### NOISE - UCSB ECE Dept.

```ECE145A/ECE218A
Performance Limitations of Amplifiers
1. Distortion in Nonlinear Systems
The upper limit of useful operation is limited by distortion. All analog systems and
components of systems (amplifiers and mixers for example) become nonlinear when
driven at large signal levels. The nonlinearity distorts the desired signal. This distortion
exhibits itself in several ways:
1. Gain compression or expansion (sometimes called AM – AM distortion)
2. Phase distortion (sometimes called AM – PM distortion)
3. Unwanted frequencies (spurious outputs or spurs) in the output spectrum. For a
single input, this appears at harmonic frequencies, creating harmonic distortion or
HD. With multiple input signals, in-band distortion is created, called
intermodulation distortion or IMD.
When these spurs interfere with the desired signal, the S/N ratio or SINAD (Signal to
noise plus distortion ratio) is degraded.
Gain Compression.
The nonlinear transfer characteristic of the component shows up in the grossest sense
when the gain is no longer constant with input power. That is, if Pout is no longer
linearly related to Pin, then the device is clearly nonlinear and distortion can be expected.
Pout
Pin
P1dB, the input power required to compress the gain by 1 dB, is often used as a simple to
measure index of gain compression. An amplifier with 1 dB of gain compression will
generate severe distortion.
Distortion generation in amplifiers can be understood by modeling the amplifier’s
transfer characteristic with a simple power series function:
Vout = a1Vin − a3Vin3
Of course, in a real amplifier, there may be terms of all orders present, but this simple
cubic nonlinearity is easy to visualize. The coefficient a1 represents the linear gain; a3 the
1
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
distortion. When the input is small, the cubic term can be very small. At high input
levels, much nonlinearity is present. This leads to gain compression among other
undesirable things. Suppose an input Vin =A sin (ωt) is applied to the input.
Vout
⎡
3a3 A2 ⎤
1
3
= A ⎢ a1 −
⎥ sin(ωt ) + a3 A sin(3ωt )
4 ⎦
4
⎣
Gain Compression
Third Order Distortion
Gain compression is a useful index of distortion generation. It is specified in terms of an
input power level (or peak voltage) at which the small signal conversion gain drops off
by 1 dB.
The example above assumes that a simple cubic function represents the nonlinearity of
the signal path. When we substitute Vin(t) = A sin (ωt) and use trig identities, we see a
term that will produce gain compression:
2
A(a1 - 3a3A /4).
If we knew the coefficient a3, we could predict the 1 dB compression input voltage.
Typically, we obtain this by measurement of gain vs. input voltage.
Harmonic Distortion
We also see a cubic term that represents the third-order harmonic distortion (HD) that
also is caused by the nonlinearity of the signal path. Harmonic distortion is easily
removed by filtering; it is the intermodulation distortion that results from multiple signals
that is far more troublesome to deal with.
Note that in this simple example, the fundamental is proportional to A whereas the third3
order HD is proportional to A . Thus, if Pout vs. Pin were plotted on a dBm scale, the
HD power will increase at 3 times the rate that the fundamental power increases with
input power. This is often referred to as being “well behaved”, although given the choice,
we could easily live without this kind of behavior!
2
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
Intermodulation Distortion
Let’s consider again the simple cubic nonlinearity a3vin3. When two inputs at ω1 and ω2
are applied simultaneously to the RF input of the system, the cubing produces many
terms, some at the harmonics and some at the IMD frequency pairs. The trig identities
show us the origin of these nonidealities. [4]
We will be mainly concerned with the third-order IMD. (actually, any distortion terms
can create in-band signals – we will discuss this later). IMD is especially troublesome
since it can occur at frequencies within the signal bandwidth. For example, suppose we
have 2 input frequencies at 899.990 and 900.010 MHz. Third order products at 2f1 - f2
and 2f2 - f1 will be generated at 899.980 and 900.020 MHz. These IM products may fall
within the filter bandwidth of the system and thus cause interference to a desired signal.
The spectrum would look like this, where you can see both third and fifth order IM.
