4/1/2009

5_5 The Theory of Small Reflections

1/1

5.5 The Theory of Small Reflections

Reading Assignment: pp. 244-246
An important and useful approximation when considering multi-section matching networks is the Theory of Small Reflections. HO: THE THEORY OF SMALL REFLECTIONS EXAMPLE: THE THEORY OF SMALL REFLECTIONS The Theory of Small Reflections provides a simpler mathematical form for analyzing the frequency response of many microwave devices. HO: THE FREQUENCY RESPONSE OF THE QUARTER-WAVE MATCHING NETWORK We can also use the Theory of Small Reflections to provide an approximate analysis of a multi-section impedance transformer (i.e., multi-section matching network). HO: THE MULTI-SECTION TRANSFORMER

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Theory of Small Reflections

1/9

The Theory of Small Reflections

Recall that we analyzed a quarter-wave transformer using the multiple reflection view point.
V +(z ) = a Z 0 e j
(z + )

Z0
V ( z ) = b Z 0 e + j
(z + )

Z1 =

Z 0RL
=

RL

4 We found that the solution could thus be written as an infinite summation of terms (the propagation series):

b = a pn
n =1

where each term had a specific physical interpretation, in terms of reflections, transmissions, and propagations. For example, the third term was path:

Z0

Z 0RL

RL

p3 a

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Theory of Small Reflections

2/9

a1

b2

a5

e j
L

p3 = 2 ( L ) e j 2
2

b1

a2

e j

b5

Now lets consider the magnitude of this path:

p3 = L e j 2
= L
2 2

Recall that = L for a properly designed quarter-wave transformer :

=

and so:
2

RL Z 1 = L RL + Z 1
2 2 3

p3 = L = L

For the case where values RL and Z 1 are numerically close in i.e., when:

RL Z 1

RL + Z 1

we find that the magnitude of the reflection coefficient will be very small: R Z1 L = L 1. 0 RL + Z 1 As a result, the value L
Jim Stiles
3

will be very, very, very small.

Dept. of EECS

The Univ. of Kansas

4/1/2009

The Theory of Small Reflections

3/9

Moreover, we know (since the connector is lossless) that:

1 = + = L +
2 2 2 2

and so:
= 1 L
2

We can thus conclude that the magnitude of path p3 is likewise very, very, very small:

p3 = L L

This is a classic case where we can approximate the propagation series using only the forward paths!! Recall there are two forward paths:

p2 p1
Z 0RL
RL

Z0

= +2 L e j 2 a

( p1 + p2 ) a

e j
L

p1 =

e j

p2 = 2 e j 2
Jim Stiles The Univ. of Kansas Dept. of EECS

4/1/2009

The Theory of Small Reflections

4/9

Therefore IF Z0 and RL are very close in value, we find that we can approximate the reflected wave using only the direct paths of the infinite series:

b
Therefore:

= +2 L e j 2 a

( p1 + p2 ) a

V ( z ) = b Z 0 e + j (z +

+2 L e j 2 a Z 0 e + j (z +

Now, if we likewise apply the approximation that

1.0 , we

conclude for this quarter wave transformer (at the design frequency): b ( p1 + p2 ) a
= + L e j 2 a

Therefore:

V ( z ) = b Z 0 e + j (z +

) )

+ L e j 2 a Z 0 e + j (z +

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Theory of Small Reflections

5/9

This approximation, where we: 1. use only the direct paths to calculate the propagation series, 2. approximate the transmission coefficients as one (i.e., = 1 ). is known as the Theory of Small Reflections, and allows us to use the propagation series as an analysis tool (we dont have to consider an infinite number of terms!).

Consider again the quarter-wave matching network SFG. Note there is one branch ( = S22 of the connector), that is not included in either direct path.

e j
L

p1 =

e j

p2 = 2 e j 2

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Theory of Small Reflections

6/9

With respect to the theory of small reflections (where only direct paths are considered), this branch can be removed from the SFG without affect.

e j
L

p1 =

e j

p2 = 2 e j 2
Moreover, the theory of small reflections implements the approximation = 1 , so that the SFG becomes:

1 .0

e j
L

p1 = p2 = e j 2

1 .0

e j

Reducing this SFG by combining the 1.0 branch and the e j branch via the series rule, we get the following approximate SFG:

e j
L

in =

= + L e j 2

b a

e j

The approximate SFG when applying the theory of small reflections!

