Berkeley
Distortion Metrics
Prof. Ali M. Niknejad
U.C. Berkeley
Copyright
c 2016 by Ali M. Niknejad
March 7, 2016
1 / 34
Gain Compression
2 / 34
Gain Compression
dVo
dVi dVo
dVi
Vo Vo
Vi Vi
The large signal input/output relation can display gain
compression or expansion. Physically, most amplifier
experience gain compression for large signals.
The small-signal gain is related to the slope at a given point.
For the graph on the left, the gain decreases for increasing
amplitude.
3 / 34
1 dB Compression Point
Po,−1 dB
! !
! Vo !
! !
! Vi !
Vi Pi,−1 dB
Gain compression occurs because eventually the output signal
(voltage, current, power) limits, due to the supply voltage or
bias current.
If we plot the gain (log scale) as a function of the input
power, we identify the point where the gain has dropped by
1 dB. This is the 1 dB compression point. It’s a very
important number to keep in mind.
4 / 34
Apparent Gain
Recall that around a small deviation, the large signal curve is
described by a polynomial
so = a1 si + a2 si2 + a3 si3 + · · ·
For an input si = S1 cos(ω1 t), the cubic term generates
1
S13 cos3 (ω1 t) = S13 cos(ω1 t) (1 + cos(2ω1 t))
2
1 2
= S13 cos(ω1 t) + cos(ω1 t) cos(2ω1 t)
2 4
Recall that 2 cos a cos b = cos(a + b) + cos(a − b)
3 1 1
= S1 cos(ω1 t) + (cos(ω1 t) + cos(3ω1 t))
2 4
5 / 34
Apparent Gain (cont)
Collecting terms
3 1
= S13 cos(ω1 t) + cos(3ω1 t)
4 4
The apparent gain of the system is therefor
So,ω1 a1 S1 + 34 a3 S13
G= =
Si,ω1 S1
3 2 3 a3 2
= a1 + a3 S1 = a1 1 + S = G (S1 )
4 4 a1 1
If a3 /a1 < 0, the gain compresses with increasing amplitude.
6 / 34
1-dB Compression Point
Let’s find the input level where the gain has dropped by 1 dB
3 a3 2
20 log 1 + S = −1 dB
4 a1 1
3 a3 2
S = −0.11
4 a1 1
s
4 a1 √
S1 = × 0.11 = IIP3 − 9.6 dB
3 a3
The term in the square root is called the third-order intercept
point (see next few slides).
7 / 34
Intermodulation Distortion Intercept Points
8 / 34
Intercept Point IP2
Pout
(dBm)
OIP2 IP2
0
-10
10 dBc
-20
nd
d
2n
Fu 20 dBc
-30
-40
IIP2
-50 -40 -30 -20 -10 Pin
(dBm)
The extrapolated point where IM2 = 0 dBc is known as the
second order intercept point IP2 .
9 / 34
Properties of Intercept Point IP2
Since the second order IM distortion products increase like si2 ,
we expect that at some power level the distortion products
will overtake the fundamental signal.
The extrapolated point where the curves of the fundamental
signal and second order distortion product signal meet is the
Intercept Point (IP2 ).
At this point, then, by definition IM2 = 0 dBc.
The input power level is known as IIP2 , and the output power
when this occurs is the OIP2 point.
Once the IP2 point is known, the IM2 at any other power level
can be calculated. Note that for a dB back-off from the IP2
point, the IM2 improves dB for dB
10 / 34
Intercept Point IP3
Pout
(dBm)
OIP3 IP3
10
-10
nd 20 dBc
Fu
-20
rd
dBc
Thi
-30
IIP3
-50 -40 -30 -20 -10 Pin
(dBm)
The extrapolated point where IM3 = 0 dBc is known as the
third-order intercept point IP3 .
11 / 34
Properties of Intercept Point IP3
Since the third order IM distortion products increase like si3 ,
we expect that at some power level the distortion products
will overtake the fundamental signal.
The extrapolated point where the curves of the fundamental
signal and third order distortion product signal meet is the
Intercept Point (IP3 ).
At this point, then, by definition IM3 = 0 dBc.
The input power level is known as IIP3 , and the output power
when this occurs is the OIP3 point.
Once the IP3 point is known, the IM3 at any other power level
can be calculated. Note that for a 10 dB back-off from the
IP3 point, the IM3 improves 20 dB.
12 / 34
Intercept Point Example
From the previous graph we see that our amplifier has an
IIP3 = −10 dBm.
What’s the IM3 for an input power of Pin = −20 dBm?
Since the IM3 improves by 20 dB for every 10 dB back-off, it’s
clear that IM3 = 20 dBc
What’s the IM3 for an input power of Pin = −110 dBm?