3
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
Pout (dBm)
x = IIP3 - PIN
OIP3
P1
am
d
fun
tal
n
e
x
2x
third-order IMD
PIMD
Pin (dBm)
IIP3
PIN
IIP3 = PIN +
1
( P1 − PIMD )
2
IMD power, just as HD power, will have a slope of 3 on a Pout vs Pin (dBm) plot. A
widely-used figure of merit for IMD is the third-order intercept (TOI) point. This is a
fictitious signal level at which the fundamental and third-order product terms would
intersect. In reality, the intercept power is 10 to 15 dBm higher than the P1dB gain
compression power, so the circuit does not amplify or operate correctly at the IIP3 input
level. The higher the TOI, the better the large signal capability of the system. If
specified in terms of input power, the intercept is called IIP3. Or, at the output, OIP3.
This power level can’t be actually reached in any practical amplifier, but it is a calculated
figure of merit for the large-signal handling capability of any RF system.
It is common practice to extrapolate or calculate the intercept point from data taken at
least 10 dBm below P1dB. One should check the slopes to verify that the data obeys the
expected slope = 1 or slope = 3 behavior. The TOI can be calculated from the following
geometric relationship:
OIP3 = (P1 − PIMD)/2 + P1
Also, the input and output intercepts (in dBm) are simply related by the gain (in dB):
OIP3 = IIP3 + power gain.
Other higher odd-order IMD products, such as 5th and 7th, are also of interest, and can
also be defined in a similar way, but may be less reliably predicted in simulations unless
the device model is precise enough to give accurate nonlinearity in the transfer
characteristics up to the 2n-1th order.
4
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
Cross Modulation
In addition, the cross-modulation effect can also be seen in the equation above. The
amplitude of one signal (say ω1) influences the amplitude of the desired signal at ω2
through the coefficient 3V12V2a3/2. A slowly varying modulation envelope on V1 will
cause the envelope of the desired signal output at ω2 to vary as well since this
fundamental term created by the cubic nonlinearity will add to the linear fundamental
term. This cross-modulation can have annoying or error generating effects at the output.
Second Order Nonlinearity
In the simplified model above, we have neglected second order nonlinear terms in the
series expansion. In many cases, an amplifier or other RF system will have some evenorder distortion as well. The transfer function then would look like this:
Vout = a1Vin + a2Vin2 + a3Vin3
If we once again apply two signals at frequencies ω1 and ω2 to the input, we obtain:
Vout 2 = a2 ⎡⎣V12 sin 2 (ω1t ) + V22 sin 2 (ω2t ) + 2V1V2 sin(ω1t )sin(ω2t ) ⎤⎦
The sin2 terms expand into:
1
1
a2V12 [1 − cos(2ω1t ) ] + a2V22 [1 − cos(2ω2t )]
2
2
From this, we can see that there is a DC term and a second harmonic term present for
each input. The DC term is proportional to the square of the voltage, therefore power.
This is one use of second-order nonlinearity – as a power sensor. The HD term is also
proportional to the square of the voltage. Thus, on a power out vs. power in plot, it has a
slope of 2.
5
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
When the next term is expanded, the product of two sine waves is seen to produce the
sum and difference frequencies.
a2V1V2 [ cos(ω2 − ω1 )t − cos(ω2 + ω1 )t ]
This can be both a useful property and a problem. The useful application is as a
frequency translation device, often called a mixer, a downconverter, or an upconverter.
The desired output is selected by inserting a filter at the output of the device.
Second order distortion, if generated by out-of-band signals, can also lead to interference
in-band as shown below. Preselection filtering can generally suppress this in narrowband
amplifiers, but it can be a big problem for wideband circuits.
A SOI, or second-order intercept can also be defined as shown below:
Pout (dBm)
OIP2
P1
PIMD
me
a
d
fun
l x
a
t
n
x
second-order IMD
Pin (dBm)
Pin
IIP2
The second-order IMD slope = 2. IIP2 can be calculated from measurement by:
IIP2 = Pin + P1 – PIMD
OIP2 = IIP2 + Power Gain = 2 P1 - PIMD
6
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
Measuring Intermodulation Distortion
Set the amplitude of generators at f1 and f2 to be equal.
Start at a very low input power using the variable attenuator, then increase power in steps
until you begin to see the IMD output on the spectrum analyzer. The resolution
bandwidth should be narrow so that the noise floor is reduced. This will allow visibility
of the IMD signal at lower power levels.
