

S.

Mihaly

³⁶

A continuous-time logarithmic photoreceptor cell

for parallel VLSI image processing . ^{M. Ol6h and} L. Liptdk-Fegd

39

NOISE PROPERTIES OF OPTICAL RECEIVERS

USING DISTRIBUTED AMPLIFICATION

G. JARO, A. HILT A. ZOLOMY and T BERCELI

BME-MHT TECHNICAL UNIVERSITY OF BUDAPEST DEPARTMENT OF MICROWAVE TELECOMMUNICAIIONS FI-1111 BUDAPEST GOLDMANN GYORGY TER 3., HUNGARY

TKI RT, INNOVATION COMPANY FOR TELECOMMUNICATIONS H-1142 BUDAPEST UNGVAR urcA 64-66., HUNGARY

ln wideband optical communications the bandwidth and the noise of optical receivers are crucial problems. The application of a distributed amplifier instead of a transimpedance amplifier in an optical receiver has

the

advantage

of

high speed and low noise

t1], p]

The characteristics of an optical receiver are not only determined by the parameters

of

the photodiode (PD) _and the distributed ampli{ier driven by the diode,

but

they are also greatly influenced by ^the matching circuit [3]. ^{ln this paper the gain and} the input equivalent noise current of the amplifier are compared

in

diflerent matching configurations. 0ne of the matching circuits has been experimentally verified as well.

1

II\lTRODUCTIOt\l

There is an increasing

demand

for high

speed digital

or

analog microwave

optical

communication

links (Fig. t).

tror

this purpose

optical

receivers

with

extreme bandwtdth and low noise are required _[4].

Fig.

L

^Microwave fiber optic link

The bandwidth of optical receivers is limited

^either

by the

physical parameters

of the photodiode or by

the external electricai

circuit [5].

The electrical

circuit

contains a microwave

amplifier

and a matching

circuit

(Fig. 2).

Fig. 2. High speed optical receiver

Distributed amplification is

a

very attractive solution

in

ultra-wideband signal

processing

[6]. ^{In this paper}

^the

effect

of

the matching

circuit

between

the

photodiode and the distributed

amplifier

is investigated.

2

FOUR STAGE DISTRIBUTED AI\nPLIFIER t\/lODEL

A

four-stage distributed

amplifier (DA)

matched

to

-50 Q

input and output

impedances

is modeled with

lumped elements. Fig. 3 shows the schematic

of

^the amp1ifier.

The

fie1d

effect transistors in the amplifier are

repre- sented

by gate and drain

capacitances,

and voltage

con-

trolled current

sources as shorvn

in Fig.

4.

4th stage

Fig. 3. Schematic of a four-stage distributed amplifier with lumped clem.ent.s

Fig. 4. Equivalent circuit model of the transistor

Inserting the transistor model

into

the

amplifler

ieads to an

LC

transmission

line

structure, as shown

in

Fig. -5.

Fig. 5. Lumped model of the four-stage DA

For typical FET element values, the gain of the

^am-

plifier is ,Szt : 14 dB, while the input and the output

reflections

(511 and

522) are smaller

than -20

^dB

ⁱⁿ

^the

frequenry

range

of 0-1-5 GHz. Asuming Cs : _0.2

pF,

Ca : 0.1 pF

and

g* : 50

^mS,

the value of -tn

^and

L 4

are

449

pH

and 322

pH,

respectively. Fig. 6 shows the equivalent

circuit

model

(in

this case

up to

1-5

GHz) of

the four-stage distributed

ampliher of

Fig. _-5.

c*

rtl

-

x

]v".

JJJ

32

30.3 ps 200mS

50 Q resistor. The receiver gain is

described

by

the transimpedance:

zr,:\!. zi, _[dBel

lin

lz,"l

:20 los' : -1o

Fig. 6. Equivalent circuit model of the distibuted ampffier

This

general

distributed amplifier

model

which

consists

of an ideal amplifier and a delay line

has

been

used to compare

the different photodiode

and

amplifier

intercon- nections based on

computer

simulations _[7].

3.

