Download Low-Frequency Noise Measurements with the E4727A

Keysight Technologies
Low-Frequency Noise Measurements with the
E4727A and Their Applications
Application Note
Introduction
Introduction to Noise and Noise Modeling
Any undesirable turbulence that corrupts an information-bearing signal is generally referred to as noise.
Electrical noise is inherent in every circuit, ranging from current flowing through a resistor or transistor, to
leakage current through a tantalum capacitor. To minimize its effects, it becomes necessary to measure and
quantify the noise of the constituent parts, and then connect the constituent noise contributions to overall
circuit performance. In the case of a voltage-controlled oscillator, transistor noise results in small frequency
fluctuations measured as phase noise. In the case of a low-noise amplifier or an operational amplifier circuit,
resistor noise in the feedback can greatly increase the resultant voltage noise at the output. A typical
amplifier will not only amplify the signal and noise at the input (which is desirable), but also contribute its
own noise (which is undesirable).
The E4727A Advanced Low-Frequency Noise Analyzer (A-LFNA) enables a closer, deeper look at noise. A
device designer may extract noise models using powerful device modeling software like the Model Builder
Program (MBP) and the Integrated Circuit Characterization and Analysis Program (IC-CAP). The models may
be then passed along to circuit designers, who may then push the envelope in low noise circuitry.
10
Sid W=50 L=1 T=25 Vbs=0 Vds=20 Vgs=5
-12
R
R10
R=400 Ohm
10
-14
R
R12
R=700 Ohm
model
R
R7
R=100 Ohm
-16
L
L2
L=0.5 nH
R=
C
C4
C=1 uF
R
R3
R=50 Ohm
BJT_NPN
BJT5
Model=BJTM1
Area=1
Region=
Trise=
C
C3
C=1.0 uF
Sid (A
-18
RF_port
Num=1
10
10
-20
BJT_NPN
BJT3
Model=BJTM2
Area=1
Region=
Trise=
L
L1
L=0.5 nH
R=
Isc=1.16160000E-12
C4=
Nc=2.00E+00
Cbo=
Gbo=
Vbo=
Rb=6.12458791E+01
Rbm=1.91785714E+01
Re=3.57142857E+00
Rc=6.96232096E+01
Rcv=
Rcm=
Dope=
Cex=
Cco=
Imax=1.0 A
Imelt=
Cje=9.67680017E-14
Vje=8.49999987E-01
Mje=3.99999994E-01
Cjc=6.05000011E-14
Vjc=7.49999989E-01
Mjc=4.99999993E-01
Xcjc=2.52561980E-01
Cjs=8.68000015E-14
Vjs=6.99999990E-01
Mjs=4.99999993E-01
Fc=7.99999988E-01
Xtf=3.35472224E+00
Tf=8.91292803E-12
Itf=1.08500002E-02
Ptf=1.80E+01
Tr=1.60000005E-09
Kf=0.00E+00
Af=1.00E+00
Ab=1.00
Fb=1.00
Rbnoi=
Iss=0 A
Ns=1.00
Nk=0.50
Ffe=1.00
RbModel=MDS
Approxqb=yes
Tnom=25.
BJT_NPN
BJT2
Model=BJTM3
Area=1
Region=
Trise=
Trise=
Eg=1.11
Xtb=2.20E+00
Xti=8.00E+00
Itf=4.34000010E-02 Trise=
Ptf=1.80E+01
Eg=1.11
Tr=1.60000005E-09 Xtb=2.20E+00
Kf=0.00E+00
Xti=8.00E+00
Af=1.00E+00
Ab=1.00
Fb=1.00
Rbnoi=
Iss=0 A
Ns=1.00
Nk=0.50
Ffe=1.00
RbModel=MDS
Approxqb=yes
Tnom=25.
