9/13/2006 Amplifiers 1/2

2. Amplifiers

We will find that the signal power collected by a receiver

antenna is often ridiculously small (e.g., less than one
trillionth of a Watt!)

To accurately recover the information impressed on this

signal, we must increase the signal power a whole bunch—
without modifying or distorting the signal in any way.

But first, a few comments about the decibel!

HO: dB, dBm, dBw

Q: So what is an amplifier exactly? What is it suppose to

accomplish?

A: HO: Amplifiers

Q: By how much will an amplifier increase signal power?

A: HO: Amplifier Gain

Q: Can we increase this signal power an unlimited amount?

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Amplifiers 2/2

A:

HO: Amplifier Output Power

Q: So, just how precisely does an amplifier reproduce a signal

at its output?

A: HO: Intermodulation Distortion

Q: Is intermodulation distortion really that big of a problem?

A:

HO: Two-Tone Intermodulation Distortion

Every good radio engineer knows and understands that

parameters of the amplifier spec sheet!

HO: The Amplifier Spec Sheet

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 dB 1/9

dB, dBm, dBw

Decibel (dB), is a specific function that operates on a unitless
parameter:

dB  10 log10 (x)

where x is unitless!

Q: A unitless parameter! What good is that ! ?

A: Many values are unitless, such as ratios and coefficients.

For example, amplifier gain is a unitless value!

E.G., amplifier gain is the ratio of the

output power to the input power:

Pout
=G
Pin

∴ Gain in dB = 10 log10G  G (dB )

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 dB 2/9

Q: Wait a minute! I’ve seen statements such as:

…. the output power is 5 dBw ….

or
…. the input power is 17 dBm ….

Of course, Power is not a unitless parameter!?!

A: True! But look at how power is expressed; not in dB, but in

dBm or dBw.

Q: What the heck does dBm or dBw refer to ??

A: It’s sort of a trick !

Say we have some power P. Now say we divide this value P by

one 1 Watt. The result is a unitless value that expresses the
value of P in relation to 1.0 Watt of power.

For example, if P = 2500 mW , then P 1W = 2.5 . This simply

means that power P is 2.5 times larger than one Watt!

Since the value P 1W is unitless, we can express this value in

decibels!

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 dB 3/9

Specifically, we define this operation as:

⎛ P ⎞
P (dBw )  10 log10 ⎜ ⎟
⎝1W ⎠

For example, P = 100 Watts can alternatively be expressed as

P (dBw ) = +20 dBw . Likewise, P = 1 mW can be expressed as
P (dBw ) = −30 dBw .

Q: OK, so what does dBm mean?

A: This notation simply means that we have normalized some

power P to one Milliwatt (i.e., P 1mW )—as opposed to one
Watt. Therefore:

⎛ P ⎞
P (dBm )  10 log10 ⎜ ⎟
⎝ 1 mW ⎠

For example, P = 100 Watts can alternatively be expressed as

P (dBm ) = +50 dBm . Likewise, P = 1 mW can be expressed as
P (dBm ) = 0 dBm .

Make sure you are very careful when doing math with
decibels!

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 dB 4/9

Standard dB Values

Note that 10 log10 (10 ) = 10dB

Therefore an amplifier with a gain G = 10 is likewise said to

have a gain of 10 dB.

Now consider an amplifier with a gain of 20 dB……

Q: Yes, yes, I know. A 20 dB

amplifier has gain G=20, a 30 dB
amp has G=30, and so forth.

Please speed this lecture up and

quit wasting my valuable time
making such obvious statements!

A: NO! Do not make this mistake!

Recall from your knowledge of logarithms that:

10log10 ⎡⎣ 10n ⎤⎦ = n 10log10 [ 10 ] = 10n

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 dB 5/9

Therefore, if we express gain as G = 10n , we conclude:

G = 10n ↔ G (dB ) = 10n

In other words, G =100 = 102 (n =2) is expressed as 20 dB,

while 30 dB (n =3) indicates G = 1000 = 103.

Likewise 100 mW is denoted as 20 dBm, and 1000 Watts is

denoted as 30 dBW.

Note also that 0.001 mW = 10-3 mW is denoted as –30 dBm.

Another important relationship to keep in mind when using

decibels is 10log10 [ 2 ] ≈ 3.0 . This means that:

10log10 ⎡⎣ 2n ⎤⎦ = n 10log10 [ 2 ]  3n

Therefore, if we express gain as G = 2n , we conclude:

G = 2n ↔ G (dB )  3n

As a result, a 15 dB (n =5) gain amplifier has G = 25 = 32.

Similarly, 1/8 = 2-3 mW (n =-3) is denoted as –9 dBm.