Plot the IMD power vs. input power and verify that the slope is close to 3. Then, you can
calculate the IIP3 as described previously.
Refer to the first part of the Harmonic Balance Simulation Tutorial on the course web
page.
7
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
How is the Third-Order Intercept Point affected by cascaded stages?
Gains multiply in a cascade: PO = Pi G(1) G(2) G(3)
(or add them if in dB)
Individual intercept points must be referred to the same reference plane. It can be either
at the input or the output. In this example, the output TOI, OIP3, is specified for each
stage.
1. Convert all OIPs from dBm to mW and gains from dB to a power ratio.
2. Let’s refer all of these OIPs to the output plane.
OIP3
G(3) OIP3(2)
G2) G(3) OIP3(1)
3. The third order intercept cascading relationship is:
1
1
1
1
=
+
+
OIP3 G ( 2) G (3) OIP3 (1) G (3) OIP3 ( 2) OIP3 (3)
IIP3 =
OIP3
G (1) G ( 2) G (3)
4. Convert the results back to dBm if desired.
8
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
Second order intercept cascading is accomplished by the following equations:
1
OIP2
IIP2 =
=
1
G (2) G (3) OIP2 (1)
+
1
G (3) OIP2 (2)
+
1
OIP2 (3)
OIP2
G (1)G (2) G (3)
Example: Third-order intercept of a receiver front end
1. Convert dBm to mW: OIP3(1) = 1 mW, OIP3(2) = 100 mW
Convert dB to a power ratio: G(1) = 10,
G(2) = 1
2. Refer to the output plane:
1/OIP3 = 1 + 1/100 = 1.01
3. IIP3 = OIP3/10 = 0.1
OIP3 = 1
(0 dBm)
(-10 dBm)
We can see that the LNA completely dominates the IIP3 in this example. IF we
eliminated the LNA, then OIP3 = OIP3(2) = 20 dBm and IIP = 20 dBm, a 30 dB
improvement!
What do we lose by eliminating the LNA?
9
rev. 12/29/10
€awt
Sec, 6,3
uV,
r{dyt^rar0,tnt>o. {D tZF
W:V,
A.Qec Pregg.
\4q4.
229
Distortion in Amplifiers ond the Intercepl Concept
Figure 6.'18 A coscode of two omplifiers, eoch with o known outPut inter-
*"L
cept. /,ir is the output intercept of the
frst stoge renormolized to the output
plone, ochieved by increosing lot bY
Q, the second stoge goin. lf the distortion products ore ossumed coherent,
ond oll intercepts ore normolized to
one plone, the equivolent intercept is
colculoted just os the net resistonce
of oorollel resistors is evoluoted,
Consider now the more general case where both amplifiers have finite output
intercepts. The analysis will be confined to third order imd although the approach
is easily extende/ to distortion of any order. Assume that the intercepts of both
stages have beer( normalized to the same plane in the cascade. The intercepts will
be designated by 1,, where the subscript n denotes the stage. D" will refer to a distortion
power while P" will describe the desired output power of the nth stage normalized
to the plane of interest.
If the fundamental defining concepts of the intercept are invoked in algebraic
terms instead of logarithmic units, the distortion power of the nth stage is Dn :
P3"/Ik. This power appears in a load resistance, rt. Hence, the corresponding distortion
voltage is Yo : (RDnltt, : (psft)tr2/Ir. The total distortion will come from the
As was the case with noise voltages, distortion voltages must be added with
care. If the voltages are phase related, they should be added algebraically. However,
if they are completely uncorrelated, they will add just as thermal noise voltages do,
as the root of the sum of the squares. There is usually a well-defined phase relationship
between signals with amplifiers. The worst case is when distortions from two stages
add exactly in phase. This will lead to the largest distortion. Some cases may exist
where distortion voltages are coherent (phase related) and cancel to lead to a distortionless amplifier. Like most physical phenomena, this is unusual and not the sort of
thing that a designer can depend upon. We rvill take the conservative approach of
choosing the worst possible case, that of algebraic addition of the distortion voltage,
assuming them to be in phase.
