Electronic Noise

Measurements III Electronic Noise Helmuth Spieler
2002 ICFA Instrumentation School, Istanbul LBNL

1
III. Electronic Noise
1. Why? 2
2. What Determines Resolution? 4
3. Basic Noise Mechanisms 7
Thermal Noise in Resistors 8
Shot Noise 9
Derivation of thermal noise spectral density 10
Derivation of shot noise spectral density 12
Noiseless Resistances 13
Noise Characteristics 15
4. Noise in Amplifiers 17
Amplifier Noise Model 18
Noise Matching with Transformers 21
Correlated Noise 22
Noise Bandwidth vs. Signal Bandwidth 23
Equivalent Noise Charge 24
S/N with Capacitive Signal Sources 26
Charge-Sensitive Preamplifier Noise vs.
Detector Capacitance 30
Strip Detector Noise Model 33
Quantum Noise Limits in Amplifiers 34
2
Resolution and Electronic Noise
Resolution: the ability to distinguish signal levels
1. Why?
a) Recognize structure in amplitude spectra
Comparison between NaI(Tl) and Ge detectors
(J.Cl. Philippot, IEEE Trans. Nucl. Sci. NS-17/3 (1970) 446)
3
b) Improve sensitivity
Signal to background ratio improves with better resolution
(signal counts in fewer bins compete with fewer
background counts)
G.A. Armantrout et al., IEEE Trans. Nucl. Sci. NS-19/1 (1972) 107
4
2. What determines Resolution?
1. Signal Variance >> Baseline Variance
Electronic (baseline) noise not important
Examples:
High-gain proportional chambers
Scintillation Counters with High-Gain PMTs
e.g. 1 MeV -rays absorbed by NaI(Tl) crystal
Number of photoelectrons
N
pe
8
.
10
4
[MeV
-1
] E
QE 2.4
.
10
4
Variance typically

pe
= N
pe
1/2
160 and
pe
/ N
pe
5 - 8%
Signal at PMT anode (assume Gain= 10
4
)
Q
sig
= G
PMT
N
pe
2.4
.
10
8
el and

sig
= G
PMT

pe
1.2
.
10
7
el
whereas electronic noise easily < 10
4
el
SIGNAL BASELINE NOISE SIGNAL + NOISE +
BASELINE BASELINE BASELINE
5
2. Signal Variance << Baseline Variance
Electronic (baseline) noise critical for resolution
Examples
Gaseous ionization chambers (no internal gain)
Semiconductor detectors
e.g. in Si
Number of electron-hole pairs
N
ep
= E
dep
/ (3.6 eV)
Variance
(where F= Fano factor 0.1)
For 50 keV photons
ep
40 el
ep
/ N
ep
= 7.5
.
10
-4
obtainable noise levels are 10 to 1000 el.
SIGNAL BASELINE NOISE SIGNAL + NOISE +
BASELINE BASELINE BASELINE
ep ep
N F
6
Baseline fluctuations can have many origins ...
pickup of external interference
artifacts due to imperfect electronics
etc.,
but the (practical) fundamental limit is electronic noise.
7
3. Basic Noise Mechanisms
Consider n carriers of charge e moving with a velocity v through a
sample of length l. The induced current i at the ends of the sample is
The fluctuation of this current is given by the total differential
where the two terms are added in quadrature since they are
statistically uncorrelated.
Two mechanisms contribute to the total noise:
velocity fluctuations, e.g. thermal noise
number fluctuations, e.g. shot noise
excess or '1/ f ' noise
Thermal noise and shot noise are both white noise sources, i.e.
power per unit bandwidth is constant:
spectral density)
or
whereas for 1/ f noise
(typically = 0.5 2)
. const
df
dP
noise
2
2
.
n
noise
e const
df
dV

f
df
dP
noise
1
2 2
2 ne ev
di dv dn
l l
_ _
+

, ,

.
n e v
i
l

8
1. Thermal Noise in Resistors
The most common example of noise due to velocity fluctuations
is the thermal noise of resistors.
Spectral noise power density vs. frequency f
k = Boltzmann constant
T = absolute temperature
Since
R = DC resistance
the spectral noise voltage density
and the spectral noise current density
The total noise depends on the bandwidth of the system.
For example, the total noise voltage at the output of a voltage
amplifier with the frequency dependent gain Av