GAIt\J CHARACTERIZATIOt\l

To connect

the photodiode

and

the DA,

^{some possible} matching methods are indicated

in Fig. 7. For

a reference transimpedance

in the case of Fig. 7a, the

^amplifier

is driven by a combination of a current source and

a

The photodiode is modeled by a current source

and

a ^parallel

^0.-5

^pF capacitance (in all

^cases

the PD

is

substituted

with

these elements). The transimpedance gain

of the receiver has been calculated for each

case

as

a

function of frequenry and it is

presented

in Fig. 8.

^The

circuit

parameters

were

choosen

for maximally flat

^gain (except Case e

where the

matching

circuit

^{elements were}

choosen

for

maximally

flat input

impedance).

The

simplest

connection between the photodiode

and

the amplifier is the direct connection as it is shown in Fig. 7b. The gain at low frequency is 6 dB higher

than

in the

resistive generator case

but

at

high frequenry it

^has

about the

same value, due

to the

monotonous descent

of the

transimpedance.

Fig, 7. Different matching circuits

Using a

series

inductance (.t" : 625 pH) to

^connect

the

diode

to the amplifier (Fig. 7c),

the gain can be more

flat than in the direct connection,

as

it

is shown

in Fig.

8 Curve c.

The photodiode

connected

to the amplilier through

^a transmission

line is drawn in Fig. 7d. If ^the

^electrical

length

(O) of

the transmission iitre is small' the situation ^is similar

to

the inductive matching case (Fig.

7c)'

Supposing

Z I Zo: ^50Q

^and

^O

^{is great}

()

^tr _f

2),thete

are ripples

in

the gain

at

high frequencies.

It

is

not

depicted

in Fig'

8 because the shape of the curve is depending either on O or

on

the transmission

line

impedance

Z. ^The

strong ripples

in the

^response

of the complete optical receiver can

^be avoided by using a short line.

Fig.7e

shows a

7

section matching

with

a resistive load

(L^ : L" : 625 pH). The ^gain

^has

a peak

(around

9 GHz) but it ^{is 6 dB lower than the}

non-resistive matching.

Using an LC

^T-sectron

^(-L"1 : I2B7 pH, L,2 :

312 pH, Cp:0.35 ^{pF) for} connecting the diode

and

the amplifier

^shown

in Fig. 7f the ^gain

^can

be more flat

compared

to the other

non-resistive

matching

cases

but has a 6 dB grater transfer

impedance

than the

resistive matching.

From

the above calculation,

it

can be seen

that for

high gain the best solution is a non-resistive matching when the

PD

is connected

to

the

amplifler through anLC

T:section (Fig. 7f).

-1-1 VOLUME XLVIII AUGUST 1997

50

48

ldBol

46

44

42

40

**_\_

\ \

\

46

frequency IGIIz]

Fig. 8. Comparison of transimpedance gain

4.

t\lOISE PROPERTIES

To describe the signal

to

noise

ratio,

the equivalent

input

noise

current of

^the

optical

receiver was calculated. (Only

the

noise

arising from the electrical circuit is

calculated.

The noise generated

in

the photodiode (dark current, shot norse) is

only

an additive

term.) For

calculating

the input

noise

current

density,

the

noise sources

in the

dtstributed

amplifier shown in Fig. 3 were modeled. The

^thermal

noise generated

by the

resistors was modeled

by

a para11el

noise current source. If the resistor is

-50

O, the

noise

current is

18.2

pN\/Hz. The

noise equivalent

circuit of the

transistors comprises

two correlated

noise sources as

shown

in

Fig.

9

_[8].

Fig. 9. Noise model of the transistor

The current densities of these gate and drain

noise sources

were calculated using the

equations

of Van

der

Ziel:

..2 /_ 2

ilo

tJ(

: 4kToR -= _9^

^R

ih: ^{4kToBg^P}

where the

parameters

fi ^and P are varying with

drain

current. It is

shown

that the correlation coeflicient

can be

writen as: c : c, * i" ^ci : 0 + ^0.35i [1].

^Using

the above equations the values

of the

noise currents have been calculated

with

R.

: 0.2, P : 0.6

and

c : 0.35i.

Fig. 10. Noise equivalent of the ampffier

Using latter noise

models

the noise of the

four-stage distributed

amplifler

can be calculated.