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Theory of Small Reflections

7/9

Note this approximate SFG provides precisely the results of the theory of small reflections! Q: Why is that? A: The approximate theory of small reflections SFG Contains all of the significant physical propagation mechanisms of the two forward paths, and only the two significant propagation mechanisms of the two forward paths. Namely: 1. The reflection at the connector (i.e., ). 2. The propagation down the quarter-wave transmission line (e j ), the reflection off the load ( L ), and the propagation back up the quarter-wave transmission line ( e j ).

p2 p1 p2
e j

Z0

Z 0RL

RL

a
The approximate SFG when applying the theory of small reflections!

p1 b

e j

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Theory of Small Reflections

8/9

a
From series rule From parallel rule

a
L e j 2 + L e j 2

however, provides the approximate result:

in + L e j 2

Q: But wait! The quarter-wave transformer is a matching network, therefore in = 0 . The theory of small reflections,

Is this approximation very accurate? How close is this approximate value to the correct answer of in = 0 ?
A: Lets find out! Recall that = L for a properly designed quarter-wave matching network, and so:
in + L e j 2 = L 1 + e j 2

Likewise,

= 4 (but only at the design frequency!) so that:

2 2 = 2 = 4

where you of course recall that = 2 !

Jim Stiles The Univ. of Kansas Dept. of EECS

4/1/2009

The Theory of Small Reflections

9/9

Thus:

in L 1 + e j 2 = L =0
j

( ) (1 + e )
!!!

= L (1 1 )

Q: Wow! The theory of small reflections appears to be a

perfect approximationno error at all!?!

A: Not so fast.

The theory of small reflections most definitely provides an approximate solution (e.g., it ignores most of the terms of the propagation series, and it approximates connector transmission as = 1 , when in fact 1 ). As a result, the solutions derived using the theory of small reflections willgenerally speakingexhibit some (hopefully small) error. We just got a bit lucky for the quarter-wave matching network; the approximate result in = 0 was exact for this one case!

The theory of small reflections is an approximate analysis tool!

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

Example Theory of Small Reflections

1/3

Example: The Theory of Small Reflections

Use the theory of small reflections to determine a numeric value for the input reflection coefficient in , at the design frequency 0 .

= 30 8

= 0 8

Z0

in

Z1

Z2

ZL

0 = 0.1

1 = 0.05

L = 0.15

Note that the transmission line sections have different lengths!

Solution
Applying the theory of small reflections, the approximate signal flow graph of the structure becomes:

a0

a1 e
0
j
1

a2 e
j
2

e
L

=e

2 3 j 8

=e

3 j 4

b0
Jim Stiles

=e

2 j 8

=e

j 4

b1

b2
The Univ. of Kansas Dept. of EECS

4/1/2009

Example Theory of Small Reflections

2/3

Note there are three direct propagation paths:

Path 1
1

= 30 8

= 0 8

in 1

Z1

Z2

ZL

0 = 0.1

a0

a1 e j
0
1

a2 e j
2

p1 = 0 = 0.1

b0
Path 2

e j

e j

b1

b2

This path includes propagation down and back a transmission line length 1 ! = 30 8 = 0 8 1 2

in 2

Z1

Z2
1 = 0.05

ZL

p2 = e j 1 e j
1

1 3

a0
4

a1 e j
0
1

a2 e j
2

= e j = e j

0.05e j 0.05

= + j 0.05
Jim Stiles

b0

e j

e j b1

The Univ. of Kansas

b2

Dept. of EECS

4/1/2009

Example Theory of Small Reflections

3/3

Path 3

This path includes propagation down and back transmission line lengths of 1 + 2 !
1

= 30 8

= 0 8

in 3

Z1

Z2

ZL
L = 0.15

p3 = e j (

1+ 2

L e j (

1+ 2

a0

a1 e j
0
1

a2 e j
2

= e j 0.15 e j = e j 2 0.15 = 0.15

b0

e j

e j b1

b2

Thus, using the theory of small reflections we can determine approximately the input reflection coefficient:

in =

b0 a0 = p1 + p2 + p3 = 0.1 + j 0.05 + 0.15 = 0.25 + j 0.05

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Frequency Response of a Quarter

1/9

The Frequency Response of a Quarter-Wave Matching Network

perhaps useless information. The quarter-wave matching network has an exact SFG of:
a

Q: You have once again provided us with confusing and

e j
L

e j

Using our reduction rules, we can quickly conclude that:

in 2 L e j 2 b =+ a 1 L

You could have left this simple and precise analysis alone BUT NOOO!! You had to foist upon us a long, rambling discussion of the propagation series and direct paths and the theory of

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Frequency Response of a Quarter

2/9

small reflections, culminating with the approximate (i.e., less accurate!) SFG: a e j
L

e j

From which we were able to conclude the approximate (i.e., less accurate!) result:
in

b = + L e j 2 a

The exact result was simpleand exact! Why did you make us determine this approximate result?
A: In a word: frequency response*. Although the exact analysis is about as simple to determine as the approximation provided by the theory of small reflections, the mathematical form of the result is much simpler to analyze and/or evaluate (e.g., no fractional terms!). Q: What exactly would we be analyzing and/or evaluating? A: The frequency response of the matching network, for one thing.
* OK, two words.