R
R13
R=500 Ohm
BJT_NPN
BJT4
Model=BJTM2
Area=1
Region=
Trise=
C
C2
C=1 uF
BJT_Model
BJTM3
NPN=yes
PNP=no
Is=2.04315924E-16
Bf=1.66178934E+02
Nf=1.03E+00
Vaf=2.50E+01
Ikf=1.16250003E-02
Ise=6.09123888E-13
C2=
Ne=2.00E+00
Br=5.12295082E+00
Nr=1.00E+00
Ikr=6.00000013E-02
Ke=
Kc=
Isc=2.31840001E-12
C4=
Nc=2.00E+00
Cbo=
Gbo=
Vbo=
Rb=1.17166379E+02
Rbm=2.70221484E+01
Re=1.66666667E+00
Rc=3.58072512E+01
Rcv=
Rcm=
Dope=
Cex=
Cco=
Imax=1.0 A
Imelt=
Cje=1.56160003E-13
Vje=8.49999987E-01
Mje=3.99999994E-01
Cjc=1.20750002E-13
Vjc=7.49999989E-01
Mjc=4.99999993E-01
Xcjc=1.95445132E-01
Cjs=1.18900002E-13
Vjs=6.99999990E-01
Mjs=4.99999993E-01
Fc=7.99999988E-01
Xtf=3.16117215E+00
Tf=7.97477712E-12
L
L3
L=0.5 nH
R=
IF_port
Num=3
Itf=2.32500005E-02 Trise=
Ptf=1.80E+01
Eg=1.11
Tr=1.60000005E-09 Xtb=2.20E+00
Xti=8.00E+00
Kf=0.00E+00
Af=1.00E+00
Ab=1.00
Fb=1.00
Rbnoi=
Iss=0 A
Ns=1.00
Nk=0.50
Ffe=1.00
RbModel=MDS
Approxqb=yes
Tnom=25.
BJT_NPN
BJT12
Model=BJTM3
Area=1
Region=
Trise=
BJT_NPN
BJT1
Model=BJTM3
Area=1
Region=
Trise=
R
R4
R=170 Ohm
BJT_NPN
BJT14
Model=BJTM5
Area=1
Region=
Trise=
R
R5
R=170 Ohm
R
R17
R=45 Ohm
BJT_NPN
BJT13
Model=BJTM3
Area=1
Region=
Trise=
R
R15
R=1000 Ohm
R
R14
R=200 Ohm
BJT_Model
BJTM5
NPN=yes
PNP=no
Is=4.08631847E-16
Bf=1.66178934E+02
Nf=1.03E+00
Vaf=2.50E+01
Ikf=2.32500005E-02
Ise=1.21824776E-12
C2=
Ne=2.00E+00
Br=5.12295082E+00
Nr=1.00E+00
Ikr=1.199999998E-01
Ke=
Kc=
Isc=4.08480002E-12
C4=
Nc=2.00E+00
Cbo=
Gbo=
Vbo=
Rb=5.91784277E+01
Rbm=1.41063123E+01
Re=8.33333321E-01
Rc=2.20144369E+01
Rcv=
Rcm=
Dope=
Cex=
Cco=
Imax=1.0 A
Imelt=
Cje=3.12320005E-13
Vje=8.49999987E-01
Mje=3.99999994E-01
Cjc=2.12750004E-13
Vjc=7.49999989E-01
Mjc=4.99999993E-01
Xcjc=2.21856636E-01
Cjs=1.65300003E-13
Vjs=6.99999990E-01
Mjs=4.99999993E-01
Fc=7.99999988E-01
Xtf=3.16117215E+00
Tf=7.97477712E-12
Itf=4.65000010E-02
Ptf=1.80E+01
Tr=1.60000005E-09
Kf=0.00E+00
Af=1.00E+00
Ab=1.00
Fb=1.00
Rbnoi=
Iss=0 A
Ns=1.00
Nk=0.50
Ffe=1.00
RbModel=MDS
Approxqb=yes
Tnom=25.