(f) is
Note: Since spectral noise components are non-correlated,
one must integrate over the noise power.
R I
R
V
P
2
2

2 2
0
2
( )
on n v
v e A f df
kT
df
dP
noise
4
kTR e
df
dV
n
noise
4
2
2

R
kT
i
df
dI
n
noise
4
2
2

9
2. Shot noise
A common example of noise due to number fluctuations is
shot noise, which occurs whenever carriers are injected into a
sample volume independently of one another.
Example: current flow in a semiconductor diode
(emission over a barrier)
Spectral noise current density:
q
e
= electronic charge
I = DC current
A more intuitive interpretation of this expression will be given
later.
Shot noise does not occur in ohmic conductors. Since the
number of available charges is not limited, the fields caused
by local fluctuations in the charge density draw in additional
carriers to equalize the total number.
3. 1/f Noise
The noise spectrum becomes non-uniform whenever the
fluctuations are not purely random in time, for example when
carriers are trapped and then released with a time constant .
With an infinite number of uniformly distributed time constants
the spectral power density assumes a pure 1/f distribution.
However, with as few as 3 time constants spread over one or
two decades, the spectrum is approximately 1/f, so this form of
noise is very common.
For a 1/f spectrum the total noise depends on the ratio of the
upper to lower cutoff frequencies, rather than the absolute
bandwidth.
I q i
e n
2
2

10
Spectral Density of Thermal Noise
Two approaches can be used to derive the spectral distribution of thermal noise.
1. The thermal velocity distribution of the charge carriers is used to calculate the time
dependence of the induced current, which is then transformed into the frequency
domain.
2. Application of Plancks theory of black body radiation.
The first approach clearly shows the underlying physics, whereas the second hides the
physics by applying a general result of statistical mechnics. However, the first requires
some advanced concepts that go well beyond the standard curriculum, so the black
body approach will be used.
In Plancks theory of black body radiation the energy per mode
/
1
h kT
h
E
e

and the spectral density of the radiated power

/
1
h kT
dP h
d e

i.e. this is the power that can be extracted in equilibrium. At low frequencies h kT =
,
1 1
dP h
kT
h d
kT

_
+

,
so at low frequencies the spectral density is independent of frequency and for a total
bandwidth B the noise power that can be transferred to an external device
n
P kTB .
To apply this result to the noise of a resistor, consider a resistor R whose thermal noise
gives rise to a noise voltage V
n
. To determine the power transferred to an external device
consider the circuit
R
R
V
I
L
n
n
11
The power dissipated in the load resistor R
L
2 2
2
2
( )
nL n L
n L
L L
V V R
I R
R R R

+
The maximum power transfer occurs when the load resistance equals the source
resistance R
T
= R, so
2
2
4
n
nL
V
V
Since the power transferred to R
L
is kTB
2 2
2
4
4
nL n
n
n
V V
kTB
R R
V
P kTB
R

and the spectral density of the noise power
4
n
dP
kT
d

12
Spectral Density of Shot Noise
If an excess electron is injected into a device, it forms a current pulse of duration . In a
thermionic diode is the transit time from cathode to anode (see IX.2), for example. In a
semiconductor diode is the recombination time (see IX-2). If these times are short with
respect to the periods of interest 1/ f = , the current pulse can be represented by a
pulse. The Fourier transform of a delta pulse yields a white spectrum, i.e. the amplitude
distribution in frequency is uniform
,
2
n p k
e
dI
q
df

Within an infinitesimally narrow frequency band the individual spectral components are
pure sinusoids, so their rms value
2
2
2
n e
n e
dI q
i q
df