Using the worst case assumption, the total distortion voltage is Vr: V1 * Yz
,r:(;*;)
(PtPltrz
(6.3-10)
The corresponding power is then
Dr:i:V7
P'
;.;)'
(6.3-1 1)
Proctical Amplifers ond
230
Mixers
Chop. 6
From the earlier definition, the net or total intercept at the plane of definition is
I7:
(Pt
/Dfltrz
(6.3-12)
Further manipulation yields the final result
o: (i *;)-'
(6.3-13)
Equation 6.3-13 has a familiar form with an easy to remember analogy. If
intercepts are normalized to a single plane and are expressed as powers in milliwatts
or watis rather than logarithmic units, the total intercept at the plane of definition
is a sum similar to that for resistors in parallel. This applies only for the case of
coherent addition of distortion voltages for third order imd. Not only is this analysis
conservative to the extent that it is "worst case," but it works well in practice, predicting measured results with reasonable accuracy.
Consider an example, two identical amplifiers with a gain of 10 dB and an
output intercept of *15 dBm. If the two intercepts are normalized to the corresponding
ones at the output, they are *15 and *25 dBm. Converting to milliwatts, the two
intercepts are 31.62 and 316.2. Application of the resistors-in-parallel rule yields an
equivalent output intercept of 28.75 mW, or 14.59 dBm. Essentially, the imd is completely dominated by the output stage.
A more realistic design would be one with a "stronger" second stage. Assume
that the output intercept of the second stage is increased to *25 dBm. That of the
first stage is still *15 dBm, while both gains remain at 10 dB. The result is an
ouput intercept of *22 dBm. The output intercept of the first stage equals the input
intercept of the second to yield equal distortion contribution from each and a 3-dB
degradation over the intercept of an individual stage.
Generally, the last stage in a chain will determine the third order imd performance. This will be maintained so long as the output intercept of the previous stage
is greater than the input intercept of the last.
Consider first the question of gain compression in a common emitter bipolar amplifier.
From an intuitive viewpoint, we would expect the gain to begin to decrease significantly
when the collector signal current reaches a peak value equaling the dc bias current.
The signal current will then be varying from the bias level to twice that value and
ta zero on negative-going peaks. This assumes that the supply voltage is high enough
that no voltage limiting occurs. The load also effects the possibility of voltage limiting.
It is found experimentally that the l-dB gain compression point is well approximated by the current limiting described. Gain will still be present at higher levels
and the continued gain compression is gradual until a "saturated" output is reached.
Distortion is severe at high levels above the point of l-dB compression. A bipolar
transistor with a 50-fI collector termination will have a l-dB compression point of
ECE145A/ECE218A
Performance Limitations of Amplifiers
2. Next Topic: NOISE
Noise determines the minimum signal power (minimum detectable signal or MDS) at the
input of the system required to obtain a signal to noise ratio of 1. A S/N = 1 is usually
considered to be the lower acceptable limit except in systems where signal averaging or
processing gain is used. Noise figure is a figure of merit used to describe the amount of
degradation in S/N ratio that the system introduces as the signal passes through.
For some applications, the minimum signal power that is detectable is
important.
o 802.11
Noise limits the minimum signal that can be detected for a given signal input
power from the source or antenna.
We will identify sources of noise, and define related quantities of interest:
o S/N = Signal to noise ratio
o MDS = Minimum Detectable Signal
o F = Noise factor
o NF = 10 * log(F) = Noise figure
10
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
Noise Basics:
What is noise? How is it evident to us? Why is it important?
vn
vn
t
P
What:
1. Any unwanted random disturbance
2. Random carrier motion produces a current. Frequency and phase
are not predictable at any instant in time
3. The noise amplitude is often represented by a Gaussian probability
density function.
The cumulative area under the curve represents the probability of the event
occurring. Total area is normalized to 1.
Because of the random process, the average value is zero:
1 t +T
v n = lim
∫ [vn (t )] dt = 0
T
T →∞
t
1
1
We cannot predict vn(t), but the variance (standard deviation) is finite:
11
rev. 12/29/10
ECE145A/ECE218A
2
vn =
Performance Limitations of Amplifiers
2
1 t +T
2
lim
∫ [vn (t )] dt = σ
T →∞ T t
1
1
Often we refer to the rms value of the noise voltage or current:
2
vn,rms = v n
Sources of Noise in Circuits:
o Shot noise
forward-biased junctions
o Thermal Noise
any resistor
o Flicker (1/f) noise
trapping effects
Shot noise: This is due to the random carrier flow across a pn junction.
