Optical Interconnections,Optical True-Time Delays, andMore... All Based on the White

Cell

Betty Lise AndersonThe Ohio State University

Introduction

What are optical true time delays andwhat are they for?What is an optical interconnection andwhat’s so special about OSU’s?What is a White cell, anyway?

Organization

Motivate true-time delay (TTD) work:phased array antennas» What TTD is» How other people do it

Explanation of the White cellAdapting the White cell to TTDExperimental results

Organization continued

Motivate optical interconnections (as ifthat’s needed)» How other people do it

Adapting the White cell to opticalinterconnectionsSummary and Conclusions

The White cell

Three spherical mirrors- all identicalLight bounces back and forth repeatedlyLight focused to a new spot on every bounceWe’ll eventually replace Mirror A with a spatial lightmodulator

A

B

C

in

out

CC (B)CC (C)

A Tale of Three Mirrors

Mirror AC

B

Center ofCurvature A

C.C. (B)

C.C. (C)

J. U. White, J. Optical Society of Am., 32,.285 (1942)

Key: mirrors refocus beam on every pass through cell

Device based on White Cell

A

C

B

input spot

first image

C.C.(B)

input turning mirror

White Cell - Operation

White Cell - continued

B

C

A

secondimage

first image

CC(C)

input spot

Sequence of Spot Images(for one input beam)

1

2

3

4

5

7

6

8 exits cell

enters cell0

•long optical path possible•no divergence•diffraction limited•compact

End View of Mirror

Spots on axis(as before)

C.C.(B)

C.C.(C)

0

1

2

3

4

5

6

8

7

Spots off axis

(different input beam)

input spot

output spot

Significance of spots

Spot patterndepends on inputlocation and MirrorB,C alignmentWill map spots topixels on SLM

0

1

2

3

4

5

6

8

7

Outputspot

Inputspot

Bring in a second light beam

0

2

3

4

5

6

8

3

5

0

2

4

6

8

input

1 1

77

output

anotherinput

anotheroutput

Complete Spot Set

N input spotsinput turning mirror

N output spots

outputturningmirror

Can add even more spots...

replace eachinput beam witha set of dots

The whole group bouncesaround, still striking uniquepixels

Summary of White cell

beams bounce back and forth a setnumber of times mm determined by alignment of mirrorsmany beams can circulate through cellat same timeNow, on to the problem we want tosolve…

Consider an array of antennas

One element alone produces a broad patternTheir signal add coherently to produce ahighly directional beamBeam emitted perpendicular to the array

To steer the radar beam

Can put the array on a stick and rotate itmechanically» It’s a pain to grease the bearings if the

array is on an airplane or worse a satellite

Use phase shifting to steer the beam

Phase shifting

In phase shifting, 2p isthe same as p

Will only get the rightphase shift for onefrequencyDifferent frequencies goin different directions(beam squint)Lousy for broadbandantennas

True-time delay

A delay of 6p corresponds to someactual timeIf delay all the signals by the right timesinstead of phases, works at allfrequenciesGreat for broadband antennas

Here’s one way

This could be switches and lengths of coax (ech tui)Could be optical switches and lengths of fiber» Modulate each light beam with an RF signal

» Delaying the light beam also delays the RF signal

There is a snag…

Some delays may be as long as 100’sof nsThat’s meters of coax or striplineHeavy, expensive, and temperaturesensitiveNaturally we want to do this optically…

But, need a set for every antennaelement in the array

That’s a lot ofhardwareSome arrays maycontain hundreds,even thousands ofelementsUgh.

Here’s a free space approach

liquid crystalswitch polarizing beamsplitters

See, for example, Riza, J. Lightwave Tech.. 12(6), pp. 1440-1447, 1994

Features

Free space doesn’t weigh muchEach switch can be a spatial lightmodulator, so can run multiple beams inparallelOur solution also is free space: theWhite cell

The White cell approach

Adapt the White cell to time delaysAlso a free-space approach“Hardware compressive”Avoids divergence problems sincebeams refocused on every bounceanywayWe’ll use liquid crystals and MEMs

Replace Mirror A with an SLM

Operation is the same opticallyReplace flat mirror with a spatial lightmodulatorSLM can be liquid crystal ormicroelectromechanical device

spherical mirror flat mirror lens

The SLM controls the path

Liquid crystal-based SLM

Polarizingbeamsplitter

Or using a MEM

Micromirrors tip to(in this case) twoanglesBeams are deflectedin either of twodirectionsPut this into a Whitecell

Replace Mirror A with MEM and lens

Let MEM be TI DMDCan form two White cellsPath length different for White cells but spotpattern is the same regardless of path

DMD

+20∞

-20∞

B

C

Elens

one White cell

another White cell

We call this a linear TTD cell

Number of delays is proportional to m:

where m is the number of bouncesNote number of bounces is fixed- set byspherical mirrors

Nm

=2

Linear cell apparatus

We built it

Time increment was 1 nsUsed pulsed laser (green) to measure delaysLoss ª 1.2 dB/bounce» Diffraction a big problem

– Pixels smaller (16 mm) than our beam

– We used a 50x50 pixel “macropixel”

– Get diffraction off interpixel gaps and holes in mirrors

» Mirrors are aluminum (gold has higher reflectivity)

Next design (using an LC)

B C

SLME

F’

