```Demystify Noise Figure Measurements
8 Oct 2013
14:00h - 14:40h
Presented by: Stefano Balzarini
EuMW Seminars 2013
Agenda
• Background
• Definition and Importance
• Methods
• Y-factor (or hot/cold)
• Cold source
• Implementations
• Importance of linearity and gain adequacy
• Measurement steps
• Uncertainties
• The basic model, terms of interest
• What really matters?
• The Uncertainty Calculator
• Summary
EuMW Seminars 2013
2
Definition
F = Noise Factor
Si = Input signal power
Ni = Input noise power
So = Output signal power
No = Output noise power
F
GDUT = Gain of amplifier
S
N
S
N
input

1
GDUT

N input  GDUT  N added
N input
output
Noise figure (NF) is:
NF (dB) = 10 * log (F)
Easy enough…a quasi-normalized measure of noise power added by the DUT.
the type of signals in the measurement, temperatures, etc.
EuMW Seminars 2013
3
Noise Parameters
What are noise parameters?
A measured NF value is only valid for a source impedance of 50 ohms (usually). At other
impedances, the practical noise figure varies and is DUT dependent (usually).
Source reflection coefficient
for minimum NF
Equivalent noise resistance
F  Fmin 
4  R n  Γ S  Γ opt
2
2
Z 0  1  Γ S   1  Γ opt


2
+j1.0
+j0.5
+j2.0
5.0
2.0
1.0
0.5
0.0
+j5.0
0.2
+j0.2
-j0.2

-j5.0
-j0.5
Concept: present different source reflection coefficients (with tuner)
and measure NF; solve for the unknowns
-j2.0
-j1.0
EuMW Seminars 2013
4
Importance
Why is NF important?
- Cellular systems  voice quality
- Digital communications  bit-error rate
Relative cost for amplifiers within one 3350 GHz family
3
Normalized price
Because of this importance,
- NF is still a critical customer-vendor
specification
- It sometimes has a steep cost-function
(especially for ultra-low-noise cases
and at mm-wave frequencies)
2
1
0
2
EuMW Seminars 2013
5
3
4
Specified noise figure (dB)
5
NF Methods: Y-factor (hot-cold)
What is it? Historic method that is based on a ratio of noise powers
(avoids absolute power cal).
Noise
source
Noise
preamps…
DUT
Composite
@ TC and TH
Noise power is measured when driven with a source in a cold state (near room temp) and a
hot state (often with 10-20 dB more noise power).
Y 
Y-Factor
output
Excess Noise Ratio = [(TH - TC) / T0]
ENRdB = 10 log10 [(TH - TC) / T0]
To = 290o Kelvin
input
Noise
power
Cold state
Hot state
F 
Time
EuMW Seminars 2013
Noutput _ hot
Noutput _ cold
6
ENR  1
Y 1
NF Methods: Y-factor
Noise Source Issues:
• Difficult to calibrate
• Significant uncertainty-adder (often > 0.15 dB)
• Fragile (will go way out-of-cal if dropped)
• Impedances in hot and cold states are not equal
(problems both from power delivery and from
noise parameters)
High gain broadband amps can actually be damaged
by the hot state.
|Gs| for hot and
cold states
EuMW Seminars 2013
7
Methods: Cold Source
What is it? An absolute noise power difference (not a relative one like before)
Implication: Need an
cal/power cal
The only input to the DUT is the noise
power (kT0B) from a termination at T0,
From the S/N definition:
F
EuMW Seminars 2013
S
N
S
N
input

1 kT0B  Nadded Bandwidth (B) is set by the IF

system; can be known a priori
G
kT0B
output
FDUT 
1
GDUT
8

NDUT rcvr  Nrcvr
GDUTkT0B
Methods: Cold Source
Cold source is the generally accepted methodology used by VNA and applied up to
~70 GHz until the recent announcement from Anritsu …
Industry-first!
70 kHz to 125 GHz VNA
Noise figure measurement
capability
Industry-first!
Optimized noise
measurements from
30 GHz to 125 GHz
EuMW Seminars 2013
9
Measuring Noise Power
Noise power can be measured statistically
A deterministic
background signal
IF
time
By acquiring many samples and performing an RMS-like calculation,
slowly-varying or constant interferers are removed.