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 dB 6/9

Multiplicative Products and Decibels

Other logarithmic relationship that we will find useful are:

10log10 [ x y ] = 10log10 [ x ] + 10log10 [ y ]

and its close cousin:

⎡x ⎤
10log10 ⎢ ⎥ = 10log10 [ x ] − 10log10 [ y ]
⎢⎣ y ⎥⎦

Thus, the relationship Pout = G Pin is written in decibels as:

Pout = G Pin
Pout G Pin
=
1mW 1mW
⎡ P ⎤ ⎡ G Pin ⎤
10log10 ⎢ out ⎥ = 10log10 ⎢ ⎥
⎣⎢ 1mW ⎦⎥ ⎢⎣ 1mW ⎥⎦
⎡ P ⎤ ⎡ P ⎤
10log10 ⎢ out ⎥ = 10log10 [ G ] + 10log10 ⎢ in ⎥
⎢⎣ 1mW ⎥⎦ ⎢⎣ 1mW ⎥⎦

Pout (dBm ) = G (dB ) + Pout (dBm )

It is evident that “deebees” are not a unit! The units of the

result can be found by multiplying the units of each term in a
summation of decibel values.

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 dB 7/9

For example, say some power P1 = 6 dBm is combined with

power P2 = 10 dBm . What is the resulting total power
PT = P1 + P2 ?

Q:This result really is obvious—

of course the total power is:

PT (dBm ) = P1 (dBm ) + P2 (dBm )

= 6 dBm + 10 dBm
= 16 dBm

A: NO! Never do this either!

Logarithms are very helpful in expressing products or ratios

of parameters, but they are not much help when our math
involves sums and differences!

10log10 [ x + y ] = ????

So, if you wish to add P1 =6 dBm of power to P2 =10 dBm of

power, you must first explicitly express power in Watts:

P1 =10 dBm = 10 mW and P2 =6 dBm = 4 mW

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 dB 8/9

Thus, the total power PT is:

PT = P1 + P2
= 4.0 mW + 10.0 mW
= 14.0 mW

Now, we can express this total power in dBm, where we find:

⎛ 14.0 mW ⎞
PT (dBm ) = 10 log10 ⎜ ⎟ = 11.46 dBm
⎝ 1 . 0 mW ⎠

The result is not 16.0 dBm !.

We can mathematically add 6 dBm and 10 dBm, but we must

understand what result means (nothing useful!).

⎡ 4mW ⎤ ⎡ 10mW ⎤
6 dBm + 10 dBm = 10log10 ⎢ ⎥ + 10log10 ⎢ ⎥
⎢⎣ 1mW ⎥⎦ ⎢⎣ 1mW ⎥⎦
⎡ 40 mW 2 ⎤
= 10log10 ⎢ ⎥
⎢⎣ 1mW 2 ⎥⎦
= 16 dB relative to 1 mW2

Thus, mathematically speaking, 6 dBm + 10 dBm implies a

multiplication of power, resulting in a value with units of
Watts squared !

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 dB 9/9

A few more tidbits about decibels:

1. 1 .0 ↔ 0 dB

2. 0.0 ↔ − ∞ dB

3. 5n ↔  7n dB (can you show why?)

I wish I had a
nickel for every
time my software
has crashed-oh
wait, I do!

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Amplifiers 1/7

Amplifiers
An ideal amplifier takes an input signal and reproduces it
exactly at its output, only with a larger magnitude!

vin( t ) Avo vout( t ) = Avo vin( t )

where Avo is the open-circuit voltage gain of the amplifier.

Now, let’s express this result using our

knowledge of linear circuit theory !

Recall, the output vout( t ) of a linear device can be determined

by convolving its input vin( t ) with the device impulse response
g (t ) :
t
vout (t ) = ∫ g (t −t ′)vin (t ′)dt ′
−∞

Jim Stiles The Univ. of Kansas Dept. of EECS

 9/13/2006 Amplifiers 2/7

The impulse response for the ideal amplifier would therefore

be:
g (t ) = Avo δ (t )
so that:
t
vout (t ) = ∫ g (t −t ′)vin (t ′)dt ′
−∞
t
= ∫ Avo δ (t −t ′)vin (t ′)dt ′
−∞

= Avo vin (t )

We can alternatively represent the ideal amplifier response in

the frequency domain, by taking the Fourier Transform of the
impulse response:
∞
T (ω ) = ∫ g (t )e
− j ωt
dt
−∞
∞

∫ Avo δ (t )e dt
− j ωt
=
−∞

= Avo + j 0

This result, although simple, has an interesting interpretation.

It means that the amplifier exhibits gain of Avo for sinusoidal
signals of any and all frequencies!

T (ω )

Avo

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Amplifiers 3/7

Moreover, the ideal amplifier does not alter the relative phase
of the sinusoidal signal (i.e., no phase shift).