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Frequency Response of a Quarter

3/9

Remember, all matching networks must be lossless, and so must be made of reactive elements (e.g., lossless transmission lines). The impedance of every reactive element is a function of frequency, and so too then is in . Say we wish to determine this function in ( ) . Q: Isnt in ( ) = 0 for a quarter wave matching network? A: Oh my gosh no! A properly designed matching network will typically result in a perfect match (i.e., in = 0 ) at one frequency (i.e., the design frequency). However, if the signal frequency is different from this design frequency, then no match will occur (i.e., in 0 ).

Recall we discussed this behavior before:

Reflection coefficient magnitude versus frequency for a singlesection quarter-wave matching transformer with various load mismatches.

Figure 5.12 (p. 243)

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Frequency Response of a Quarter

4/9

Q: But why is the result:

2 L e j 2 in = + 1 L

or its approximate form:

in = + L e j 2

dependent on frequency? I dont see frequency variable anywhere in these results!

A: Look closer!

Remember that the value of spatial frequency (in radians/meter) is dependent on the frequency of our eigen function (aka the signal):
1 = v p

where you will recall that v p is the propagation velocity of a wave moving along a transmission line. This velocity is a constant (i.e., v p = 1 LC ), and so the spatial frequency is directly proportional to the temporal frequency . Thus, we can rewrite:

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Frequency Response of a Quarter

5/9

= T vp

Where T = v p is the time required for the wave to

propagate a distance

down a transmission line.

As a result, we can write the input reflection coefficient as a function of spatial frequency :
in ( ) = + L e j 2

Or equivalently as a function of temporal frequency :

in ( ) = + L e j 2T

Frequently, the reflection coefficient is simply written in terms of the electrical length of the transmission line, which is simply the difference in relative phase between the wave at the beginning and end of the length of the transmission line. = = T

So that:

in ( ) = + L e j 2

Note we can simply insert the value = into the expression above to get in ( ) , or insert = T into the expression to get in ( ) .

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Frequency Response of a Quarter

6/9

Now, we know that = L for a properly designed quarterwave matching network, so the reflection coefficient function can be written as: in ( ) = L 1 + e j 2

Note that: And that: And so:

1 = e j 0 = e j ( ) = e j e + j

e j 2 = e j ( + ) = e j e j

in ( ) = L 1 + e j 2 = L
j

( (e

e + j + e j e j )

= L e j ( e + j + e j ) = L e j (2 cos )

Where we have used Eulers equation to determine that:

e + j + e j = 2 cos
Now, lets determine the magnitude of our result:
in ( ) = L e j 2 cos = 2 L cos Note that in ( ) is zero-valued only when cos = 0 . This of course occurs when = 2 : in ( ) = = 2 L cos 2 = 0
2

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Frequency Response of a Quarter

7/9

In other words, a perfect match occurs when = 2 !!

Q: What the heck does this mean? A: Remember, = . Thus if = 2 :

= =

!!

As we (should have) suspected, the match occurs at the frequency whose wavelength is equal to four times the matching ( Z 1 ) transmission line length, i.e. = 4 . In other words, a perfect match occurs at the frequency where = 4 . Note the physical length of the transmission line does not change with frequency, but the signal wavelength does:

vp f

Q: So, at precisely what frequency does a quarter-wave

transformer with length

provide a perfect match?

A: Recall also that = T , where T = v p . Thus, for = 2 :

= T

1 vp = 2T 2

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Frequency Response of a Quarter

8/9

This frequency is called the design frequency of the matching networkits the frequency where a perfect match occurs. We denote this as frequency 0 , which has wavelength 0 , i.e.:

0 =

vp = 2 2 T

f0 =

vp 0 1 = = T 4 2 4

0 =

vp = 4v pT = 4 f0

Given this, yet another way of expressing = is:

= =
Thus, we conclude:

vp

vp 20

f = = 20 2f0

in (f ) = 2 L cos

2f0

From this result we can determine (approximately) the bandwidth of the quarter-wave transformer! First, we must define what we mean by bandwidth. Say the maximum acceptable level of the reflection coefficient is value . This is an arbitrary value, set by you the microwave m engineer (typical values of range from 0.05 to 0.2). m We will denote the frequencies where this maximum value m occurs fm . In other words:

in (f = fm ) = m = 2 L cos

2f0

fm

)
Dept. of EECS

Jim Stiles

The Univ. of Kansas

4/1/2009

The Frequency Response of a Quarter

9/9

There are two solutions to this equation, the first is:

fm 1
And the second:

m cos = 2 L
1

2f0

fm 2 =

2f0

cos 1

m 2 L

Important note! Make sure cos-1x is expressed in radians! You will find that fm 1 < f0 < fm 2 so, the values fm 1 and fm 2 define the lower and upper limits on matching network bandwidth.

m
fm 1 fm 2

All this analysis was brought to you by the simple mathematical form of in (f ) that resulted from the theory of small reflections!