If N electrons are emitted at the same average rate, but at different times, they will have
the same spectral distribution, but the coefficients will differ in phase. For example, for
two currents i
p
and i
q
with a relative phase the total rms current
( ) ( )
2 2 2
2 cos
i i
p q p q p q p q
i i i e i i e i i i i

+ + + +
For a random phase the third term averages to zero
2 2 2
,
p q
i i i +
so if N electrons are randomly emitted per unit time, the individual spectral components
simply add in quadrature
2 2
2
n e
i Nq
The average current
,
e
I Nq
so the spectral noise density
2
2
2
n
n e
dI
i q I
df

13
Noiseless Resistances
a) Dynamic Resistance
In many instances a resistance is formed by the slope of a devices
current-voltage characteristic, rather than by a static ensemble of
electrons agitated by thermal energy.
Example: forward-biased semiconductor diode
Diode current vs. voltage

/
0
( 1)
e
q V kT
I I e
The differential resistance

d
e
dV kT
r
dI q I
i.e. at a given current the diode presents a resistance, e.g. 26 at
I = 1 mA and T = 300 K.
Note that two diodes can have different charge carrier concentrations,
but will still exhibit the same dynamic resistance at a given current, so
the dynamic resistance is not uniquely determined by the number of
carriers, as in a resistor.
There is no thermal noise associated with this dynamic resistance,
although the current flow carries shot noise.
14
b) Radiation Resistance of an Antenna
Consider a receiving antenna with the normalized power pattern
( , )
n
P pointing at a brightness distribution ( , ) B in the sky. The
power per unit bandwidth received by the antenna

( , ) ( , )
2
e
n
A
w B P d
where
e
A is the effective aperture, i.e. the capture area of the
antenna. For a given field strength E, the captured power
e
W EA .
If the brightness distribution is from a black body radiator and were
measuring in the Rayleigh-Jeans regime,

2
2
( , )
kT
B
and the power received by the antenna

2
.
e A
kT
w A
A
is the beam solid angle of the antenna (measured in rad
2
), i.e. the
angle through which all the power would flow if the antenna pattern
were uniform over its beamwidth.
Since
2
e A
A (see antenna textbooks), the received power
w kT
The received power is independent of the radiation resistance, as
would be expected for thermal noise.
However, it is not determined by the temperature of the antenna, but
by the temperature of the sky the antenna pattern is subtending.
For example, for a region dominated by the CMB, the measured
power corresponds to a resistor at a temperature of ~3K, although
the antenna may be at 300K.
15
Noise characteristics
Both thermal and shot noise are purely random.
amplitude distribution is gaussian
noise modulates baseline
baseline fluctuations superimposed on signal
output signal has gaussian distribution
16
Correlated Noise
Generally, noise power is additive.
, 1 1
...
nt ot n n
P P P + +
However, in a coherent system (i.e. a system that preserves phase),
the power often results from the sum of voltages or currents, which is
sensitive to relative phase.
For two correlated noise sources
1
N and
2
N the total noise
2 2
1 2 1 2
2 N N N CN N + +
where the correlation coefficient C can range from 1 (anti-correlated,
i.e. identical, but 180 out of phase) to +1 (fully correlated).
For uncorrelated noise components 0 C and then individual current
or voltage noise contributions add in quadrature, e.g.
V V
n tot ni
i
,

2
17
4. Noise in Amplifiers
Consider a chain of two amplifiers (or amplifying devices), with gains
A1 and A1, and input noise levels N1 and N2 .
A signal S is applied to the first amplifier, so the input signal-to-noise
ratio is
1
/ S N .
At the output of the first amplifier the signal is A1S and the noise A1N.
Both are amplified by the second amplifier, but in addition the second
amplifier contributes its noise, so the signal-to-noise ratio at the
output of the second amplifier
( )
( ) ( )
_

