Demonstration of Negative Group

Delays in a Simple Electronic Circuit

Masao Kitano
kitano@kuee.kyoto-u.ac.jp, Room #1213

Kyoto University

Workshop on Quantum Optics, UCSB p.1/34

Group Delays in Circuits

Group delays in lumped systems (L = 0 or L
c/ )
Mitchell and Chiao: Am. J. Phys. 66, 14 (1998)
Bandpass Amplifier (LC + opamp)
Arbitrary Waveform Generator
Nakanishi, Sugiyama, and MK: quant-ph/0201001
(to appear in Am. J. Phys.)
Highpass Amplifier (RC + opamp)
Baseband pulse (No carriers)
Band limited signal from rectangular pulser +
lowpass filter

Workshop on Quantum Optics, UCSB p.2/34

Negative delay circuit

The output LED is lit earlier than the input LED.
Negative delay
Time constants could be order of seconds.
We can see it!

Workshop on Quantum Optics, UCSB p.3/34

Experiment
button

pulse

pulse generator

input

output

LED LED
negative delay circuit

1. rectangular pulse generator + low-pass filter

2. negative delay circuit

Workshop on Quantum Optics, UCSB p.4/34

Group delays ideal case

Group delay for base-band signals (delay time: td )
VOUT (t) = (h VIN)(t) = VIN (t td )
h(t) = (t td )
Fourier Transformed: VOUT ( ) = H( )VIN ( )

F h)( ) =

H( ) = (

Vin (t)

dt h(t)ei t = exp(i td )

H( )
td

Vout (t)

time

Workshop on Quantum Optics, UCSB p.5/34

Positive and Negative delays

td
>0
=0
<0

Causality
causal
(locally, mutually) causal
non causal

Physical realization
distributed system
lumped system
impossible

Positive delays are easy, if you have an appropriate

space.
Record and play is also possible.
No way to make ideal (unconditional) negative delays.
No lumped systems (L = 0) can produce ideal positive
or negative delays.

Workshop on Quantum Optics, UCSB p.6/34

Approximate delay with lumped systems

Ideal response function H( )
A( ) = |H( )| = 1,
( ) = arg H( ) = td
Approximate realization #1 with lumped systems
1 + i T
(1 pole, 1 zero)
H( ) =
1 i T
A( ) = 1 (flat response)

( ) = 2 tan1 T 2T
Stability condition T 0 (only positive delays)

Workshop on Quantum Optics, UCSB p.7/34

Approximate delay (2)

Approximate realization #2 with lumped systems
H( ) = 1 + i T (1 zero)
2


T
)
(
A( ) =
1 + ( T )2 1 +
2
( ) = tan1 T T

( distortion)

No sign restriction on T (positive and negative delay)

Workshop on Quantum Optics, UCSB p.8/34

Asymmetry positive / negative

Transfer function

positive delay T < 0

Im

negative delay T > 0

Im
zero

Re

H( ) = 1 + i T

zero

Im

Im

pole

zero

Re

1 + i T
H( ) =
1 i T

Re

zero

Re

pole

The arrows shows the direction of the phase increase of H( ).

Stability condition requires that poles can be only in the upperhalf plane, but zero can be anywhere.

Workshop on Quantum Optics, UCSB p.9/34

Negative Group Delay Circuit

Vin
V-

Vout
R
1/i C

(i C)1
1
Vout
V =
Vout =
1
R + (i C)
1 + i CR
Vin V for large gain of operational amplifier
Vout = (1 + i CR)Vin
A pole is converted into a zero by the feedback circuit.