Trise=
Eg=1.11
Xtb=2.20E+00
Xti=8.00E+00
-22
1
100
10k
Frequency (Hz)
Device characterization
C
C1
C=1 uF
Imax=1.0 A
Imelt=
Cje=3.87072007E-13
Vje=8.49999987E-01
Mje=3.99999994E-01
Cjc=1.92500003E-13
Vjc=7.49999989E-01
Mjc=4.99999993E-01
Xcjc=3.17506489E-01
Cjs=1.54000003E-13
Vjs=6.99999990E-01
Mjs=4.99999993E-01
Fc=7.99999988E-01
Xtf=3.35472224E+00
Tf=8.91292803E-12
C
C5
C=1 uF
R
R2
R=50 Ohm
R
R6
R=20 Ohm
BJT_Model
BJTM1
NPN=yes
PNP=no
Is=1.07411147E-16
Bf=1.30035647E+02
Nf=1.03E+00
Vaf=2.50E+01
Ikf=5.42499974E-03
Ise=3.78086717E-13
C2=
Ne=2.00E+00
Br=5.12295082E+00
Nr=1.00E+00
Ikr=2.80000006E-02
Ke=
Kc=
Isc=3.69600001E-12
C4=
Nc=2.00E+00
Cbo=
Gbo=
Vbo=
Rb=1.61795253E+01
Rbm=5.66269841E+00
Re=8.92857130E-01
Rc=2.27441087E+01
Rcv=
Rcm=
Dope=
Cex=
Cco=
BJT_NPN
BJT8
Model=BJTM1
Area=1
Region=
Trise=
BJT_NPN
BJT11
Model=BJTM3
Area=1
Region=
Trise=
R
R1
R=50 Ohm
10
BJT_Model
BJTM4
NPN=yes
PNP=no
Is=4.29644587E-16
Bf=1.30035647E+02
Nf=1.03E+00
Vaf=2.50E+01
Ikf=2.17000005E-02
Ise=1.51234685E-12
C2=
Ne=2.00E+00
Br=5.12295082E+00
Nr=1.00E+00
Ikr=1.11999998E-01
Ke=
Kc=
R
R9
R=100 Ohm
BJT_NPN
BJT6
Model=BJTM1
Area=1
Region=
Trise=
2
/Hz)
LO_PORT
Num=4
10
BJT_NPN
BJT7
Model=BJTM1
Area=1
Region=
Trise=
Bias_port
Num=2
R
R11
R=800 Ohm
BJT_NPN
BJT9
Model=BJTM4
Area=1
Region=
Trise=
BJT_NPN
BJT10
Model=BJTM3
Area=1
Region=
Trise=
R
R8
R=400 Ohm
data
Modelling
1M
BJT_Model
BJTM2
NPN=yes
PNP=no
Is=2.1482293E-16
Bf=1.30035647E+02
Nf=1.03E+00
Vaf=2.50E+01
Ikf=1.08500002E-02
Ise=7.56173434E-13
C2=
Ne=2.00E+00
Br=5.12295082E+00
Nr=1.00E+00
Ikr=5.600000013E-02
Ke=
Kc=
Isc=2.00640001E-12
C4=
Nc=2.00E+00
Cbo=
Gbo=
Vbo=
Rb=3.16646062E+01
Rbm=1.06309524E+01
Re=1.78571429E+00
Rc=3.75704756E+01
Rcv=
Rcm=
Dope=
Cex=
Cco=
Imax=1.0 A
Imelt=
Cje=1.93536003E-13
Vje=8.49999987E-01
Mje=3.99999994E-01
Cjc=1.04500002E-13
Vjc=7.49999989E-01
Mjc=4.99999993E-01
Xcjc=2.92440187E-01
Cjs=1.09200002E-13
Vjs=6.99999990E-01
Mjs=4.99999993E-01
Fc=7.99999988E-01
Xtf=3.35472224E+00
Tf=8.91292803E-12
Itf=2.17000005E-02 Tnom=25.0
Trise=
Ptf=1.80E+01
Tr=1.60000005E-09 Eg=1.11
Xtb=2.20E+00
Kf=0.00E+00
Xti=8.00E+00
Af=1.00E+00
Kb=
Ab=1.00
Fb=1.00
Rbnoi=
Iss=0 A
Ns=1.00
Nk=0.50
Ffe=1.00
RbModel=MDS
Approxqb=yes
Circuit design in ADS
Figure 1. High-level flow of information from device characterization to modeling to low-noise circuit design.
The E4727A addresses the first function shown in Figure 1: noise data measurement. In this application note,
we will cover the following topics:
–– Basics of noise spectral density
–– Noise measurement applications
–– Practical considerations in noise measurements
–– How the E4727A addresses noise test challenges
03 | Keysight | Low-Frequency Noise Measurements with the E4727A and Their Applications - Application Note
Understanding Noise Power Spectral Density
Noise may be quantified in terms of its aggregate effect in the time domain, as in peak to peak
fluctuations about some mean voltage, or in terms of its standard deviation about some mean
voltage (RMS). Underlying these time domain manifestations is the frequency content of the
voltage fluctuations, which is expressed as voltage (“en”) or current (“in”) spectral noise density
with units of either V/√Hz or A/√Hz. In order to discuss RMS noise power, it is imperative to
describe the bandwidth over which one would integrate the spectral density.