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Multisection Transformer

1/6

The Multi-section Transformer

Consider a sequence of N transmission line sections; each section has equal length , but dissimilar characteristic impedances:
0
1

N 1

Z0

in

Z1

Z2

ZN

RL

Where the marginal reflection coefficients are:

0

Z1 Z 0 Z1 + Z 0

Z n +1 Z n Z n +1 + Z n

RL Z N RL + Z N

If the load resistance RL is less than Z 0 , then we should design the transformer such that:

Z 0 > Z1 > Z2 > Z 3

> Z N > RL

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Multisection Transformer

2/6

transformer such that:

Conversely, if RL is greater than Z 0 , then we will design the

Z 0 < Z1 < Z2 < Z3

< Z N < RL

In other words, we gradually transition from Z0 to RL ! Note that since RL is real, and since we assume lossless transmission lines, all n will be real (this is important!). Likewise, since we gradually transition from one section to another, each value:

will be small.

Z n +1 Z n

As a result, each marginal reflection coefficient n will be real and have a small magnitude. This is also important, as it means that we can apply the theory of small reflections to analyze this multi-section transformer! The theory of small reflections allows us to approximate the input reflection coefficient of the transformer as:

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Multisection Transformer

3/6

N 1

Z0

in

Z1

Z2

ZN

RL

a0
The approximate SFG when applying the theory of small reflections!

a1 e j
0
1

a2 e j

......
2

aN 1
e j
N 1

aN

L = N

b0

e j

e j

b1

b2

......

e j

bN 1

bN

b0 = in ( ) a0
0 + 1 e j 2 + 2 e j 4 + = n e j 2n
n =0 N

+ N e j 2N

We can alternatively express the input reflection coefficient as a function of frequency ( = ): T

in ( ) = 0 + 1e j 2T + 2e j 4T + = n e j (2nT )
n =0 N

+ N e j 2N T

where:

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Multisection Transformer

4/6

vp

= propagation time through 1 section

We see that the function in ( ) is expressed as a weighted set of N basis functions! I.E.,
in ( ) = cn ( )
N

where:

n =0

cn = n

and

( ) = e

j (2nT )

we can synthesize any function in ( ) of frequency , provided that: 1. in ( ) is periodic in = 1 2 T 2. we have sufficient number of sections N. Q: What function should we synthesize? A: Ideally, we would want to make in ( ) = 0 (i.e., the reflection coefficient is zero for all frequencies). Bad news: this ideal function in ( ) = 0 would require an infinite number of sections (i.e., N = )!

We find, therefore, that by selecting the proper values of basis weights cn (i.e., the proper values of reflection coefficients n ),

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Multisection Transformer

5/6

Instead, we seek to find an optimal function for in ( ) , given a finite number of N elements. Once we determine these optimal functions, we can find the values of coefficients n (or equivalently, Zn ) that will result in a matching transformer that exhibits this optimal frequency response. To simplify this process, we can make the transformer symmetrical, such that:
0 = N , 1 = N 1 , 2 = N 2 ,

Note that this does NOT mean that:

Z 0 = Z N , Z1 = Z N 1 , Z2 = Z N 2 ,
We find then that:
( ) = e jN T 0 e jN T + e jN T + 1 e j (N 2)T + e j (N 2)T + 2 e j (N 4 )T + e j (N 4 )T +

and since:

e j x + e j x = 2 cos ( x )

we can write for N even:

Jim Stiles

The Univ. of Kansas

Dept. of EECS

4/1/2009

The Multisection Transformer

6/6

( ) = 2 e j N T 0 cosN + 1 cos (N 2 ) T T 1 + + n cos (N 2n ) + + N 2 T 2

whereas for N odd:

( ) = 2 e j N T 0 cosN + 1 cos (N 2 ) T T + + n cos (N 2n ) + T + (N 1) 2 cos T

The remaining question then is this: given an optimal and realizable function in ( ) , how do we determine the necessary

number of sections N, and how do we determine the values of all reflection coefficients n ??

Jim Stiles

The Univ. of Kansas

Dept. of EECS