, + _
+

,
_
_

, _ ,
+

,
2 2
2
1 2
2 2 2
1 1 2 2 2 2
2
1
1
2
2
2
1
2
1 1
1
1
S A A
S S
N
N A A N A
N
N
A
S S
N N
N
A N
The overall sign-to-noise ratio is reduced, but the noise contribution
from the second-stage can be negligible, provided the gain of the first
stage is sufficiently high.
In a well-designed system the noise is dominated by
the first gain stage.
A
N N
A
1
1 2
2
18
Amplifier Noise Model
The noise properties of any amplifier can be described fully in terms
of a
voltage noise source
and
current noise source.
at the amplifier input. Typical magnitudes are n V/ Hz and
p A/ Hz .
Here the magnitude of the noise sources is characterized by the
spectral density
The noise sources do not have to physically present at the input.
Noise also originates within the amplifier. Assume that at the output
the combined contribution of all internal noise sources has the
spectral density
no
e . If the amplifier has a voltage gain
V
A , this is
equivalent to a voltage noise source at the input /
n no V
e e A .
It is convenient to express the input noise in terms of spectral density,
so that the effect of amplifier bandwidth can be assessed separately.
e
i
n
n
19
Assume that a sensor with resistance
S
R is connected to an amplifier
with voltage gain
V
A and an infinite input resistance, so no current
flows into the amplifier.
The input noise current
n
i flows through the source resistance
S
R to
yield a noise voltage
n S
i R , which adds to the thermal noise of the
source resistance and the noise voltage of the amplifier.
All terms add in quadrature, since they are not correlated.
The total noise voltage at the input of the amplifier
( )
+ +
2
2 2
4
ni S n n S
e k T R e i R
and at the output of the amplifier
( )
1
+ +
]
2
2 2 2 2
( ) 4
no V ni V S n n S
e A e A kTR e i R
The signal-to-noise ratio at the amplifier output
( )
_

1
,
+ +
]
2
2 2
2
2 2
4
V S
V S n n S
S A V
N
A kTR e i R
is independent of the amplifier gain and equal to the input S/N, as
both the input noise and the signal are amplified by the same amount.
e R 4kTR
A i V e
e
n
S S
V
n
S
ni
no
20
In the preceding example the amplifier had an infinite input
resistance, so no current flowed into the amplifier. Is the signal-to-
noise ratio affected by a finite input resistance?
The signal at the input of the amplifier
+
i
Si S
S i
R
V V
R R
The noise voltage at the input of the amplifier
( )
_ _
+ +

+ +
, ,
2 2
2 2 2
4
i i S
ni S n n
i S i S
R R R
e kTR e i
R R R R
where the bracket in the
2
n
i represents the parallel combination of
i
R
and
S
R . The signal-to-noise ratio at the output of the amplifier
( )
( )
_

+
_
,

,
_ _
+ +

+ +
, ,
_

+ + ,
2
2
2
2 2
2 2 2 2
2 2
2
2
2 2 2
4
,
4
i
S
S i V S i
V ni
i i S
S n n
i S i S
S
S n n S
R
V
R R
S A V
N A e
R R R
kTR e i
R R R R
S V
N
kTR e i R
the same as for an infinite input resistance.
This result also hold for a complex input impedance, i.e. a
combination of resistive and capacitive or inductive components.
S /N independent of amplifier input impedance.
e R 4kTR
R i V e
e
n
S S
i
n S ni
no
21
Noise matching with transformers
The sensor is coupled to the amplifier through a transformer with the
turns ratio /
S P
N N N .
Assume unity coupling in the transformer. Then the sensor voltage
appearing at the secondary
SS S
V NV
The thermal noise of the sensor at the secondary
2 2
4
nS S S
e N kTR
Because the transformer also converts impedances, the source
resistance appears at the secondary as
2
SS S
R N R
Thus, the signal is increased, but so is the noise contribution due to
the input noise current.
+ +
2 2 2 2 4 2
4
ni S n S n
e kTR N e R N i
and the signal-to-noise ratio
_