There are two noise spectral density shapes that capture the vast majority of electronic device
noise modeling situations.
1. White noise, where the spectral content is flat versus frequency. This is always present.
in = in_white
en = en_white
2. 1/f or flicker noise (also known as “pink noise”), where the spectral density is inversely
proportional to the frequency.
en (f)
2
 v2

 Hz
 Kv 2
=
 f
in(f)
2
 A2

 Hz
 Ki 2
=
 f
The two scenarios are pictured in Figure 2.
White noise
4
2
Amplitude
Amplitude
2
0
-2
-4
0
1000
2000
3000
Samples
4000
-4
5000
Pink noise
0
200
400
Counts
600
Pink noise histogram
4
2
2
Amplitude
Amplitude
0
-2
4
0
-2
-4
White noise histogram
4
0
-2
0
1000
2000
3000
4000
5000
Samples
Figure 2. Example time domain data of white noise and 1/f noise.
-4
0
200
400
Counts
600
04 | Keysight | Low-Frequency Noise Measurements with the E4727A and Their Applications - Application Note
The histograms shown to the right in Figure 2 show a more peaked distribution for the white
noise as compared to pink noise. Another type of noise that results in a bimodal distribution is
random telegraph noise (RTN), also known as “burst” or “popcorn” noise. With RTN, sudden
step-like transitions between 2 or more levels are observed, as shown in Figure 3.
NMOS drain current vs time
100.15
Drain current histogram
Id ( A)
100.1
100.05
100
99.95
0
20
40
60
80
100
120
140
0
500
1000
1500
2000
Counts
Time ( sec)
Figure 3. Example NMOS drain current in the time domain showing random telegraph noise (RTN).
Figure 4 presents a typical measurement scenario showing both 1/f and white noise.
Measured and modeled noise
10 -5
Noise density (V/sqrt(Hz))
Model
Measured data
10 -6
10 -7
10 -8
10 -9
100m
1
10
100
1k
10k
100k
Frequency (Hz)
Figure 4. Measured noise power density showing 1/f noise and white noise.
1M
10M
100M
05 | Keysight | Low-Frequency Noise Measurements with the E4727A and Their Applications - Application Note
The two effects are presented in the following equation, which was fitted to the above data.
Noise (f )


20 ⋅ log Noise_Floor ⋅ 1 +
Corner_Freq 
f


The extracted noise floor is 9.8 nV/√Hz, and the corner frequency is 45.3 kHz.
There is a strong similarity between the equation above and the way that SPICE models noise
current in a diode2.
Id (f )
2
KF⋅Idc
f
EF
AF
+ 2⋅q ⋅Idc
The parameters AF, KF and EF typically need to be fitted to the measured data. In particular,
EF may be used to set the slope with respect to frequency. If we initially assume EF = AF = 1,
then the observations on current noise floor and 1/f frequency may be combined with the above
equation as follows2.
Idc
Noise_Floor
2
KF
2 ⋅ q ⋅ Corner_Freq
2⋅ q
Building on this, the modeling engineer will model current noise as a function of DC bias and
device size. The noise data from the E4727A is easily ported to noise modeling software like
MBP or IC-CAP, thus streamlining the model process.
06 | Keysight | Low-Frequency Noise Measurements with the E4727A and Their Applications - Application Note
Noise Measurement Applications
Across the electronics industry, the need to characterize noise is ubiquitous. Wherever
there is a sensor, detector or receiver accompanied by signal processing circuitry, numerous
impairments are possible — both systematic and stochastic. While the systematic errors can
often be calibrated out, the noise floor sets an absolute level to the sensitivity of the detector
circuit. Noise considerations are also important on the signal-creation side. For example,
an ultralow noise voltage reference circuit provides a temperature stable bias voltage that
may be applied to a digital to analog converter (DAC), the active element inside a low noise
oscillator, a low noise amplifier (LNA), or even industrial process control circuitry that provides
programmable voltage and current outputs. By influencing voltage thresholds or timing jitter,
voltage noise can easily translate to amplitude noise or phase noise in an RF communications
signal, leading to degraded modulation quality. The following provides more details on two
applications where high voltages or very low frequency noise data are required.
Application: GaN Devices for Radar Transmitters
To resolve the locations and velocities of unfriendly targets, high-power transmitters are
required in radar systems and weather tracking systems. A typical radar transmitter - receiver
block diagram is shown in Figure 5. The high-power amplifier (HPA) has been implemented
using solid-state laterally diffused MOS (LDMOS) devices.