N
Noise_powe r 
EuMW Seminars 2013

k 1
10
b 2,k
N
2
 N

b
 k 1 2,k

N











2
Selection of acquisition parameters (IFBW) and # of RMS points:
- Trade-off between data jitter and sweep time
- 100 kHz/3000 is default
- 10 kHz/6000 for ~0.1 dB jitter at reasonable gains
NF (dB)
Data jitter vs. acqusition settings
6
10kHz/6000
5
10kHz/2000
100kHz/6000
The dependence is stronger at
parameter extremes and in lownet-gain situations.
4
3
6
EuMW Seminars 2013
7
Frequency (GHz)
8
11
Methods: Cold Source (cont.)
Considerations with the cold source method:
Accuracy of the power/receiver calibration: The power calibration can be affected
by mismatch, sensor linearity, harmonic contributions,…
Tighter linearity constraints: Since it is more of an absolute noise power
measurement, receiver linearity problems map directly to NF uncertainty
(including linearity between receiver cal level and measurement level).
EuMW Seminars 2013
12
Accurate noise figure measurements require high composite receiver gain and
appropriate filtering.
External Component Benefits
B
•
•
•
Greater flexibility in composite
Can optimize for best
performance with DUT
Easier to service and maintain
The difference is whether the components for the composite receiver are inside or
outside. If inside, a lot of switching and IF receiver modifications are needed - which
can degrade performance of the standard s-parameter measurements.
EuMW Seminars 2013
13
Implementation: VS mm-wave
• Based on unique 3743A Freq Ext Mod
o 8% the weight and 2% the volume of other
solutions
• coupling loss deleted,
• un-needed multiplier paths removed
 Still need to incorporate into a composite
EuMW Seminars 2013
14
Two key components: pre-amplifiers and filter(s):
• Select the proper amplification such that the resultant noise power into the VNA
•
•
•
Not too low.
Not too high.
For VectorStar™, generally want composite receiver gain + DUT gain between ~40
dB and 70 dB for linearity optimization.
•
•
Using multiple pre-amplifiers is ok.
Lower pre-amplifier noise figures are better. But if the DUT gain exceeds ~10 dB, it
does not matter much.
EuMW Seminars 2013
15
Example: What may cause low end linearity issue?
-Discretization floor of flipping a few least significant bits (LSBs)
-Integral nonlinearity (INL) of the ADC – sometimes related to charge
storage in the sample and hold circuit
dBFS output
ADC integral nonlinearity effect (14 bit,
typical)
-40
Rarely plotted this way by
can be important for noise
signals
-50
ideal
practical
-60
-60
EuMW Seminars 2013
-50
dBFS input
-40
16
Select the proper filtering to minimize unwanted images:
•
•
•
Most VNAs incorporate harmonic downconversion
Any downconverter has responses for at least some harmonics of the LO (commonly
3rd, 5th, 7th… for balanced mixers and all harmonics for some structures).
Isolate the harmonic of interest
•
Example: Measuring amplifier NF at 18 GHz
9
18 27 36
EuMW Seminars 2013
– The receiver will be “listening” at sub- and
super-multiples of 18 GHz.
– If the pre-amps have gain there, use filters.
– If the preamps only covered 1-20 GHz, a
high pass filter might be adequate.
17
Limited
The example below highlights the effects of having a limited
noise figure measurement range.
Another
instrument:
Too
low pre-amp
gain Too low a preamp gain
just right (~35 dB)
Just right
too(~40
much
Too much gain
dB)gain...(~40 dB)
4
4
3
3
2
-2
NF (dB)
NF (dB)
NF (dB)
6
2
1
1
30 dB gain preamp
2
20 dB gain preamp
-6
0
0
1.5
1.7
1.9
2.1
Frequency (GHz)
EuMW Seminars 2013
2.3
2.5
1.5
1.7
18
1.9
2.1
Frequency (GHz)
2.3
2.5
1.5
1.7
1.9
2.1
Frequency (GHz)
2.3
2.5
Implementation: VS- Measurement Process
FDUT 
1
GDUT

NDUT rcvr  Nrcvr
GDUTkT0B
Three main steps:
1. Receiver calibration (may include a power cal)
2. Noise calibration
3. DUT measurement
B
DUT gain (GDUT) should be
measured beforehand.