The noise of

the

l0

amplifier

can be substituted

with a

correlated voltage and

current

noise source

pair

as shown

in Fig.

10.

The

calculated noise parameters

of the amplifier

as a

function of

the frequency are shown

in

Thble 1.

Table

l.

Calculated noise parameters of the

DA

Freq.

IGHz]

un

Ipvlt/nzl

1,-

bN'/n"1

Correlation

I

^257.067 18.669 0.226-0.369i

2 307.851 18.260 0.170-0.607i

3 374.163 17.595 0.115-0.732i

4 445.034 16.696 0.066-0.792i

5 513.896 15.600 0.020-0.817i 6 576.561 l4.J)() -0.024-0.816i 7 630.052 13.037 -0.072-0.794 8 672.ttO 1,1,.742 -0.127 -0.7 441 9 701,.049 10.608 -0.189-0.657i 10 715.809 9 807 -0.256-0.526i

Using the noise equivalent model of the

amplifier

(Fig. 10) the noise properties of the different

matching

circuits can be compared. Usually the noise of

optical receivers

are

described

with the equivalent input

noise

current

_[9].

Fig.

11 compares

the

calculated equivalent

input

noise

current

densities

of

the

photodiode -

distributed amplifier combinations

for the different matching circuits.

(Case d is omitted

for

the same reason as described

in

the previous section.)

As

it

can be seen that

in

the non-resistive generator case

the

noise

is lower

compared

to the

resistive

matching

(at

low frequenry

the difference is 3.3 dB).

In the

case

of

the

direct

connection

the

noise

current

is

shown

in Fig.

11 Curve

b. In the

inductive matching case (Fig. 7c)

the

shape

of the

noise curve has a

minimum but at

high frequencies

it

has a steep increase. This result was experimentally verified as well.

Using an

LCrT

section

the

noise

of the optical

receiver does

not

increase significantly

at high

frequencies

(Fig.

11

Curve

f)

compared

to

the indr.rctive matching (Fig. 7c).

4t)

30

vsa

20

10

0

0246810

frequency [Gl{z]

Fig. 11. Calculated equivalent input noise cunent densities case a.) case e.) ---t(-

case b.,1 ---O- case f.)

^o.a -\ +

---'i--'

=x-'-\-

x \ -\ _,/ ,/- *

>C

34

5.

EXPERIl\/lE1\lTAL II\JVESTIGATIOt\j

The

series

inductive matching

^(see

Fig. 7c) has

been experimentally

verified [10]. A hybrid integrated

optrcal recelver was designed using distributed

amplification.

The

receiver

consists

of a high speed photodiode and for simplicity only a two-stage distributed amplifier.

^The

distributed

amplifier

was designed

to

^provide

8

dB

gain in

a very broad band and a

low

noise figure.

The measurement set

up

consists

of

a

HP network

ana- lyzer

with

a lightwave test set extension and a spectrum an- alyzer. The optical

transmitter

contains a

DFB

(distributed feedback) laser along

with

an external

modulator.

The op- tical wavelength is

i.3

plm and the

modulation fiequenry

is swept

from

40

M}{z to

12

GHz. That

set up was used

for

cbarccler

uing

the o p tical/microwave trans

fer

^p erfo rmance

of

the optical receiver.

The

measured

responsivity of the opticai receiver

is relatively

flat from

40

MHz up to

10

GHz. The

measured equivalent noise

current

density

of the optical

receiver is depicted

in Fig.

12 showing

a low

noise

level. The

shape

of

the curve follows the

theoretical

one.

6.

COr\lCLUSIOt\l

Various matching circuits connecting a photodiode

to

a

distributed

amplif,er

have

been

ana\yzed and compared.

As a conclusion the optical receiver applying

resistive

natching has 6 dB

^less

amplification and greater

noise

REFEREl\lCES

[1]

^{C. S.} Aitchison: "The Intrinsic Noise Figure of the MESITET Distributed "Arnplifier",

IEEE

Tiansactions

on

Micrawave Theory and kchniques,

Vol.

33,

No.

6.

pp.

460-466, June 1985.