In other words, if:

vin (t ) = cos (ωt )

then at the output of the ideal amplifier we shall see:

vout (t ) = T (ω ) cos (ωt + ∠T (ω ))

= Avo cos (ωt )

BUT, there is one big problem with an ideal amplifier:

They are impossible to build !!

Q: Why is that ??

A: Two reasons:

a) An ideal amplifier has infinite bandwidth.

b) An ideal amplifier has zero delay.

Not gonna happen !

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Amplifiers 4/7

Let’s look at this second problem first. The ideal amplifier

impulse response g (t ) = Avo δ (t ) means that the signal at the
output occurs instantaneously with the signal at the input.

This of course cannot happen, as it takes some small, but non-

zero amount of time for the signal to propagate through the
amplifier. A more realizable amplifier impulse response is:

g (t ) = Avo δ (t − τ )

resulting in an amplifier output of:

t
vout (t ) = ∫ g (t −t ′) vin (t ′)dt ′
−∞
t
= ∫ Avo δ (t − τ −t ′) vin (t ′)dt ′
−∞

= Avo vin (t − τ )

In other words, the output is both an amplified and delayed

version of the input.

* Ideally, this delay does not distort the signal, as the output
will have the same form as the input.

* Moreover, the delay for electronic devices such as

amplifiers is very small in comparison to human time scales (i.e.,
τ  1 second ).

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Amplifiers 5/7

* Therefore, propagation delay τ is generally not considered

a problem for most amplifier applications.

Let’s examine what this delay means in the frequency domain.

Evaluating the Fourier Transform of this modified impulse

response gives:

∞
T (ω ) = ∫ g (t ) e − j ωt dt
−∞
∞

∫ Avo δ (t − τ ) e dt
− j ωt
=
−∞

= Avo cos (ωτ ) + j Avo sin (ωτ )

= Avo e j ωτ

We see that, as with the ideal amplifier, the magnitude

T (ω ) = Avo . However, the relative phase is now a linear function
of frequency:
∠T (ω ) = ωτ

As a result, if vin (t ) = cos (ωt ) , the output signal will be:

vout (t ) = T (ω ) cos (ωt − ∠T (ω ))

= Av cos (ωt − ωτ )

In other words, the output signal of a real amplifier is phase

shifted with respect to the input.

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Amplifiers 6/7

In general, the amplifier phase shift ∠T (ω ) will not be a

perfectly linear function (i.e., ∠T (ω ) ≠ ωτ ), but instead will be a
more general function of frequency ω.

However, if the phase function ∠T (ω ) becomes too “non-

linear”, we find that signal dispersion can result—the output
signal can be distorted!

Now, let’s examine the first problem with the ideal amplifier.
This problem is best discussed in the frequency domain.

We discovered that the ideal amplifier has a frequency

response of T (ω ) = Avo . Note this means that the amplifier gain
is Avo for all frequencies 0 < ω < ∞ (D.C. to daylight !).

The bandwidth of the ideal amplifier is therefore infinite !

* Since every electronic device will exhibit some amount of

inductance, capacitance, and resistance, every device will have a
finite bandwidth.

* In other words, there will be frequencies ω where the

device does not work !

* From the standpoint of an amplifier, “not working” means

T (ω )  Avo (i.e., low gain).

* Amplifiers will therefore have finite bandwidths.

Jim Stiles The Univ. of Kansas Dept. of EECS

 9/13/2006 Amplifiers 7/7

There is a range of frequencies ω between ω L and ωH where

the gain will (approximately) be Avo. For frequencies outside
this range, the gain will typically be small (i.e. T (ω )  Avo ):

⎧ ≈ Avo ω L < ω < ωH

T (w ) = ⎨
⎩  Avo ω < ω L , ω > ωH

The width of this frequency range is called the amplifier

bandwidth:
Bandwidth  ωH − ω L (radians/sec)
 fL − fH (cycles/sec)
T (ω )

Avo

ω
ωL ωH

One result of having a finite bandwidth is that the amplifier

impulse response is not an impulse function !

∞
g (t ) = ∫T (ω ) e + j ωt dt ≠ Avo δ (t − τ )
−∞

The ideal amplifier is not possible!

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Amplifier Gain 1/4

Amplifier Gain
Note that an amplifier is a two-port device.

Port 1 Port 2
Av

As a result, we can describe an amplifier with a 2 x 2 scattering

matrix:
⎡S11 S12 ⎤
S( ω ) = ⎢ ⎥
⎣S21 S22 ⎦

Q: What is the scattering matrix of an ideal amplifier??

A: Let’s start with S11 and S22.