Duplexer
High power amplifier
Lowpass filter
B
Waveform generator
+
−
A
High-power transmit section
(~100 W to ~MW)
Receiver
To signal processor
ADC
LNA
I
RF
Q
To signal processor
ADC
Figure 5. Simplified radar transmitter/receiver system block diagram.6
Bandpass filter
Antenna
07 | Keysight | Low-Frequency Noise Measurements with the E4727A and Their Applications - Application Note
GaN devices have recently attracted significant interest for high power RF amplifiers due
to their high electron mobility, high breakdown voltages and excellent reliability at elevated
temperatures.3 However, the high-power amplifier must be both rugged and low noise.
Broadband noise and 1/f noise of the final stage device modulates onto the carrier, causing
an overall degradation to the received signal-to-noise ratio. Pulse-Doppler radar enables the
detection of both location and velocity based on the timings and frequency shifts of the echoes
received. The frequency spectrum near the carrier is estimated from these echoes by FFT
analysis. For a small, slowly moving target, the Doppler shift will be small, making it difficult to
separate the target from the close-to-carrier noise of the radar. As shown in Figure 6, target 2
is a low amplitude target with a small frequency shift that can easily be overshadowed by the
phase noise sidebands of the transmitter signal or receiver LO.
Amplitude
Carrier power
Noise floor
-PRF/2
Target 1
Main lobe
Target 2
Doppler frequency or f d
Figure 6. Doppler shift exhibited by 3 moving targets. With the noise floor and 1/f noise on the main lobe,
radar return clutter can mask target 2.
The E4727A leads the industry in its ability to handle output voltages of up to 200 V, making it
well suited to the characterization of GaN, SiC and other high-voltage devices.
Target 3
PRF/2
08 | Keysight | Low-Frequency Noise Measurements with the E4727A and Their Applications - Application Note
Application: Optical Image Stabilization
Optical Image Stabilization (OIS) technology in handheld digital cameras is effective in
mitigating the effects of involuntary movements of the hand. This movement inevitably results
in unwanted blurring. Major developments in the market have made this an increasingly
important technology.4 Since 2010, digital still cameras have been steadily replaced by
smart phones that feature large LCD displays and built-in digital cameras. The cameras are
manufactured with increasingly smaller size, lighter weight and higher resolution. Lighter
cameras result in greater blurring, an effect that is further exacerbated by the use of selfie
sticks and users taking pictures with one arm outstretched. With image stabilization, the
user may feel free to increase the exposure time in low light conditions. Earlier digital image
stabilization (DIS) methodology involved elaborate and computationally intensive software
to shift an image during the photo shoot so as to counteract the effects of hand motion.
More recently, camera manufacturers are shifting to OIS, which senses and counteracts the
movements of the lens itself. This is enabled by three key components at the heart of a typical
lens shift scheme.
–– Two MEMs-based gyroscopes are used to measure the rate of rotation in terms of yaw
(bearing) or pitch. The voltage output is digitized and integrated to get angle of rotation.
–– Two Hall-effect magnetic sensors are used to measure the displacement of the camera lens
relative to the chassis or cell phone.
–– Voice coil motors (VCM) are used to actuate the movement of the lens to track the position
of an image.
The OIS control loop block diagram is shown in Figure 7. Assuming the mobile phone is oriented
vertically with the camera on its backside, any aberrations in camera direction along the
horizon will be picked up by the “yaw” gyroscope. Any vibrations in tilt will be picked up by the
“pitch” gyroscope. Spiral movements of the camera would be considered “roll” movements and
these are not corrected.
Gyro
yaw
ωyaw
∫
θyaw +
X-Axis
VCM driver
Controller
–
θHSens_X
Gyro
pitch
ωpitch
∫
θpitch
X-Axis
Hall sensor
Y-Axis
VCM driver
Controller
+
–
θHSens_Y
Y-Axis
Hall sensor
Figure 7. Optical image stabilization control loop block diagram.4
09 | Keysight | Low-Frequency Noise Measurements with the E4727A and Their Applications - Application Note
The gyroscopes, magnetometers, and their associated post-processing circuitry must each
be of low enough noise so as to not corrupt the operation of the control loop. Manufacturers
represent gyroscope noise density as “dps/√Hz” or angular rate (in degrees) noise power
density. Magnetometer noise density is represented in “nT/√Hz.” The frequency range for this
characterization must extend below 1 Hz, as shown by hand tremor data presented in Figure 8.