Workshop on Quantum Optics, UCSB p.10/34

Finite bandwidth
Transfer Function
H( ) = 1 + i T (T = CR > 0)
A( ) = 1 + O( 2 T 2 ),

( ) = T + O( 3 T 3 )
Spectral condition for input signals
1
| | <
T

Workshop on Quantum Optics, UCSB p.11/34

Bandwidth
Works only for band-limited signals
otherwise outputs are distorted.
Phase ( )/degree

Amplitude A( )

2.5
2
1.5
1
0.5
0

80
60
40
20
0
-20
-40
-60
-80

-2 -1.5 -1 -0.5 0 0.5 1 1.5 2

-2 -1.5 -1 -0.5 0 0.5 1 1.5 2

Frequency T

Frequency T

Workshop on Quantum Optics, UCSB p.12/34

Band-Limit Circuit (Low-pass filter)

(1)R2
R2

Vin

R1

Vout

R1
C1
C1

Bessel filter (2nd order; m = 2)

HLP ( ) =
1 + i TLP (3 ) + (i TLP)2
TLP = R1C1 ,

= (1 + R3 /R2 ) = 1.268

Cutoff frequency: C = 0.7861/TLP

Workshop on Quantum Optics, UCSB p.13/34

Low-pass filters
A rectangular pulser and a series of lowpass filters (m stages)
are used to generate band-limited pulses.
....
Rectangular
Pulser

LPF 1

LPF 2

LPF m

Pushing the button (at t = 0) starts the event.

The band-limited output has a smooth leading edge.
A delay comparable to the pulse width is unavoidable.

Workshop on Quantum Optics, UCSB p.14/34

Low-pass filters(2)
0.6
m=2

amplitude /A.U.

0.5

6 8

0.4

10

0.3
0.2
0.1
0
0

3
4
time c t

Workshop on Quantum Optics, UCSB p.15/34

Circuit Diagram
Pulse generator
9V
10k

timer IC
7555 9V

R3

TRG DIS
OUT TH
RES VC

SW

Low-pass filters

R0
C0

20k

R1

R2
R1
C1

R3

TL082

R2
R1
R1 C1

C1

TL082
C1

Trec

input
Vin
LED
100

Negative delay circuit

C
10k

R C

C
R C

20k
TL082

output
Vout
LED
100

TL082

Workshop on Quantum Optics, UCSB p.16/34

Circuit parameters
Pulse generator
R0 6.8 M
C0 0.22 F
Trec 1.5 s
R1 2.2 M
C1 0.22 F
R2 10 k
R3 2.2 k
c 1.6 Hz

Negative delay circuit

R
C
R
C
T (= RC)

1 M
0.22 F
10 k
22 nF
0.22 s

Workshop on Quantum Optics, UCSB p.17/34

Circuit Board

Workshop on Quantum Optics, UCSB p.18/34

Experimental result
m = 4, n = 2
2.5

input

amplitude / V

2
1.5
1
0.5
0
-0.5
-1

time / s

Workshop on Quantum Optics, UCSB p.19/34

Experimental result
m = 4, n = 2
2.5

output

amplitude / V

input

1.5
1
0.5
0
-0.5
-1

time / s

Workshop on Quantum Optics, UCSB p.19/34

Interference in time domain

In time domain (cf. H( ) = 1 + i T )


dvin
d
(t)
vin (t) = vin (t) + T
vout (t) = 1 + T
dt
dt
Two terms interfere constructively at the leading edge and
destructively at the trailing edge.
vout
vin
dv /dt
1.2

exp(-x**2)
-2*x*exp(-x**2)
(1-0.2*x)*exp(-x**2)

0.8
0.6

in

0.4
0.2

0
-0.2
-0.4

constructive

-0.6

destructive

-0.8
-1
-4

-2

time

Workshop on Quantum Optics, UCSB p.20/34

All-passive circuit
i(t)
v(t)

R


d
1
d
1
1 + RC
i(t) = v(t) +C v(t) =
R
dt
R
dt


v(t)

This circuit gives negative group delays, if we consider v(t) as

the input and i(t) as the output.

Workshop on Quantum Optics, UCSB p.21/34

Unbalanced interferometer
Reflectivity of beam splitters: R = 1/2 ,

Delay: = 2L/c



d
Ein (t)
Edark (t) = (1R)Ein (t)REin (t ) 2 1 +
4 dt
1.2

input
bright
dark
dark (blow up)

dark port
input

amplitude

0.8
0.6
0.4

bright port
0.2

0
-0.2
-4

-3

-2

-1

= 0.08, = 0.15T , T : gaussian pulse width

0
t/T

Workshop on Quantum Optics, UCSB p.22/34

Cascading for larger advancement

Transfer function for n stages: H n ( ) = (1 + i T )n

2
2
(

T
)
n

T
)
n(
= 1+
, n ( ) n T
An ( ) 1 +
2
2
1.5

n=3

1.3

n=2

1.2
1.1

n=3

60

n=1

1
0.9

Phase arg[ ( )]

1.4

Amplitude A( )

80

n=2

40
20

n=1

0
-20
-40
-60

0.8

-80
-0.6

-0.4

-0.2

0.2

Frequency T

0.4

0.6

-0.6

Bandwidth is reduced as n increased.