To insure maximum power transfer, the input and output ports

would ideally be matched:

S11 = S22 = 0

Now, let’s look at scattering parameter S21 . We know that:

2
P2− = S21 P1 +

Jim Stiles The University of Kansas Dept. of EECS

9/13/2006 Amplifier Gain 2/4

or, stated another way:

2
Pout = S21 Pin

Therefore, we can define the amplifier power gain as:

Pout 2
G  = S21
Pin

As the purpose of an amplifier is to boost the signal power, we

can conclude that ideally:
S21  1

Clearly, an amplifier must be an active device!

As discussed earlier, the gain of an amplifier will change with

signal frequency:
2
G (ω ) = S21 (ω )

When radio engineers speak of amplifier gain, they almost

always are speaking of this power gain G. However, they do not
generally state it as a specific function of frequency!

Rather, amplifier gain is typically specified as a numeric value

such as G =20 or G =13 dB. This value is a statement of the
approximate amplifier gain within the amplifier bandwidth.

Jim Stiles The University of Kansas Dept. of EECS

9/13/2006 Amplifier Gain 3/4

G (ω )
G

ω
ωL ωH

Thus, amplifier gain and bandwidth are the two most

fundamental performance specifications of any microwave
amplifier—together they (approximately) describe the amplifier
transfer function!

Additionally, radio engineers almost always speak of amplifier

gain in decibels (dB):

G (dB ) = 10 log10 G

Finally, let’s consider S12. This scattering parameter relates the

wave into port 2 (the output) to the wave out of port 1 (the
input).
V1 − V2+

Q: Are amplifiers reciprocal devices? In other words, is

S12 = S21 ??

Jim Stiles The University of Kansas Dept. of EECS

9/13/2006 Amplifier Gain 4/4

A: No! An amplifier is strictly a directional device; there is a

specific input, and a specific output—it does not work in
reverse!

Ideally, S12 = 0. Any other value can just cause problems!

Typically though, S12 is small, but not zero. Generally speaking,

radio engineers express S12 as a value called reverse isolation:

2
reverse isolation  −10 log10 S12

Note when S12 =0, reverse isolation will be infinite. Thus, the
larger the reverse isolation, the better!

Summarizing, we find that the scattering matrix of the ideal

amplifier is:

⎡0 0⎤
Sideal =⎢ where S21  1
⎣S21 0 ⎥⎦

Sort of like an isolator with gain!

The non-ideal reality is that the zero valued terms will be small,
but not precisely zero. Moreover, each scattering parameter
will change with signal frequency—although they remain
approximately constant within the amplifier bandwidth.

Jim Stiles The University of Kansas Dept. of EECS

09/13/06 Amplifier Output Power 1/5

Amplifier Output Power

Say we have an amplifier with gain G = 30 dB (i.e., G = 1000).

If the input power to this amplifier is 0 dBw (i.e., Pin = 1W), then
the output power is:
Pin G = Pout
(1 W) 1000 = 1000 W

Or, in dB:
0 dBw + 30 dB = 30 dBw

Pin = 1W Pout = 1 kW

WOW! We created 999 Watts !

The energy crisis is solved !

Of course, the amplifier cannot create energy.

Q: Then, where does the power come from ???

A: The D.C. power supply ! (Every amplifier has one).

Jim Stiles The Univ. of Kansas Dept. of EECS

09/13/06 Amplifier Output Power 2/5

+V

Pin Pout
G

D. C. Power =PDC =V I > Pout

The output power Pout cannot exceed the power delivered by the
D.C. supply.

Q: What happens to the D.C. power not converted to signal

power Pout ??

A:

So, if we were to plot Pout vs. Pin for a microwave amplifier, we

would get something like this:

Jim Stiles The Univ. of Kansas Dept. of EECS

09/13/06 Amplifier Output Power 3/5

Pout Pout = G Pin

PDC

Pin

We notice that the output power compresses, or saturates.

Note there is one point on this curve where the amplifier

output power Pout is 1 dB less than its ideal value of G Pin. In
other words, there is one (and only one!) value of Pin and Pout that
will satisfy the equation:

Pout (dB ) = ⎡⎣Pin (dB ) + G (dB ) ⎤⎦ − 1 dB

At this point, the amplifier is said to be compressed 1 dB.

Therefore, a 10 dB amplifier would appear to be a 9 dB
amplifier!

The output power when the amplifier has compressed 1dB is

called the 1 dB compression point  P1dB of the amplifier.

Jim Stiles The Univ. of Kansas Dept. of EECS

09/13/06 Amplifier Output Power 4/5

The 1 dB compression point is generally considered to be the

maximum power output of the amplifier.

The input power at the 1 dB compression point is said to be the

maximum input power ( Pinmax ) of the amplifier. We of course
can put more than Pinmax into the amplifier—but we won’t get
much more power out!

Pout Pout = G Pin

PDC

1 dB
P1dB

Pinmax Pin

Note the equation Pout (dB ) = ⎡⎣Pin (dB ) + G (dB ) ⎤⎦ − 1 dB alone is

not sufficient to determine the 1 dB compression point, as we
have two uknowns (Pin and Pout). We need another equation!