+ +
,
+ +
2
2 2 2
2 2 2 2 4 2
2 2 2
2
,
4
4
S S
n S n S n
S S n
S V N V
e N kTR N e R N i
kTR N R i
N
which attains a maximum for
2 n
S
n
e
R N
i
e R
N N
4kTR
A
i V e
e
n
S
P S
S
V
n
S
ni
no
22
Correlated Noise
The noise sources can be correlated, for example
+ +
2 2 2
1 2 1 2
2
n n n n n
e e e Ce e
where C is the correlation coefficient. C can range from 1 to +1
(anti-correlated to fully correlated).
If 0 C the noise components are uncorrelated and they simply add
in quadrature.
Thus, in the above example, if the input noise voltage and current are
correlated, the input noise voltage
+ + +
2 2 2
4 2
ni S n n n n S
e kTR e i Ce i R
The total noise at the output is obtained by integrating over the
spectral noise power
2
( ) ( )
n no
P f e f .
The frequency distribution of the noise is determined both by the
spectral distribution of the input noise voltage and current and by the
frequency response of the amplifier.

2
2 2 2
0 0
( ) ( )
no no ni V
v e f df e f A df
The amplifier gain factor is shown as magnitude squared, as in
general the amplifier has a frequency-dependent gain and phase, so
it is a complex number.
23
Noise Bandwidth vs. Signal Bandwidth
Consider an amplifier with the frequency response A(f). This can be
rewritten
where A
0
is the maximum gain and G(f ) describes the frequency
response.
For example, in the simple amplifier described above
and using the above convention
If a white noise source with spectral density e
ni
is present at the
input, the total noise voltage at the output is
f
n
is the noise bandwidth.
Note that, in general, the noise bandwidth and the signal bandwidth
are not the same. If the upper cutoff frequency is determined by a
single RC time constant, as in the simple amplifier, the signal
bandwidth
and the noise bandwidth
) ( ) (
0
f G A f A
_
+

+
,
1
1 1

1
V m o m L
L L o
A g i C g R
R i R C

+
0
1
and ( )
1 (2 )
m L
L o
A g R G f
i f R C
2
2 2
0 0 0
0 0

( ) ( )
no ni ni ni n
V e A G f df e A G f df e A f
RC
f f
u s
2
1

u n
f
RC
f
2 4
1

24
Equivalent Noise Charge
A detector readout systems that measure signal charge can be
characterized in terms of Equivalent Noise Charge,
Equivalent Noise Charge (ENC) =
= the signal charge that yields a signal-to-noise ratio of one.
For a given detector material, the signal charge can be translated into
absorber energy, so the noise can be express in terms of energy, i.e.
eV or keV.
For an ionization energy
i

n i
E ENC
The relationship between ENC and basic amplifier noise parameters
will be derived in the next chapter.
25
Continuous Signals vs. Individual Pulses
Consider a semiconductor detector detecting visible light.
At low intensities, where the mean time between successive photons
is much longer than the collection time, the detector signal consists of
individual pulses.
As the light intensity increases, so does the photon rate. At some
point the signals from individual photons overlap and the detector
output appears as a continuous current.
The average current of a sequence of pulses ( ) i t of duration T
occur r ing at a rate R
( )
av
i R i t dt
If each individual pulse has a DC component, the DC component of
the pulse train will grow as the rate increases.
Each individual pulse has a characteristic Fourier spectrum. Since
this is a linear superposition process, the sum of all pulses has the
same frequency spectrum as an individual pulse.
signal-to-noise can be analyzed using either pulses or
continuous signals.
If a filter is chosen to optimize the signal-to-noise ratio for a
single pulse, it will also optimize S /N at high rates.
However, the need to resolve individual pulses or measure their
amplitude accurately adds an additional constraint that modifies
the choice of filter at high rates.
26
S/N with Capacitive Signal Sources
Equivalent Circuit

charges moving in detector capacitance
detector induce change discharges into amplifier
of charge on detector
electrodes
Assume an amplifier with constant noise. Then signal-to-noise ratio
(and the equivalent noise charge) depend on the signal magnitude.
R
AMPLIFIER
V
in
DETECTOR
C
DETECTOR
C R
AMPLIFIER
i v
i
s in det
in
27
Pulse shape registered by amplifier depends on the input time
constant RC
det
.
Assume a rectangular detector current pulse of duration T and
magnitude I
s
.
Equivalent circuit
Input current to amplifier
At short time constants RC << T the amplifier pulse approximately
follows the detector current pulse.
RC= 0.01 T RC= 0.1 T
( )
( )
RC t RC T
s in
RC t
s in
e e I t i t T
e I t i T t
/ /
/
1 ) ( :
1 ) ( : 0