Electrons and holes randomly diffuse across the junction producing noise
current pulses that occur randomly in time. The steady state current
measured across a forward biased diode junction is really a large number of
discrete current pulses.
p
ID
I
The variance of this current:
2
1T
i = lim ∫ (I − I D ) dt = σ 2
T →∞ T 0
2
It can be shown that this mean square noise current can be predicted by
2
i = 2qI D B
12
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
where
q = charge of an electron = 1.6 x 10-19
ID = diode current
B = bandwidth in Hertz (sometimes called Δf)
The noise current spectral density:
2
i / B = 2qI D
o Independent of frequency (white noise)
o Independent of temperature for a fixed current
o Proportional to the forward bias current
o Gaussian probability distribution
1 mA of current corresponds to a noise current spectral density of
18 pA/√Hz
read: 18 picoamp per root Hertz
Thermal Noise: Thermal noise, sometimes called Johnson noise, is due to
random motion of electrons in conductors. It is unaffected by DC current
and exists in all conductors. Its spectral density is also frequency
independent, but is directly proportional to temperature. The noise
probability density is Gaussian.
2
v = 4kTRB
2
i = 4kTB / R
4kT = 1.66 x 10 –20 V-C
13
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
A 50 ohm resistor produces a noise voltage spectral density of
0.9 nV/√Hz
or a Norton equivalent noise current spectral density of
18 pA/√Hz
Flicker or 1/f noise. This noise source is most evident at very low
frequencies. It is hard to localize its physical mechanisms in most devices.
There is usually some 1/f noise contribution due to charge traps with long
time constants. The trap charge then is randomly released after some
relatively long period of time. 1/f noise is modeled by:
2
i /B = K
I
f
 K is a fudge factor. It can vary wildly from one type of transistor to
the next or even from one fabrication lot to the next.
 I is the current flowing through the device.
 B is the bandwidth.
Log (i2/B)
Log f
Corner frequency
 1/f noise can be described by a corner frequency.
 Carbon resistors exhibit 1/f noise; metal film resistors do not.
14
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
Noise can be modeled as a Thevenin equivalent voltage source or a Norton
equivalent current source. The noise contributed by the resistor is modeled
by the source, thus the resistor is considered noiseless.
R
in
vn
R
It is important to note that noise sources:
o Do not have polarity (the arrow is just to distinguish current
from voltage)
o Do not add algebraically, but as RMS sums
v
vn1
2
n , total
2
n1
2
n2
R1
vn2
= v + v = 4kTBR1 + 4kTBR2
R2
If the sources are correlated (derived from the same physical noise source),
then there is an additional term:
2
2
2
v n,total = v n1 + v n 2 + 2Cvn1 vn 2
C can vary between –1 and 1.
15
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
The available noise power can be calculated from the RMS noise voltage or
current:
2
2
v
i R
Pav = n = n = kTB
4R
4
That is, the available noise power from the source is
o independent of resistance
o proportional to temperature
o proportional to bandwidth
o has no frequency dependence
Pav = 4 x 10 -21 watt
in a 1 Hz bandwidth at the standard noise room temperature of 290 K. If
converted to dBm = 10 log(P/10-3), this power becomes
- 174 dBm/Hz
We are generally interested in the noise power in other bandwidths than 1 Hz. It’s easy
to calculate:
P = kTB
where kT = -174 dBm
To convert bandwidth in Hertz to dB: 10 log B
EX: Suppose your B = 1000 Hz.
P = kTB.
In dBm, P = -174 + 10 log (1000) = -174 + 30 = -144 dBm
16
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
Can a resistor produce infinite noise voltage?
Vn 2 = 4kTBR
R
Equivalent circuit for noisy resistor.
Vn
Vno
C
Always some shunt capacitance.
Low Pass
log10 Vno
Vno = Vn
1
1+ ω 2 C 2 R 2
9R
4R
R
f
to find total noise power:
∫
∞
0
2
Vno df =
kT
= Vno2
C
total noise power is independent of R
17
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
Noise Equivalent Bandwidth
An amplifier or filter has a nonideal frequency response. Noise power transmitted
through is determined by the bandwidth.