[2] A.

Z6lamy,

A" Hilt, A.

Baranyi,

G. Jdr6:

"Microwave Distributed

Amplifier in Hybrid

Integrated Technology", Proc. of the ECCTD'97, Hungary, Sept. 1997.

[3]

^{G. JArd,}

^A.

^Z6lomy,

T

Berceli. J. Ladvdnsz$,

A.

^Baranyi,

C.

S. Aitchison and

J. Y. Liang:

"Noise Minimization in Photodiode Driven Distributed Amplifiers", Proc. of the 25th European Microwave Conference , pp. 1,'/9-L84, Bologna, Italy, Sept. 1995.

[4] K.

Yang,

A. L.

^G.-Aitken

et al.:

"Design, Modelling, and Characterization of Monolithically Integrated ^InP-Based (1.-55

pm)

High-Speed

Qa ^Gbls)

p-i-n,41BT Front-End Photoreceivers' _, IEEE J. of Lighnuave kchn., YoL 14, No. B,

pp. 1831-18-19, Aug. 1996.

[5] A. Hilt, G. Jar6,

A-. Z6lomy,

B.

Cabon,

T

^{Berceli and}

Gribor Jdr6 received the

M. Sc.

degree

in

electrical engineering lrom the Techni- cal University

of

Budapest

in t994.

In 1994

he joined the

Departmenl

of

the

Microwave lblecommunication,

TUB, where he is working loward his Ph.D. de- gree. FIis research interest are in the ar-

L

^{eas of noise in} high speed optical receiver

,r.r'.':

^and optical system. millimeter-wave ^signal generatiorl in optical systenrs"

than

the

others.

Comparing

the

fi.ve analyzed circuits, the

best solution is the

non-resistive

LC Tsection (Fig.

7t) considering

both

the gain and noise performance.

30

[pAdHz]

20

246810

frequency

_[GI{z]

Fig. 12. Measured equivalent input",lr7u

r"o*t

density of the

7.

ACKt\lOWLEDGI\/lEl\lT

The authors thank 'OTK,A, the National

Scientific Research

Fund fbr continuous support with the

projects

No.

^T011295

and F024I13. This work was

performed

within

the frame

of the COPERNICUS No.

666-5 arrd the

FRANS

projects

of

the

European

Union.

T

Marozsdk: "Microwave Characterization

of

High SpeeC pin Photodiodes", Proc. of the COMI'L'I:'97. Czech Republic, October 1997.

[6] T Y.

Wong: "Fundamentals

of

Distributed Amplification", Artech Hotre _, Boston, London, 1993.

[7]

Helsinki University of Technology: "API-AC: Aaalysis Pro- gram for Linear Active Circuits", version 6.2, Espoo, Finland.

[8] A. Ambr6zy:

"Electronic Noise", Akaddmiai Konyvkiad6, Hungary, 1982.

[9] A.

K. Petersen,

E

Ebskamp, R. J. S. Pedersen,

X.

Zhang:

"Wide-band Low-Noise Distributed Front-End

for

Multi- Gigabit CPFSK Receivers", IEEE MTI]S Digest,

pp.

1375- 1,378,1994.

[10]

A.

Zolomy,

T

Berceli,

A. Hilt, G.

Jdrd, C. Aitchison. A.

Baranyi,

J.

Ladvdnszlry and

J. Y. Liang:

"Eight-Octave Bandwidth Optical Receiver Using Distributed Amplifica- tion", Proc. of the IEEE MTT:S llopical Meeting on Optical Microwave Interactions, Duisburg, Germany. Sept. 1997.

50

40

10

Attila Z6lomy received the M. Sc. degree in electrical engineering from the Technical University of Budapest in 1994. Now he is

working at the Department of Microwave Telecommunication

as a

^{Ph.D. student.}

His

research interest are

in

the field of wideband microwave distributed amplifiers, high speed photodetectors, millimeter-wave signal generalion in optical systems.

photograph and biography, see p. 1.

q

a\

ill

'-*'4

Attila

l{ilt

for ^a photograph and biography, see p. 30"

3.5

Tibor Berceli for a

VOLUME XLVIII AUGUST 1997