This second “equation” is the actual curve or table of data

relating Pin to Pout for a specific amplifier.

Jim Stiles The Univ. of Kansas Dept. of EECS

09/13/06 Amplifier Output Power 5/5

Amplifier Efficiency

We can define amplifier efficiency e as the ratio of the

maximum output power ( P1dB) to the D.C. power:

P1dB
e = (don’t use decibels here!)
PDC

For example, if e=0.4, then up to 40% of the D.C. power can be

converted to output power, while the remaining 60% is
converted to heat.

We require high power amps to be very efficient!

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Intermodulation Distortion 1/11

Intermodulation Distortion
The 1 dB compression curve shows that amplifiers are only
approximately linear.

Actually, this should be obvious, as amplifiers are constructed

with transistors—non-linear devices!

So, instead of the ideal case:

vout = Av vin

Actual amplifier behavior requires more terms to describe!

vout = Av vin + B vin2 + C vin3 + "

This representation is simply a Taylor Series representation of

the non-linear function:
vout = f (vin )

Q: Non-linear! But I thought an

amplifier was a linear device?
After all, we characterized it
with a scattering matrix!

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Intermodulation Distortion 2/11

A: Generally speaking, the constants B, C, D, etc. are very small

compared to the voltage gain Av . Therefore, if vin is likewise
small, we can truncate the Taylor Series and approximate
amplifier behavior as the linear function:

vout ≈ Av vin

BUT, as vin gets large, the values vin2 and vin3 will get really
large! In that case, the terms B vin2 and C vin3 will become
significant.

As a result, the output will not simply be a larger version of the

input. The output will instead be distorted—a phenomenon
known as Intermodulation Distortion.

Q: Good heavens! This sounds

terrible. What exactly is
Intermodulation Distortion,
and what will it do to our
signal output?!?

A: Say the input to the amplifier is sinusoidal, with magnitude a:

vin = a cos ωt

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Intermodulation Distortion 3/11

Using our knowledge of trigonometry, we can determine the

result of the second term of the output Taylor series:

B vin2 = B a 2 cos2 ωt
B a2 B a2
= + cos 2ωt
2 2

We have created a harmonic of the input signal!

In other words, the input signal is at a frequency ω, while the

output includes a signal at twice that frequency (2ω).

We call this signal a second order product, as it is a result of

squaring the input signal.

Note we also have a cubed term in the output signal equation:

vout = Av vin + B vin2 + C vin3 + "

Using a trig identity, we find that:

C vin3 = C a 3 cos3 ωt
C a3 C a3
= cos ωt + cos 3ωt
2 4

Now we have produced a second harmonic (i.e., 3ω)!

As you might expect, we call this harmonic signal a third-order

product (since it’s produced from vin3 ).

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Intermodulation Distortion 4/11

Q: I confess that I am still a bit

befuddled. You said that values B and
C are typically much smaller that that
of voltage gain Av . Therefore it
would seem that these harmonic
signals would be tiny compared to the
fundamental output signal Av a cos ωt .
Thus, I don’t why there’s a problem!

To understand why intermodulation distortion can be a problem

in amplifiers, we need to consider the power of the output
signals.

We know that the power of a sinusoidal signal is proportional to

its magnitude squared. Thus, we find that the power of each
output signal is related to the input signal power as:

1rst-order output power  P1out = Av2 Pin = G Pin

out B2
2nd-order output power  P2 = Pin2 = G2 Pin2
4
out C2
3rd-order output power  P3 = Pin3 = G3 Pin3
16

where we have obviously defined G2  B 2 4 and G3  C 2 16 .

Note that unlike G, the values G2 and G3 are not coefficients

(i.e., not unitless!). The value G2 obviously has units of inverse
power (e.g., mW-1 or W-1 ), while G3 has units of inverse power
squared (e.g., mW-2 or W-2 ).

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Intermodulation Distortion 5/11

We know that typically, G2 and G3 are much smaller than G.

Thus, we are tempted to say that P1out is much larger than
P2out or P3out .

But, we might be wrong ! Q: Might be wrong! Now I’m

more confused than ever. Why
can’t we say definitively that
the second and third order
products are insignificant??

Look closely at the expressions for the output power of the

first, second, and third order products:

P1out = G Pin
P2out = G2 Pin2
P3out = G3 Pin3

This first order output power is of course directly proportional

to the input power. However, the second order output power is
proportional to the input power squared, while the third order
output is proportional to the input power cubed!

Thus we find that if the input power is small, the second and
third order products are insignificant. But, as the input power
increases, the second and third order products get big in a
hurry!