A( f )
AM
vi 2
2
A( f )
vo 2
2
f
B
Noise power ∝ V (mean square voltage) – white noise
2
vi A( f ) = vo / Hz in a 1Hz interval
2
2
2
Summation over entire frequency band
∫
∞
o
vo ( f )df = v i
2
2
∫
∞
o
A ( f ) df
2
We choose an equivalent BW, B, with rectangular profile whose area is the same.
∞
Am B = ∫o A( f ) df
2
B=
2
2
1 ∞
2 ∫o A( f ) df
Am
This is the definition of bandwidth that we will assume in subsequent
calculations.
18
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
Signal-to-noise ratio
Several definitions
SNR =
PS
S
=
PN N
generally use available power
Pav =
VS
+
-
VS 2
4R
rms voltage VS
R
S+N
S+N+D
and
N
N
Why is S/N important?
Affects the error rate when receiving information.
Ref: S. Haykin,
Communication Systems, 4th
ed., Wiley, 2001
19
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
Noise Factor, F:
Savi, Navi
Ga
Savo, Navo
is a measure of how much noise is added by a component such as an
amplifier.
F=
Savi / N avi
>1
Savo / N avo
because S/N at input will always be greater than S/N at output, F > 1.
Noise factor represents the extent that S/N is degraded by the system.
F=
=
Gav =
total noise power available at output
noise power available at output due to source @ 290k
Navo
Navi ⋅ Gav
source at 290K
Savo
Savi
(S / N )avi
F=
(S / N )avo
Noise Figure:
The higher the noise factor (or
noise figure), the larger the
amplifier.
NF = 10 log10 F
Noise Factor, F:
Si , Ni
Ga
So , No
is a measure of how much noise is added by a component such as an
amplifier.
20
rev. 12/29/10
ECE145A/ECE218A
Ex.
Performance Limitations of Amplifiers
RS = 50Ω
vS = 1.4VμV
S = 1μV
(So / N o ) = Si F/ N i
Gav
+
-
RL
Gav = 10dB
NF = 3dB
amplifier specification
B = 10 Hz
6
signal available power
vs2
2 × 10 −12
S avi =
=
= 5 × 10 −15W = − 113 dBm
8 Rs
400
vs2 10 −12
signal av. pwr. = Savi =
=
= 5 × 1015 N W
⇒ −113dBm
4RS
200
noise av. pwr. = N avi = kTB = −174 + 60 = −114dBm
Since noise power increases with B
10 log10 B = 60dB (in this example)
⎛S ⎞
⎛S ⎞
10 log⎜ avo ⎟ = 10 log⎜ avi ⎟ − NF
⎝ Navo ⎠
⎝ Navi ⎠
= −113 − (−114) − 3
= 1dB − 3dB = −2dB (not very good)
How can So / N o be improved?
1. Reduce F.
Slight room for improvement
2. Reduce B.
Major improvement if application can tolerate
reduced B.
3. Increase antenna gain. Lots of room for improving Si/Ni
21
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
say B = 105
Navi = −174 + 50 = −124 dBm
Savi
S
= 11dB and avo = 8dB
Navi
Navo
Ex.
Noise Floor of Spectrum Analyzer
typical NF ≅ 25dB for SA .
N AVO = NAVI ⋅ F ⋅ GAV
N AVI = ( −174 dBm / Hz ) + 10 log B
NF = 25dB
resolution bandwidth (RBW)
GAV = 1 (0dB)
RBW
N AVO
1kHz
10kHz
−119 dBm
−109
−99
100kHz
etc.
We will see later how this can be improved.
22
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
The excess noise added by an active circuit such as an amplifier can also be
modeled by an extra resistor at an effective input noise temperature, Te .
G
N avi = kTo B
N avo
noisy
amp
is equivalent to:
noiseless
Σ
Pav
G
kTe B
N avo = k (To + Te ) B ⋅ G
(useful when To ≠ 290k )
In terms of noise factor:
F=
=
noise out due to DUT + noise out due to source
Noise out due to source
kTe BG + kTo BG
T
= 1+ e
kTo BG
To
or Te = 290( F − 1)
(where F is a number, not dB)
23
rev. 12/29/10
ECE145A/ECE218A
Significance of Te :
Performance Limitations of Amplifiers
excess noise.