Since the IM3 improves by 20 dB for every 10 dB back-off, it’s
clear that IM3 = 200 dBc
13 / 34
Calculated IIP2/IIP3
We can also calculate the IIP points directly from our power
series expansion. By definition, the IIP2 point occurs when
a2
IM2 = 1 = Si
a1
Solving for the input signal level
a1
IIP2 = Si =
a2
In a like manner, we can calculate IIP3 s
3 a3 2 4 a1
IM3 = 1 = Si IIP3 = Si =
4 a1 3 a3
14 / 34
Blockers or Jammers
15 / 34
Blocker or Jammer
channel
Interference
Signal
LNA
Consider the input spectrum of a weak desired signal and a
“blocker”
Si = S1 cos ω1 t + s2 cos ω2 t
| {z } | {z }
Blocker Desired
We shall show that in the presence of a strong interferer, the
gain of the system for the desired signal is reduced. This is
true even if the interference signal is at a substantially
different frequency. We call this interference signal a
“jammer”.
16 / 34
Blocker (II)
Obviously, the linear terms do not create any kind of
desensitization. The second order terms, likewise, generate
second harmonic and intermodulation, but not any
fundamental signals.
In particular, the cubic term a3 Si3 generates the jammer
desensitization term
Si3 = S13 cos3 ω1 t + s23 cos3 ω2 t + 3S12 s2 cos2 ω1 t cos ω2 t+
3S1 s22 cos2 ω2 t cos ω1 t
The first two terms generate cubic and third harmonic.
The last two terms generate fundamental signals at ω1 and
ω2 . The last term is much smaller, though, since s2 S1 .
17 / 34
Blocker (III)
The blocker term is therefore given by
1
a3 3S12 s2 cos ω2 t
2
This term adds or subtracts from the desired signal. Since
a3 < 0 for most systems (compressive non-linearity), the effect
of the blocker is to reduce the gain
a1 s2 + a3 32 S12 s2
App Gain =
s2
3 2 3 a3 2
= a1 + a3 S1 = a1 1 + S
2 2 a1 1
18 / 34
Out of Band 3 dB Desensitization
Let’s find the blocker power necessary to desensitize the
amplifier by 3 dB. Solving the above equation
3 a3 2
20 log 1 + S = −3 dB
2 a1 1
We find that the blocker power is given by
POB = P−1 dB + 1.2 dB
It’s now clear that we should avoid operating our amplifier
with any signals in the vicinity of P−1 dB , since gain reduction
occurs if the signals are larger. At this signal level there is also
considerable intermodulation distortion.
19 / 34
Distortion of AM Signals
Consider a simple AM signal (modulated by a single tone)
s(t) = S2 (1 + m cos ωm t) cos ω2 t
where the modulation index m ≤ 1. This can be written as
m m
s(t) = S2 cos ω2 t + cos(ω2 − ωm )t + cos(ω2 + ωm )t
2 2
The first term is the RF carrier and the last terms are the
modulation sidebands
Note that the AM modulation can be analog or digital. In a
digital case, the actual modulation is likely to be complex so
that the two sidebands are no longer symmetric, but the
analysis that follows still applies.
20 / 34
Cross Modulation
Cross modulation occurs in AM systems (e.g. video cable
tuners, QAM digital modulation)
The modulation of a large AM signal transfers to another
carrier going through the same amplifier (or non-linear system)
Si = S1 cos ω1 t + S2 (1 + m cos ωm t) cos ω2 t
| {z } | {z }
wanted interferer
CM occurs when the output contains a term like
K (1 + δ cos ωm t) cos ω1 t
Where δ is called the transferred modulation index
21 / 34
Cross Modulation (cont)
For So = a1 Si + a2 Si2 + a3 Si3 + · · · , the term a2 Si2 does not
produce any CM
The term
a3 Si3 = · · · + 3a3 S1 cos ω1 t (S2 (1 + m cos ωm t) cos ω2 t)2 is
expanded to
= · · · + 3a3 S1 S22 cos ω1 t(1 + 2m cos ωm t + m2 cos2 ωm t)×
1
2 (1 + cos 2ω2 t)
Grouping terms we have in the output
a3 2
So = · · · + a1 S1 (1 + 3 S m cos ωm t) cos ω1 t
a1 2
22 / 34
CM Definition
unmodulated waveform (input) modulated waveform due to CM
Transferred Modulation Index
CM =
Incoming Modulation Index
a3
CM = 3 S22 = 4IM3
a1
= IM3 (dB) + 12dB
= 12HD3 = HD3 (dB) + 22dB
23 / 34
Calculation Tools
24 / 34
Series Inversion
Sometimes it’s easier to find a power series relation for the
input in terms of the output. In other words
Si = a1 So + a2 So2 + a3 So3 + · · ·
But we desire the inverse relation
So = b1 Si + b2 Si2 + b3 Si3 + · · ·
To find the inverse relation, we can substitute the above
equation into the original equation and equate coefficient of
like powers.