<
0
0.2
0.4
0.6
0.8
1
1.2
0 1 2 3 4 5 6 7 8 9 10
t/T
S
I
G
N
A
L
0
0.2
0.4
0.6
0.8
1
1.2
0 1 2 3 4 5 6 7 8 9 10
t/T
S
I
G
N
A
L
DETECTOR
C R
AMPLIFIER
i v
i
s in
in
28
As the input time constant RC increases, the amplifier signal
becomes longer and the peak amplitude decreases, although the
integral, i.e. the signal charge, remains the same.
RC = T RC = 10 T
RC = 100 T RC = 10
3
T
At long time constants the detector signal current is integrated on
the detector capacitance and the resulting voltage sensed by the
amplifier
Then the peak amplifier signal is inversely proportional to the total
capacitance at the input, i.e. the sum of
detector capacitance,
input capacitance of the amplifier, and
stray capacitances.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0 1 2 3 4 5 6 7 8 9 10
t/T
S
I
G
N
A
L
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
0 1 2 3 4 5 6 7 8 9 10
t/T
S
I
G
N
A
L
0
0.002
0.004
0.006
0.008
0.01
0.012
0 1 2 3 4 5 6 7 8 9 10
t/T
S
I
G
N
A
L
0
0.0002
0.0004
0.0006
0.0008
0.001
0.0012
0 1 2 3 4 5 6 7 8 9 10
t/T
S
I
G
N
A
L
C
dt i
C
Q
V
s
in

det
29
Maximum signal vs. capacitance
At small time constants the amplifier signal approximates the detector
current pulse and is independent of capacitance.
At large input time constants (RC/T > 5) the maximum signal falls
linearly with capacitance.
For input time constants large compared to the detector
pulseduration the signal-to-noise ratio decreases with
detector capacitance.
Caution when extrapolating to smaller capacitances:
If S/N = 1 at RC/T = 100, decreasing the capacitance to 1/10 of
its original value (RC/T = 10), increases S/N to 10.
However, if initially RC/T = 1, the same 10-fold reduction in
capacitance (to RC/T = 0.1) only yields S/N = 1.6.
0
0.2
0.4
0.6
0.8
1
1.2
0 1 2 3 4 5 6 7 8 9 10
RC/T
M
A
X
I
M
U
M

S
I
G
N
A
L
0.01
0.1
1
0.1 1 10 100
RC/T
M
A
X
I
M
U
M

S
I
G
N
A
L
30
Charge-Sensitive Preamplifier
Noise vs. Detector Capacitance
In a voltage-sensitive preamplifier
noise voltage at the output is essentially independent of
detector capacitance,
i.e. the equivalent input noise voltage v
ni
= v
no
/A
v
.
input signal decreases with increasing input capacitance, so
signal-to-noise ratio depends on detector capacitance.
In a charge-sensitive preamplifier, the signal at the amplifier output is
independent of detector capacitance (if C
i
>>C
det
).
What is the noise behavior?
Noise appearing at the output of the preamplifier is fed back
to the input, decreasing the output noise from the open-loop
value v
no
= v
ni
A
v0
.
The magnitude of the feedback depends on the shunt
impedance at the input, i.e. the detector capacitance.
Note, that although specified as an equivalent input noise, the
dominant noise sources are typically internal to the amplifier.
Only in a fed-back configuration is some of this noise actually
present at the input. In other words, the primary noise signal is
not a physical charge (or voltage) at the amplifier input, to
which the loop responds in the same manner as to a detector
signal.
S/N at the amplifier output depends on feedback.
31
Noise in charge-sensitive preamplifiers
Start with an output noise voltage v
no
, which is fed back to the
input through the capacitive voltage divider C
f
C
d
.
1 1
1
1