-0.4

-0.2

0.2

Frequency T

0.4

0.6

Workshop on Quantum Optics, UCSB p.23/34

Cascading (2)

Usable bandwidth decreases as 1/ n, if T is given.

For
a given input pulse (Tw ), T = CR must be reduced as
Tw / n.

Total advancement scales 1/ n;

td n (Tw / n) = nTw
Order m of lowpass filters must be increased
accordingly;
m>n

Workshop on Quantum Optics, UCSB p.24/34

Cascading, n = 10

normalized amplitude

1.2
1
0.8

input

0.6
0.4
0.2
0
0

time / s

Workshop on Quantum Optics, UCSB p.25/34

Cascading, n = 10

normalized amplitude

1.2
1
0.8

output

input

0.6
0.4
0.2
0
0

time / s

Workshop on Quantum Optics, UCSB p.25/34

Problems in cascading
It is possible to increase the advancement as large as the
pulse width or more, but

The advancement increases slowly; 1/ n.

The order m of lowpass filters must be increased.
The system becomes very sensitive to noises. (A huge
gain outside of the band)

Workshop on Quantum Optics, UCSB p.26/34

Pushing the limit

Ttotal=sqrt(n)T

Total=nT
1

0.5

0.5

0.5
50

0.5
8

40
30

6
n

20
10

2
0

10

15

20

25

time t/sec

Simple-minded cascading

30

10

15

20

25

1/ n cascading

time t/sec

Workshop on Quantum Optics, UCSB p.27/34

Pushing the limit two pulse case

2.5
output(linear,n=7)
2
1.5
1
0.5
0
output(n=49)
0.5
1
input
1.5
2
0

10

15

20

25

30

35

time t/sec

Workshop on Quantum Optics, UCSB p.28/34

Out-of-band gain
Outside gain Significant obstacle toward large n.
Realistic transfer function (finite gain at = ):
1 + i T
H( ) =
,
1 + i T /

>1

( ) = (1 1 )T


2
1 + ( T )
1 ( = 0)

A( ) =
2
( = )
1 + ( T / )
For = 0.5, n = 50; An () = n = 250 = 1015 !

Workshop on Quantum Optics, UCSB p.29/34

Role of lowpass filters

Predictability Your move will be predicted if some
restrictions are imposed.
 m 
d x

JERK ... 
 dt m 
ACCEL.
LIMIT
SPEED LIMIT
LIMIT
0.11
10 0.034
35

Resemblance of functions A negative delay circuit

reveals the difference.
 j

j
d f d g


dist( f , g) = max 
0 jm  dt j
dt j 

Workshop on Quantum Optics, UCSB p.30/34

Role of filters (2)

On the condition m > n
3

x 10
6
0.5

n=8
n=9
n=10
n=11
n=12
n=13

n=8
n=9
n=10
n=11
n=12
n=13

4
time t/sec

8
0

0.02

0.04

0.06

0.08

0.1

0.12

time t/sec

Workshop on Quantum Optics, UCSB p.31/34

Causality
A

Pulser

LPF (m)

ND (n)

Causal relation in a casual sense

?
OABC
Causal relation in a strict sense
ABC

( A)

Workshop on Quantum Optics, UCSB p.32/34

Conclusion
A stand-alone (battery-operated) demonstration box
Battery operated
No extra equipment such as an oscilloscope or a
function generator needed
Large negative delays 0.5 s are achieved
experimentally.
25% of pulse width
Larger delay with cascading is possible (in priciple).
Advancement can be larger than the pulse width.
Filter stage plays an essential role.
The start of event (cause) is located before the
filters.

Workshop on Quantum Optics, UCSB p.33/34