N avo (total) = kBG (To + Te )
due to
source
resistor
due to
amplifier
Example: NF = 1dB ⇒ F = 1.26
= 1 + Te To = 1 + Te 290
so Te = 75K
total output noise ⇒ 290 + 75 = 365K equivalent source temp
So what?
Not major increase in noise power. Further reduction
in F may not be justified.
But, for space application: To = 20K is possible.
Then T = To + Te = 20 + 75 = 95K
F or NF at room temperature doesn’t reveal this so
clearly.
F = 1 + 75/20 = 4.5
24
(NF = 7 dB)
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
Use Available gain.
Why available gain?
Noise power defined as available power. Cascading of noise is more convenient
when GA is used.
Second Stage Noise Contribution
F1
Ni = kTo B
F2
G1
No1
G2
No2
RS
T1 = eff.
noise temp
@ input
T2
No1 = k(To + T1 )BG1
No2 = k(To + T1 )BG1G2 + kT2 BG2
To get total input referred noise power:
No2
= Ni (equivalent) = k(To + T1 )B + kT2 B / G1
G1G2
excess noise at input:
kT1 B + kT2 B / G1
Recall that F = 1 +
Te
To
Te = T1 + T2 G1
T1
T
+ 2
To To G1
F1 + F2 − 1
G1
FTOTAL = 1 +
Third Stage:
+
25
F3 − 1
G1 G2
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
(S N )IN
F1
F2
F3
G1
G2
G3
(S N )OUT
Fi = Noise Factor ⎫
⎬ not in dB
G i = Available Gain ⎭
FTOTAL = F1 +
F2 − 1 F3 − 1
+
+ ...
G1
G1G2
= Input Total Noise Factor
(S N )IN
(S N )OUT
Or:
26
= FTOTAL
(S N )OUT dB = (S N )IN dB − NFTOTAL
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
Total available gain of cascade = Ga 1 Ga 2 Ga 3 ...
1. If noise figure is important in a receiver, it is standard procedure to design so
that the first stage sets the noise performance.
FTOTAL = F1 +
F2 − 1 F3 − 1
+
G1
G1G2
This will require a large enough G1 to diminish the noise contribution of the
second stage.
2. How is the minimum detectable signal or MDS defined?
* at a given B (very important)
PMDS ⇒
S+N
= 3dB or S = N
N
S
= O dB
N
PMDS = 10log(kTB ) + NF ( dB )
OR
PMDS = − 174 dBm / Hz + 10log B + NF ( dB )
27
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
Noise figure of Passive Networks
ex. attenuator
filter
matching network
No active components. Only resistors and reactances.
passive
network
ZS
Gav
PS
F
Navi = kTo B
Navo = kTo B
no excess noise is generated by network
Savo
= Gav
Savi
so, (S N )i =
PS
kTo B
(S N )o =
G ⋅ PS
kTo B
F=
(S N )i
(S N )o
=
1
G
Noise factor is just the inverse of gain.
or, NF = −G(dB )
ex. 10dB attenuator
Gav = −10dB
NF = 10dB
28
rev. 12/29/10
ECE145A/ECE218A
Performance Limitations of Amplifiers
Measuring NF: HOT-COLD NF Technique
You can use a calibrated noise source for measuring NF.
50Ω
noise
source
B1
B2
LNA
DUT
B3 > B2
Power
Meter
preamp
Bn >> B1 < B2
The advantage here is that we don’t need to know noise equivalent BW
accurately.
Noise source has very wide BW compared with system under test.
PH = noise power with source on = kTH B
TH = effective noise temp. of source
Po = kTo B = noise power with source off.
To = 290k
Excess Noise Ratio = ENR =
PH − Po TH
=
−1
Po
To
⎞
⎛T
ENR(dB ) = 10 log10 ⎜ H − 1⎟
⎝ To
⎠
Y factor for noise source:
YS =
29
PH TH
=
Po To
rev. 12/29/10
ECE145A/ECE218A Performance Limitations of Amplifiers
So, we can use the noise source instead of the signal generator.