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Intermodulation Distortion 6/11

For example, if we double the input power, the first order

signal will of course likewise double. However, the second order
power will quadruple, while the third order power will increase 8
times.

For large input powers, the second and third order output
products can in fact be almost as large as the first order
signal!

Perhaps this can be most easily seen by expressing the above

equations in decibels, e.g.,:

P1out (dBm ) = G (dB ) + Pin (dBm )

P2out (dBm ) = G2 (dBm −1 ) + 2 ⎡⎣ Pin (dBm ) ⎤⎦
P3out (dBm ) = G3 (dBm −2 ) + 3 ⎡⎣ Pin (dBm ) ⎤⎦

where we have used the fact that log x n = n log x . Likewise, we

have defined:
⎡ ⎤
⎢ G2 ⎥
G2 (dBm −1 ) = 10log10 ⎢ ⎥
⎢( 1
⎣ 1.0mW ) ⎦
⎥
= 10log10 [ G2 ( 1.0mW ) ]

and:
⎡ ⎤
⎢ G3 ⎥
G3 (dBm −2 ) = 10log10 ⎢ ⎥
⎢ 1
(
⎢ 1.0mW 2 ⎥
⎣
⎥
⎦ )
= 10log10 ⎡⎣ G3 ( 1.0mW 2 ) ⎤⎦

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Intermodulation Distortion 7/11

Hint: Just express everything in milliwatts!

Note the value 2 ⎡⎣ Pin (dBm ) ⎤⎦ does not mean the value 2Pin
expressed in decibels. The value 2 ⎡⎣ Pin (dBm ) ⎤⎦ is fact the value
of Pin expressed in decibels—times two!

For example, if Pin (dBm ) = −30 dBm , then

2 ⎡⎣ Pin (dBm ) ⎤⎦ = −60 dBm . Likewise, if Pin (dBm ) = 20 dBm , then
2 ⎡⎣ Pin (dBm ) ⎤⎦ = 40 dBm .

What this means is that for every 1dB increase in input power
Pin the fundamental (first-order) signal will increase 1dB;
the second-order power will increase 2dB; and the third-order
power will increase 3dB.

This is evident when we look at the three power equations (in

decibels), as each is an equation of a line (i.e., y = m x + b).

For example, the equation:

P3out (dBm ) = 3 ⎡⎣ Pin (dBm ) ⎤⎦ + G3 (dBm −2 )

y = mx +b

describes a line with slope m =3 and “y intercept” b =

G3 (dBm −2 ) (where x = Pin (dBm ) and y = P out (dBm ) ).

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Intermodulation Distortion 8/11

Plotting each of the three equations for a typical amplifier, we

would get something that looks like this:

m3 = 3
Pout(dBm)

P2int

P3int

G (dB )

Pin (dBm)
m1 = 1

G2 (dBm −1 )
G3 (dBm −2 )
m2 = 2
P1out (dBm)
P2out (dBm)
P3out (dBm)

Note that for Pin (dBm ) < 0 dBm (the left side of the plot), the
second and third-order products are small compared to the
fundamental (first-order) signal.

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Intermodulation Distortion 9/11

However, when the input power increases beyond 0 dBm (the

right side of the plot), the second and third order products
rapidly catch up! In fact, they will (theoretically) become equal
to the first order product at some large input power.

The point at which each higher order product equals the first-
order signal is defined as the intercept point. Thus, we define
the second order intercept point as the output power when:

P2out = P1out  P2int Second - order intercept power

Likewise, the third order intercept point is defined as the

third-order output power when:

P3out = P1out  P3int Third - order intercept power

Using a little algebra you can show that:

G2 G3
P2int = and P3int =
G2 G3

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Intermodulation Distortion 10/11

Or, expressed in decibels:

P2int (dBm ) = 2 G (dB ) − G2 (dBm −1 )

3 G (dB ) − G3 (dBm −2 )
P3int (dBm ) =
2

* Radio engineers specify the intermodulation distortion

performance of a specific amplifier in terms of the intercept
points, rather than values G2 and G3.

* Generally, only the third-order intercept point is provided by

amplifier manufactures (we’ll see why later).

* Typical values of P3int for a small-signal amplifier range

from +20 dBm to +50 dBm

* Note that as G2 and G3 decrease, the intercept points

increase.

Therefore, the higher the intercept point of an amplifier, the

better the amplifier !

One other important point: the intercept points for most

amplifiers are much larger than the compression point! I.E.,:

P int > P1dB

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Intermodulation Distortion 11/11

In other words the intercept points are “theoretical”, in that we

can never, in fact, increase the input power to the point that
the higher order signals are equal to the fundamental signal
power.

All signals, including the higher order signals, have a maximum

limit that is determined by the amplifier power supply.