F

PBS

foldingmirror

D

7D

“E” is the one’splaceCan count up to m/2by going to E (say, 6in 12 bounces)Make “F” longer by1 (the 7’s place)Can go there up tosix times

This is a quadratic cell

Number of delays goes as

Quadratic in m/2

Nm m m m m

= ÊË

ˆ¯

+ÊË

ˆ¯

+ ÊË

ˆ¯

= ÊË

ˆ¯

+ ÊË

ˆ¯2 2

12 2

22

2

Quadratic cell apparatus

We built it

Used Nd:YAG laser (1319 nm)Time delay increment 1 nsHad four input spots» Fiber array in silicon V-groove

Losses about 1 dB/bounceCrosstalk was lousy (we didn’t specifySLM correctly)

Want more delays

New architecture:re-image spots fromSLM onto a delayengineEach column ofspots has a delaytwice that ofprevious columnFor no delay that bit,go to a plain mirror

C

F

SLM

real mirror

imaginary mirror(replace with delayengine)

E

B

Assignment of columns

0 2 6 8 10121416

1357911131517

CC (B) CC (C)

...

..

MEM

DELAY ENGINE evenbounces land here

odd bounces land here

spots landing in this column get a delay of Dspots landing in this column get a delay of 2D

spots landing in this column get a delay of 4Dspots landing in this column get a delay of 8D

Example of delay engine

Glass slows lightImages at differentplane so longer pathEach block twice thelength of previousoneGood only for shortdelays

glass blocks

Imaginarymirror plane

reflectivesurfaces

Put that into the system

E F

C

B

SLM

PBS

For longer delays: lens train

conjugateplane CP5 CP4 CP3 CP2 CP1

III

III

IVV

IIIII

IVV

I

Reflective strips

entrance todelay engine

For longer delays: lens train

conjugateplane CP5 CP4 CP3 CP2 CP1

III

III

IVV

IIIII

IVV

I

Reflective strips

entrance todelay engine

We’re building that

Liquid crystal SLM1319 nm light7 bits in glass blocks (1 ps up to 128 ps)6 bits in lens train (512 ps up to 16.4 ns)

Glass block results

Used PMMA (faster, cheaper)Vendor unable to make without twisting,warpingHave not put resulting transporteraccident into system yet

The outcome

Lens train

6 bits means 18 lensesImplementation of strip mirrorsclever but made alignmentdarned hard» Have to offset, tilt mirror just right,

but lens goes with it» Ultimately had to separate them

(drat)

The apparatus

Lens Train

Got it all aligned at visible» Image SLM to every mirror plane» Image every field lens onto every other

Now re-aligning at infraredNo data yet

But we’re still looking ahead

Next: present some designs for higherorder cellsTwo flavors:» Polynomial (like quadratic), where Nµmx

» Exponential (like binary) where N µxm

New designs use MEM’s

Binary cell with MEM

In any two bounces, can visit delays or nullpath

Nm

= 2 2

FDelay element

MEM

Aulixiary Mirror

CB

E

Can do better

Two delay elements

First: Columns delay by D, 3D, 9D...

Second: Columns delay by 2D, 6D, 18D...

FDelay elements

MEM

Aulixiary Mirror

CB

E

“Ternary” cell

Nm

= 3 3

But, suppose a MEM had threestates

TI DMD can tilt to ±10°“Flat” position not stableBut, if it were, could have more choicesof paths on every bounce

So, with three-state MEM

delay engine F

MEMPlane

delay engine H

H

GFE

B

C

Assignment of columns…

32D8D2DDelayElementG

16D4D1DDelayElementE

Column mColumn3

Column2

Column1

4 21

m-Ê

ˈ¯ D

2 4 21

¥-Ê

ˈ¯

m

D

Nm

= 4 2 D

Can do higher order polynomialcells, too

2q

2q

MEMplane

C

B E

G

F

H

D 10D100D

1000D

This is a quartic cell

Number ofdelays goesas

2q

2q

MEMplane

C

B E

G

F

H

D 10D100D

1000D

Nm m m m

=-Ê

ˈ¯

+-Ê

ˈ¯

+-Ê

ˈ¯

+-Ê

ˈ¯

-2

44

24

62

44

24

14 3 2

Comparison of devices

100

10

102

103

104

105

20151050

Number of bounces

(m/4)2

(m/4)4

(m/2)2

2m/2

16m/5

8 m/4

5m/4

3m/3

4m/2

1

2

3

45

6

7

8

9

10

11

1213

14

15

16

Summary of key points

White cell has multiple bounce schemeRefocus each beam to spots on SLM pixelsCan switch a beam on any bounce

A

B

C

in

out

CC (B)CC (C)

White cell can support manybeams

One input spot foreach antennaelementCan support largephased arrays withone spatial lightmodulator

Got to keep m down

Lots of loss each bounce (ª1dB withcurrent SLM’s)» Almost all of loss due to SLM, not other

components» Need MEM’s with gold mirrors

Need to get a lot of delays for smallnumber of bouncesLed us to various designs

Advantages and disadvantages

Polynomial cells» Loss independent of

delay» Fewer components

to align» OK number of delays» May make more

delays for smallnumber of bounces

Exponential cells» Loss varies slightly

with delay» More complex

hardware-wise» Can get a boatload

of delays with a lot ofbounces