Angular rate spectral density of handshake movements
1.5
Mean
3
Magnitude
1
0.5
0
0
5
10
15
20
25
Frequency (Hz)
Figure 8. Angular rate spectral density of handshake movements.4
The E4727A uniquely enables measurements down to 0.03 Hz, making it well-suited for noise
characterizations on image stabilization systems.
30
10 | Keysight | Low-Frequency Noise Measurements with the E4727A and Their Applications - Application Note
Practical Considerations in Noise Measurements
One of the biggest challenges in measuring component noise is avoiding data corruption by
other noise sources in the system. Each of these noise sources will add in a root sum of the
squares (RSS) fashion, meaning that two independent noise densities of equal power will add to
give rise to a 3 dB increase in the overall noise.
Instrument noise
-
+
Measured noise
DUT noise
-
+
Figure 9. Illustration of instrument reference noise and noise from device under test.
The noise density of the measurement system must be significantly better than the device
under test (DUT) so as to not overestimate DUT noise. Figure 10 presents the apparent
error that may occur when the DUT noise is close to the system reference noise at any given
frequency.
Normalized plot of excess noise caused by reference noise
3.5
Increase in measured noise due to reference noise (dB)
3
2.5
2
1.5
1
0.5
0
0
2
4
6
8
10
12
Amount DUT noise exceeds reference noise (dB)
Figure 10. Increase in measured noise due to reference noise being close to the DUT noise.
14
16
11 | Keysight | Low-Frequency Noise Measurements with the E4727A and Their Applications - Application Note
The amplifier VAMP illustrated in Figure 12 amplifies the low-level noise signals so that the
digitizer may accurately represent the spectral density, thus capturing device noise density.
Since the noise of this amplifier adds in an RSS fashion to the noise of the device under test,
the additive noise of the amplifier becomes a critical specification for sensitivity. The previous
generation 1/f noise measurement system amplifiers proved inadequate for some devices. The
latest E4727A features a newly-developed amplifier with —183 dBV2 /Hz floor (equivalent to
the noise of a 28 Ω resistor) and a corner frequency of only 20 Hz (Table 1). Figure 11 presents
the additive voltage noise of the LNA in the latest E4727A as compared to its predecessor, the
E4725A. We find that the E4727A has a much lower noise density, enabling industry-leading
measurement sensitivity and modeling accuracy.
E4725A
—177 dBV2 /Hz
1.4 nV/√Hz
10 kHz
Noise floor
Corner frequency
E4727A
—183 dBV2 /Hz
0.67 nV/√Hz
20 Hz
Table 1. Tabulated LNA performance of predecessor E4725A and current E4727A measurement
system.
10
LNA noise of A-LFNA systems
3
E4727A
E4725A
Noise density (nV/sqrt(Hz))
10
10
10
10
2
1
0
-1
1
10
100
1k
10k
100k
Frequency (Hz)
Figure 11. Noise floor comparison of E4727A and E4725A A-LFNA systems.
The additive noise of the amp preceding the digitizer is one important contributor to system
noise, but careful attention must be paid to other noise sources, including the following:
–– Noise from the power supplies used to bias the DUT. On an MOS device, this could include
connections to the substrate or gate.
–– Thermal noise from resistors used to set source and load impedances to the device under
test.
–– Coupling of Earth ground noise to instrument chassis. Besides connecting all the
instruments to the same power circuit, we found it helpful to isolate the chassis ground
from Earth ground.
–– Thermoelectric voltage noise caused by uneven airflow from ambient room air turbulence.
As an example, let’s consider a discrete voltage reference such as Linear Technologies
LT1021. The datasheet shows output voltage noise versus time before and after a simple
foam cup shield is removed. The results showed a strong influence from the airflow in the
0.01 Hz to 1 Hz band.5
1M
12 | Keysight | Low-Frequency Noise Measurements with the E4727A and Their Applications - Application Note
E4727A Noise Measurement Solution
Fortunately, the E4727A Advanced Low-Frequency Noise Analyzer (A-LFNA) architecture helps
address each of the noise measurement challenges presented above. To measure noise in a
CMOS device, a source measurement unit like the Keysight B1500A is used to apply bias and
measure DC operating points. When measuring noise however, the source measurement unit
(SMU) noise contribution must be filtered out. The voltage noise is amplified and analyzed using
a high-speed digitizer. Figure 12 presents one possible configuration of noise measurement,
although many others are possible using these modules. The variable resistance, switching and
filtering functions are included in the A-LFNA modules, which when combined with the A-LFNA
software, leads to a powerful and accurate noise measurement solution.