f D
D
C C
f D
n o ni ni
C
D
D
n o ni
f
X X
C C
v v v
X
C
C
v v
C

+
+

_
+

,
Equivalent input noise charge
( )

n o
n i no f
Q
n i ni D f
v
Q v C
A
Q v C C

+
32
Signal-to-noise ratio
Same result as for voltage-sensitive amplifier, but here
the signal is constant and
the noise grows with increasing C.
As shown previously, the pulse rise time at the amplifier output also
increases with total capacitive input load C, because of reduced
feedback.
In contrast, the rise time of a voltage sensitive amplifier is not affected
by the input capacitance, although the equivalent noise charge
increases with C just as for the charge-sensitive amplifier.
Conclusion
In general
optimum S/N is independent of whether the voltage,
current, or charge signal is sensed.
S/N cannot be i im mp pr ro ov ve ed d by feedback.
Practical considerations, i.e. type of detector, amplifier technology,
can favor one configuration over the other.
1

( )
s s s
n i n i D f n i
Q Q Q
Q v C C C v

+
33
Strip Detector Model for Noise Simulations
Noise coupled from neighbor channels.
Analyze signal and noise in center channel.
Includes: Noise contributions from neighbor channels
Signal transfer to neighbor channels
Noise from distributed strip resistance
Full SPICE model of preamplifier
Measured Noise of Module
Simulation Results:
p-strips on n-bulk
BJT input transistor
(see Part V)
1460 el (150 A)
1230 el (300 A)
Noise can be predicted

with good accuracy.
50 100 150 200 250 300
Current in Input Transistor [A]
1200
1300
1400
1500
1600
N
o
i
s
e

[
r
m
s

e
l
]
34
Quantum Noise Limits in Amplifiers
What is the lower limit to electronic noise?
Can it be eliminated altogether, for example by using
superconductors and eliminating devices that carry shot noise?
Starting point is the uncertainty relationship

h
2
E t
Consider a narrow frequency band at frequency . The energy
uncertainty can be given in terms of the uncertainty in the number of
signal quanta
h E n
and the time uncertainty in terms of phase
, t
so that

1
2
n
We assume that the distributions in number and phase are Gaussian,
so that the equality holds.
Assume a noiseless amplifier with gain G, so that
1
n quanta at the
input yield
2 1
n Gn
quanta at the output.
Furthermore, the phase at the output
2
is shifted by a constant
relative to the input.
35
Then the output must also obey the relationship

2 2
1
2
n
However, since
2 1
n G n and
2 1

1 1
1
,
2
n
G
which is smaller than allowed by the uncertainty principle.
This contradiction can only be avoided by assuming that the amplifier
introduces noise per unit bandwidth of
h ( 1) ,
no
dP
G
d
which, referred to the input, is
_

,
h
1
1
ni
dP
d G
If the noise from the following gain stages is to be small, the gain of
the first stage must be large, and then the minimum noise of the
amplifier
h
ni
dP
d
At 2 mm wavelength the minimum noise corresponds to about 7K.
This minimum noise limit applies to phase-coherent systems. In
systems where the phase information is lost, e.g. bolometers, this
limit does not apply.
For a detailed discussion see
C.M. Caves, Phys. Rev. D 26 (1982) 1817-1839

Electronic Noise

Uploaded by

Copyright:

Available Formats

Electronic Noise

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Electronic Noise

Uploaded by

Copyright:

Available Formats

Measurements III Electronic Noise Helmuth Spieler

2002 ICFA Instrumentation School, Istanbul LBNL

Measurements III Electronic Noise Helmuth Spieler

Measurements III Electronic Noise Helmuth Spieler

and the spectral density of the radiated power

Noise can be predicted

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.