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Two-Tone Intermodulation 1/7

Two-Tone Intermodulation

Q: It doesn’t seem to me that this

dad-gum intermodulation distortion
is really that much of a problem.

I mean, the first and second

harmonics will likely be well outside
the amplifier bandwidth, right?

A: True, the harmonics produced by intermodulation distortion

typically are not a problem in radio system design. There is a
problem, however, that is much worse than harmonic distortion!

This problem is called two-tone intermodulation distortion.

Say the input to an amplifier consists of two signals at

dissimilar frequencies:

vin = a cos ω1t + a cos ω2t

Here we will assume that both frequencies ω1 and ω2 are within

the bandwidth of the amplifier, but are not equal to each other
( ω1 = ω2 ) .

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Two-Tone Intermodulation 2/7

This of course is a much more realistic case, as typically there

will be multiple signals at the input to an amplifier!

For example, the two signals considered here could represent

two FM radio stations, operating at frequencies within the FM
band (i.e., 88.1 MHz ≤ f1 ≤ 108.1 MHz and
88.1 MHz ≤ f2 ≤ 108.1 MHz ).

Q: My point exactly!
Intermodulation distortion will
produce those dog-gone second-
order products:

a2 a2
cos 2ω1t and cos 2ω2t
2 2

and gul-durn third order products:

a3 a3
cos3ω1t and cos3ω2t
4 4

but these harmonic signals will lie

well outside the FM band!

A: True! Again, the harmonic signals are not the problem. The
problem occurs when the two input signals combine together to
form additional second and third order products.

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Two-Tone Intermodulation 3/7

Recall an amplifier output is accurately described as:

vout = Av vin + B vin2 + C vin3 + "

Consider first the second-order term if two signals are at the

input to the amplifier:

v2out = B vin2
2
= B ( a cos ω1t + a cos ω2t )
= B ( a 2 cos2 ω1t + 2a 2 cos ω1t cos ω2t + a 2 cos2 ω2t )

Note the first and third terms of the above expression are
precisely the same as the terms we examined on the previous
handout. They result in harmonic signals at frequencies 2ω1 and
2ω2 , respectively.

The middle term, however, is something new. Note it involves

the product of cos ω1t and cos ω2t . Again using our knowledge
of trigonometry, we find:

2a 2 cos ω1t cos ω2t = a 2 cos (ω2 − ω1 )t + a 2 cos (ω2 + ω1 )t

Note that since cos ( −x ) = cos x , we can equivalently write this

as:

2a 2 cos ω1t cos ω2t = a 2 cos (ω1 − ω2 )t + a 2 cos (ω1 + ω2 )t

Either way, the result is obvious—we produce two new signals!

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Two-Tone Intermodulation 4/7

These new second-order signals oscillate at frequencies

( ω1 + ω2 ) and ω1 − ω2 .

Thus, if we looked at the frequency spectrum (i.e., signal power

as a function of frequency) of an amplifier output when two
sinusoids are at the input, we would see something like this:

First-order products
P/Hz
Second-order products

2ω1 2ω2

ω1 ω2 ω1 + ω2 ω
ω1 − ω2

Note that the new terms have a frequency that is either much
higher than both ω1 and ω2 (i.e., ( ω1 + ω2 ) ), or much lower than
both ω1 and ω2 (i.e., ω1 − ω2 ).

Either way, these new signals will typically be outside the

amplifier bandwidth!

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Two-Tone Intermodulation 5/7

Q: I thought you said these

“two-tone” intermodulation
products were some “big
problem”. These sons of a
gun appear to be no more a
problem than the harmonic
signals!

A: This observation is indeed correct for second-order, two-

tone intermodulation products. But, we have yet to examine the
third-order terms! I.E.,

v3out = C vin3
3
= C ( a cos ω1t + a cos ω2t )

If we multiply this all out, and again apply our trig knowledge, we
find that a bunch of new third-order signals are created.

Among these signals, of course, are the second harmonics

cos 3ω1t and cos 3ω2t . Additionally, however, we get these new
signals:

cos ( 2ω2 − ω1 )t and cos ( 2ω1 − ω2 )t

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Two-Tone Intermodulation 6/7

Note since cos ( −x ) = cos x , we can equivalently write these

terms as:
cos ( ω1 − 2ω2 )t and cos ( ω2 − 2ω1 )t

Either way, it is apparent that the third-order products include

signals at frequencies ω1 − 2ω2 and ω2 − 2ω1 .

Now lets look at the output spectrum with these new third-
order products included:

P/Hz First-order products

ω1 ω2 Second-order products
Third-order products

2ω1 2ω2

ω1 − ω2 2ω1 − ω2 2ω2 − ω1 ω1 + ω2 ω

Now you should see the problem! These third-order products

are very close in frequency to ω1 and ω2 . They will likely lie
within the bandwidth of the amplifier!