Source measurement unit (SMU)
LPF
RLoad
Drain
Gate
SMU
LPF
RSource
E4727A Input module
To digitizer
VAMP
FET
DUT
Source
E4727A Output module
Figure 12. Drain voltage noise amplified and digitized, while maintaining control of source and load
impedances and avoiding corruption of the noise data from the source measurement unit bias.
For different types of devices, different source and load impedances (RSource and RLoad)
are required. For example, if RLoad is too low, the poor noise voltage transfer to the digitizer
will degrade signal-to-noise ratio and measurement sensitivity. If Rload is too high, then
the thermal noise of the resistance will dominate the noise measurement, also reducing
measurement sensitivity. Excessive RLoad will furthermore play against the input capacitance
of the amplifier “VAMP” to roll off the apparent noise measurement. The A-LFNA software is
able to judiciously select RSource and RLoad based on device type (FET, diode, BJT, etc.) and
device current, which in turn has a strong impact on the impedances presented by the device.
Given its selection, the software can predict the frequency at which the data has rolled off due
to the RC filtering from the amplifier and chop the data at that frequency. Figure 13 illustrates
measured data on an NMOS device at a number of bias currents. In this case, since the input
impedance of the MOSFET is exceedingly high, RSource = 0 Ω. We observe evidence of varying
RLoad at different current levels based on the cutoff frequency of the measurement. An ultralow frequency amplifier (“VAMP-ULF”) is switched in for measurements below 1 Hz.
13 | Keysight | Low-Frequency Noise Measurements with the E4727A and Their Applications - Application Note
Noise measurements on NMOS device, Vds=1.7V
10
-14
VAMP-HF
VAMP-ULF
10
-16
2
Sid (A /Hz)
10 -18
10 -20
10 -22
Id=100uA
10 -24
Id=1uA
Id=100nA
10
-26
10m
1
100
10k
1M
100M
Frequency (Hz)
Figure 13. Noise data measured on a typical NMOS transistor at drain current levels of 100 µA, 1 µA and
100 nA.
14 | Keysight | Low-Frequency Noise Measurements with the E4727A and Their Applications - Application Note
Best Solution for Device 1/f Noise Measurement
The E4727A hardware consists of a series of modules paired with a PXI chassis housing a
mainframe computer and digitizer. These modules are connected to a source measurement unit
like the B1500A to enable both flexible and clean device biasing and noise signal conditioning,
as shown in Figure 14.
Source measurement unit (SMU) bias, triaxial shielded cable
Drain
BG
Gate
Source
Input module
Output module
Digital
control
Digitizer
Substrate module
Figure 14. E4727A is combined with a source measurement unit like the B1500A to create a powerful noise
modeling platform.
Due to their small size, the modules may be positioned as close as possible to the device
under test, so as to minimize the undesired parasitic capacitance from the cabling. Figure 14
illustrates how the output module may be connected across the drain and source of an FET. It
may be similarly connected across the collector and emitter of a bipolar transistor, anode and
cathode of a diode, the two terminals of a resistor or the output terminals of an operational
amplifier IC. A “fixture module,” included with the E4727A and shown to the top right in Figure
15, facilitates measurements on discrete packaged components. On-wafer testing is controlled
by an easy-to-use software interface that tracks device location, wafer lot and geometry.
The resultant data may then be ported over to powerful device modeling software like Model
Builder Program (MBP) and Integrated Circuit Characterization and Analysis Program (IC-CAP).
15 | Keysight | Low-Frequency Noise Measurements with the E4727A and Their Applications - Application Note
Figure 15. Modules of E4727A: interface module, input module, output module, substrate module and test
fixture module.
16 | Keysight | Low-Frequency Noise Measurements with the E4727A and Their Applications - Application Note
Conclusion
The Keysight E4727A Advanced Low-Frequency Noise Analyzer is the next-generation system
for characterization and analysis of 1/f flicker noise and random telegraph signal noise (RTN).