1. Source off. Noise power at meter:
P1 = F kTo B AT
total noise factor
transducer gain
2. Source on.
P2 = P1 + Ys kT0 BAT
P2
Y
= Y = 1+ S
P1
F
Divide:
again, the transducer gain cancels, and now B cancels too. We can solve for F
from the measured P2 P1 .
F=
YS
Y −1
Noise factor – numerical ratios, not dB.
and
NF = 10 log F (dB)
30
rev. 12/29/10
Block diagram of a noise figure measurement system
Noise and distortion example
B1 = 5 MHz
B2= 100 MHz
B3= 100 MHz
B4= 10 kHz
P1
S
LO
G1= -3 dB
P2
G2= -6 dB
G3= 10 dB
NF2=6 dB
NF3=6 dB
IIP(2)=+10 dBm
IIP(3)=+4 dBm
G4= -3 dB
Assume source P2 is off. What is the minimum source power P1 in dBm that will
produce an output signal to noise ratio = 1?
First calculate noise figure:
Ftotal = F1 + (F2-1)/G1 + (F3-1)/(G1G2) + (F4-1)/(G1G2G3)
= 2 + 6
+
24
+
0.8 = 32.8
(15.1 dB NF)
Minimum signal level at input to produce (S/N)out = 1? First find the minimum
bandwidth in the chain: stage 4; B4 = 10 kHz
P1 = MDS = -174 dBm/Hz + 10 log B4 + NF = -119 dBm
Could we improve the noise figure? The 3rd stage is the major contributor. We do
not need such a wide band IF amplifier for a 10 KHz bandwidth, so this stage could
be redesigned for minimum noise figure. Even then, the total NF is high due to the
losses in stages 1 and 2. The first stage filter should be replaced with one with
lower loss, since its noise figure adds directly to the receiver total NF. The best way
to improve NF is to add an LNA, but this will have an impact on the IIP3.
1
Noise and distortion example
B1 = 5 MHz
B2= 100 MHz
B3= 100 MHz
B4= 10 kHz
P1
S
LO
G1= -3 dB
P2
G2= -6 dB
G3= 10 dB
NF2=6 dB
NF3=6 dB
IIP(2)=+10 dBm
IIP(3)=+4 dBm
G4= -3 dB
Now assume both sources are on and P1 = P2. How much source power will be
required to produce a third order intermodulation component of - 100 dBm at the
output?
First, we must calculate the input third-order intercept for the chain.
Refer the intercepts of stages 2 and 3 to the input of the filter at stage 1:
IIP(2)’ = IIP(2) + 3 dB = + 13 dBm
(20 mW)
IIP(3)’ = IIP(3) + 6 + 3 = +13 dBm
(IIP3total)-1 = 1/20 + 1/20 = 1/10
IIP3total = +10 dBm
Next, refer to the output. Total gain of the 4 stages = -2 dB
OIP3total = IIP3total – 2 dB = +8 dBm
Next we can plot the Pout vs Pin and easily calculate Pin required for the –100 dBm
IMD power.
2
X
OIP3= +8 dBm
X
2X
PIMD = -100 dBm
Pin
IIP3
From the plot, you can see that 3x = 108 dB. Thus, x = 36 dB
Pin = IIP3 – x = -26 dBm
Now we can calculate SFDR
3
Spurious Free Dynamic Range
SFDR =
2
[IIP3 - (10 log kTDf + NF )]
3
Pout (dBm)
Output noise
floor
fun
d
en
am
tal
third-order IMD
Pin (dBm)
MDS = 10 log(kTDf) + NF
IIP3 = +10 dBm
IIP3
MDS = -119 dBm
SFDR = 86 dB
Spurious free dynamic range measures the ability of a receiver system to operate
between noise limits and interference limits.
SFDR = 2 (IIP3 – MDS)/3
The maximum signal power is limited by distortion, which we describe by IIP3.
The spurious-free dynamic range (SFDR) is a commonly used figure of merit to
describe the dynamic range of an RF system. If the signal power is increased
beyond the point where the IMD rises above the noise floor, then the signal-todistortion ratio dominates and degrades by 3 dB for every 1 dB increase in signal
power. If we are concerned with the third-order distortion, the SFDR is calculated
from the geometric 2/3 relationship between the input intercept and the IMD.
It is important to note that the SFDR depends directly on the bandwidth Df. It has
no meaning without specifying bandwidth.