Si = a1 (b1 Si + b2 Si2 + b3 Si3 + · · · ) + a2 ( )2 + a3 ( )3 + · · ·
25 / 34
Inversion (cont)
Equating linear terms, we find, as expected, that a1 b1 = 1, or
b1 = 1/a1 .
Equating the square terms, we have
0 = a1 b2 + a2 b12
a2 b12 a2
b2 = − =− 3
a1 a1
Finally, equating the cubic terms we have
2a22 a3
b3 = − 4
0 = a1 b3 + a2 2b1 b2 + a3 b13 a15 a1
It’s interesting to note that if one power series does not have
cubic, a3 ≡ 0, the inverse series has cubic due to the first
term above.
26 / 34
Cascade
IIP 2 GA
V
IIP 3 GA
P
IIP 2A IIP 2B
IIP 3A IIP 3B
Another common situation is that we cascade two non-linear
systems, as shown above. we have
y = f (x) = a1 x + a2 x 2 + a3 x 3 + · · ·
z = g (y ) = b1 y + b2 y 2 + b3 y 3 + · · ·
We’d like to find the overall relation
z = c1 x + c2 x 2 + c3 x 3 + · · ·
27 / 34
Cascade Power Series
To find c1 , c2 , · · · , we simply substitute one power series into
the other and collect like powers.
The linear terms, as expected, are given by
c1 = b1 a1 = a1 b1
The square terms are given by
c2 = b1 a2 + b2 a12
The first term is simply the second order distortion produced
by the first amplifier and amplified by the second amplifier
linear term. The second term is the generation of second
order by the second amplifier.
28 / 34
Cascade Cubic
Finally, the cubic terms are given by
c3 = b1 a3 + b2 2a1 a2 + b3 a13
The first and last term have a very clear origin. The middle
terms, though, are more interesting. They arise due to second
harmonic interaction. The second order distortion of the first
amplifier can interact with the linear term through the second
order non-linearity to produce cubic distortion.
Even if both amplifiers have negligible cubic, a3 = b3 ≡ 0, we
see the overall amplifier can generate cubic through this
mechanism.
29 / 34
Cascade Example
In the above amplifier, we can decompose the non-linearity as
a cascade of two non-linearities, the Gm non-linearity
id = Gm1 vin + Gm2 vin2 + Gm3 vin3 + · · ·
And the output impedance non-linearity
vo = R1 id + R2 id2 + R3 id3 + · · ·
The output impedance can be a non-linear resistor load (such
as a current mirror) or simply the load of the device itself,
which has a non-linear component.
30 / 34
IIP2 Cascade
Commonly we’d like to know the performance of a cascade in
terms of the overall IIP2. To do this, note that IIP2 = c1 /c2
c2 b1 a2 + b2 a12 a2 b2
= = + a1
c1 b1 a1 a1 b1
This leads to
1 1 a1
= A
+
IIP2 IIP2 IIP2B
This is a very intuitive result, since it simply says that we can
input refer the IIP2 of the second amplifier to the input by the
voltage gain of the first amplifier.
31 / 34
IIP2 Cascade Example
Example 1: Suppose the input amplifiers of a cascade has
IIP2A = +0 dBm and a voltage gain of 20 dB. The second
amplifier has IIP2B = +10 dBm.
The input referred IIP2B
i = 10 dBm − 20 dB = −10 dBm
This is a much smaller signal than the IIP2A , so clearly the
second amplifier dominates the distortion. The overall
distortion is given by IIP2 ≈ −12 dB.
Example 2: Now suppose IIP2B = +20 dBm. Since
IIP2Bi = 20 dBm − 20 dB = 0 dBm, we cannot assume that
either amplifier dominates.
Using the formula, we see the actual IIP2 of the cascade is a
factor of 2 down, IIP2 = −3 dBm.
32 / 34
IIP3 Cascade
Using the same approach, let’s start with
c3 b1 a3 + b2 a1 a2 2 + b3 a13 a3 b3 2 b2
= = + a + 2a2
c1 b a a1 a1 b1 1 b1
The last term, the second harmonic interaction term, will be
neglected for simplicity. Then we have
1 1 a12
= +
IIP32 IIP32A IIP32B
Which shows that the IIP3 of the second amplifier is input
referred by the voltage gain squared, or the power gain.
33 / 34
LNA/Mixer Example
A common situation is an LNA and mixer cascade. The mixer
can be characterized as a non-linear block with a given IIP2
and IIP3.
In the above example, the LNA has an IIP3A = −10 dBm and
a power gain of 20 dB. The mixer has an IIP3B = −20 dBm.
If we input refer the mixer, we have
IIP3Bi = −20 dBm − 20 dB = −40 dBm.
The mixer will dominate the overall IIP3 of the system.
34 / 34