Designed for on-wafer probed devices, discrete package devices, and low noise integrated
circuits, it uniquely allows for versatile noise measurement under diverse conditions, including:
ultra-low frequency, ultra-low current, high current, high voltage, or high power. Given its
state-of-the-art noise sensitivity and modular design, the E4727A is well suited for low
frequency device noise measurements, especially when paired with a source measurement
unit like the B1500A. For device modeling, characterization and statistical process control, this
solution features an easy-to-use software platform. Measurements ranging from 0.03 Hz to
40 MHz are now possible, enabling device and circuit designs that push the frontier in
sensitivity and noise.
Items
Measurement devices
Maximum frequency range
AMP
Voltage AMP
Current AMP
Noise floor
Corner frequency
Noise floor
Corner frequency
Maximum bias range
Specifications
FET, BJT, Diode, Resistor, Circuit
0.03 Hz ~ 40 MHz
Voltage AMP x3, Current AMP x2
—183 dBV 2 /Hz (0.67 nV/√Hz)
20 Hz
1E-23 A 2 /Hz
200 Hz
200 V/0.1 A (10 Wmax)
Table 2. Key specifications for E4727A Advanced Low-Frequency Noise Analysis system.
References
1. Beach, M., Fornetti, F., and Rathmell, J.G., “The Application of GaN HEMTs to Pulsed
PAs and Radar Transmitters”, European Microwave Integrated Circuits Conference, 7th
European EuMIC 2012, Amsterdam, p. 405-408, Oct. 29 – 30, 2012.
2. Franco, Sergio, “Design with Operational Amplifiers and Analog Integrated Circuits, 3rd
edition”, McGraw-Hill, August 8, 2001.
3. Kyung-Whan, Yeom, Microwave Circuit Design: A Practical Approach Using ADS,
Prentice-Hall, May 22, 2015.
4. La Rosa, Fabrizio, Celvisia Virzì, Maria, Bonaccorso, Filippo, Branciforte, Marco “Optical
Image Stabilization (OIS)”, STMicroelectronics.
5. Linear Technologies, “LT1021 Datasheet”, 1995.
6. O’Donnell, Dr. Robert M, “Radar Systems Engineering, Lecture 10, Transmitters and
Receivers”, http://bit.ly/1XrlOk0.
17 | Keysight | Low-Frequency Noise Measurements with the E4727A and Their Applications - Application Note
For more information on Keysight
Technologies’ products, applications or
services, please contact your local Keysight
office. The complete list is available at:
www.keysight.com/find/contactus
Download your next insight
Keysight software is downloadable expertise.
From first simulation through first customer
shipment, we deliver the tools your team needs
to accelerate from data to information to
actionable insight.
–– Electronic design automation (EDA)
software
–– Application software
–– Programming environments
–– Utility software
Learn more at
www.keysight.com/find/software
Start with a 30-day free trial.
www.keysight.com/find/free_trials
From Hewlett-Packard through Agilent to Keysight
For more than 75 years, we‘ve been helping you unlock measurement insights. Our
unique combination of hardware, software and people can help you reach your next
breakthrough. Unlocking measurement insights since 1939.
1939
THE FUTURE
myKeysight
www.keysight.com/find/mykeysight
A personalized view into the information most relevant to you.
Americas
Canada
Brazil
Mexico
United States
(877) 894 4414
55 11 3351 7010
001 800 254 2440
(800) 829 4444
Asia Pacific
Australia
China
Hong Kong
India
Japan
Korea
Malaysia
Singapore
Taiwan
Other AP Countries
1 800 629 485
800 810 0189
800 938 693
1 800 11 2626
0120 (421) 345
080 769 0800
1 800 888 848
1 800 375 8100
0800 047 866
(65) 6375 8100
Europe & Middle East
Austria
Belgium
Finland
France
Germany
Ireland
Israel
Italy
Luxembourg
Netherlands
Russia
Spain
Sweden
Switzerland
United Kingdom
0800 001122
0800 58580
0800 523252
0805 980333
0800 6270999
1800 832700
1 809 343051
800 599100
+32 800 58580
0800 0233200
8800 5009286
800 000154
0200 882255
0800 805353
Opt. 1 (DE)
Opt. 2 (FR)
Opt. 3 (IT)
0800 0260637
For other unlisted countries:
www.keysight.com/find/contactus
(BP-02-10-16)
This information is subject to change without notice.
© Keysight Technologies, 2016
Published in USA, April 28, 2016
5992-1537EN
www.keysight.com