T0 is room temperature, nominally.
B is the IF bandwidth of the VNA.
Depending on the DUT,
30-50 dB gain
EuMW Seminars 2013
19
Need to avoid
images
Implementation: DUT Gain Measurement
A note about the DUT gain measurement … make sure the DUT is nowhere near
compression.
NF vs. S-parameter compression
7
NF (dB)
6
start
5
.3 dB compressed
4
1 dB compressed
3
5
EuMW Seminars 2013
6
Frequency (GHz)
7
20
•
Establish absolute power reference
- Power calibrations help accuracy considerably
• Must be done at a low level (usually about -50 to -60 dBm) due to
the use of the composite receiver
B
Power cal at this plane
over the relevant
frequency list
EuMW Seminars 2013
21
Comparison between using a noise source or VNA source (with power meter
calibration) as the absolute power reference
Noise Source
VNA Source
Re-calibration complexity
Stability of source calibration (vibration)
More thermal instability
Match variance
Higher fundamental uncertainty
Need +28V switching and timing control
with Power Cal
Better calibration stability
Better fundamental uncertainty
Easier handling of different signal levels
EuMW Seminars 2013
22
Implementation: Noise calibration
Measurement of the noise power of the receiver with input terminated.
B
EuMW Seminars 2013
23
Implementation: DUT measurement
Insert the DUT to make the NF measurement.
B
NOTE: It is worthwhile to check the absolute power being
is not being overdriven.
EuMW Seminars 2013
24
Implementation: Comparisons
ZJL-6G amplifier
5.5
NF (dB)
5
4.5
4
VS -041
8970
upper2
lower2
3.5
0.1
0.6
1.1
Frequency (GHz)
Preamp
linearity issues
EuMW Seminars 2013
25
1.6
Uncertainties
NF measurement uncertainty can be affected by many factors:
B
Residual image responses, discrete
leakage and RMS computation scatter
DUT S-parameter
uncertainty
Mismatch
Mismatch, composite receiver linearity (net gain),
=
1
+∆
+
((+ −  )+(∆+ − ∆ ))(1+∆)
0 (+∆)
EuMW Seminars 2013
26
Uncertainties: What matters?
• Receiver gain when the DUT gain drops
Example amplifier @ 10 GHz
Noise figure (dB)
Required receiver gain for 0.3 dB
uncertainty on example amplifier
Fixed 40 dB rcvr gain
4
3
nominal
upper
lower
2
0
20
DUT gain (dB)
20
0
10
30
50
Frequency (GHz)
70
40
With fixed composite receiver gain, as DUT
gain drops, uncertainty increases
EuMW Seminars 2013
40
More composite receiver gain is needed to
keep the same uncertainty as the DUT gain
drops
27
Uncertainties: What matters?
• Match, particularly between DUT and receiver
Unc. on example 3 dB NF amp
(dB)
Uncertainty vs. mismatch
EuMW Seminars 2013
1
|S22|=-15 dB
|S22|=-6 dB
0.5
0
-30
-25
-20
-15
28
-10
Uncertainties: What matters?
Unc. on example 3 dB NF amp
(dB)
Uncertainty vs. power accuracy (into rcvr cal)
1
0.5
0
0
0.2
0.4
0.6
Power accuracy (dB)
At low power uncertainties, other terms dominate
EuMW Seminars 2013
29
0.8
1
Uncertainties: Calculator
3 basic modes of operation:
(enter fixed or actual DUT .s2p and NF data as well as various measurement parameters)
-
Uncertainty vs. frequency
-
Uncertainty vs. DUT gain (at one frequency)
-
Required receiver gain (vs. frequency) for a given uncertainty
EuMW Seminars 2013
30
Summary
• Background
• Definition and Importance
• Methods
• Y-factor (or hot/cold)
• Cold source
• Implementations
MS4640B-041
Noise Figure Measurement