For example, if f1 =100 MHz and f2 =101 MHz, then 2f2 -f1 =102
MHz and 2f1 -f2 = 99 MHz. All frequencies are well within the
FM radio bandwidth!

Jim Stiles The Univ. of Kansas Dept. of EECS

9/13/2006 Two-Tone Intermodulation 7/7

Thus, these third-order, two-tone intermodulation products

are the most significant distortion terms.

This is why we are most concerned with the third-order

intercept point of an amplifier!

I only use amplifiers with

the highest possible
3rdorder intercept point!

Jim Stiles The Univ. of Kansas Dept. of EECS

9/20/2006 The Amplifier Spec Sheet 1/4

The Amplifier Spec Sheet

Here’s a list of some of the most important amplifier
specifications:

Gain (dB)

A numeric value that specifies the average gain over the

amplifier bandwidth. Note that this value is for the case where
both the source connected to input, and the load connected to
the output, are matched to Z0.

Typical values for single-stage amplifiers are 8 dB to 25 dB.

Bandwidth (Hz)

Typically, bandwidth is defined as the frequencies where the

amplifier will produce gain within 3 dB of the nominal gain value
expressed above.

For example, if the gain of an amplifier is stated as 17dB, then

the bandwidth would specify the range of frequencies for which
the gain is 14 dB or greater.

Generally speaking then, the lower and upper frequency values

are provided (e.g., amplifier bandwidth is 2.2 GHz to 4.7 GHz).

Amplifiers can have an extremely wide bandwidth (e.g., multiple

octaves), but generally we find a trade between gain and

Jim Stiles The Univ. of Kansas Dept. of EECS

9/20/2006 The Amplifier Spec Sheet 2/4

bandwidth (sound familiar?)—the wider the bandwidth, the

lower the gain (and vice versa).

Gain Variance(dB)

Be careful! This parameter can have a variety of names (e.g.,

gain ripple) and definitions.

Generally it describes the gain flatness over the middle portion

of the amplifier bandwidth. For example, if our 17dB amplifier
has a gain variance of +/- 1.0 dB, then the gain might vary from
16dB to 18dB across the amplifier bandwidth.

This parameter sometimes also refers to the variation in gain a

function of temperature, or specifies the variation in the
manufacturing process.

Typical values are +/- 0.5 dB to +/- 2.0 dB

Input Impedance (S11, Zin, Γ, return loss, VSWR)

Amplifiers are generally well-matched over their operating

bandwidth. There are (as we have discovered) a variety of ways
to express this match. Often the worst-case value over the
operating bandwidth is provided (e.g., return loss > 30 dB over
the operating frequency).

Typically, an amplifier input port return loss is 30 dB or more,

although this value typically gets worse and bandwidth
increases.

Jim Stiles The Univ. of Kansas Dept. of EECS

9/20/2006 The Amplifier Spec Sheet 3/4

Output Impedance (S11, Zin, Γ, return loss, VSWR)

See above.

Reverse Isolation (dB)

This value can change markedly (in dB)over the amplifier

bandwidth, and so a worst-case value is often provided (e.g.
>40dB over the operating bandwidth).

Typical values are 35 dB or greater.

D.C. Power (Nominal D.C. voltage and current)

Generally, a microwave amplifier requires a regulated DC

voltage supply (e.g., 15.0 V). The DC current can vary, and
typically a maximum value is given. This leads to a maximum DC
power requirement (I’ll leave the description of this value up to
you).

A fairly standard supply voltage for low-power amplifiers is

+15.0 V DC.

1 dB compression point (Watts, dBm, dBw)

This again is determined in the operating bandwidth of the

amplifier. This value is considered to be the largest output
power the amplifier can provide.

For low-power (i.e., small-signal) amplifiers, typical values range

from +10 dBm to +25 dBm.
Jim Stiles The Univ. of Kansas Dept. of EECS
9/20/2006 The Amplifier Spec Sheet 4/4

3rd order intercept point (Watts, dBm, dBw)

Remember, the larger this value, the better!

Typical values for small signal amplifiers range from +15 dBm
to +40 dBm.

Noise Figure (dB)

A very important amplifier parameter. We will learn about this

later! The smaller the noise figure, the better.

Typical values range from 1.0 dB to 6.0 dB. Amplifiers with

the best noise figure often have comparatively lower gain.

Everything Else (??)

The above list of amplifier specifications is by no means

complete, unambiguous, or in any way standard.

The reason for this is that there is no complete, unambiguous,

or standard list of specifications!

It is up to YOU—the radio engineer—to determine if a

particular amplifier meets the needs of your particular radio
application or design specifications.

Or, you must be able to write a clear, complete, unambiguous

specification that results in an amplifier that meets your needs!

Jim Stiles The Univ. of Kansas Dept. of EECS