Download Atmospheric Noise - Stanford University

CCIR 322-2
Atmospheric Noise
1
Why do this?
• Understand the evolution atmospheric
noise measurements
• Determine what are appropriate values to
use from the CCIR documentation
• Compare CCIR data alternative noise data
• Define the noise environment so mitigation
techniques may be explored
2
What is CCIR 322-2?
• Reported background atmospheric noise levels
• Evolved into ITU P372-7
– Included man-made and galactic noise
– Removed technical background
• Used the ARN-2 Radio Noise Recorder
– 16 Stations around the globe
– Average noise power at each of eight frequencies for
fifteen minutes each hour
– 13 kHz, 11kHz, 250kHz, 500kHz, 2.5MHz, 5MHz,
10MHz, and 20MHz
– 1957-1961 (4 years) → 8640 15-minute measurements
→ 99.98%
– Tracked filtered noise envelope not instantaneous noise
– Took high speed data to obtain APDs
3
What is CCIR 322-2? (cont)
• Sectioned the year into seasons and time blocks
– Four 90-day seasons
– Six 4-hour time blocks
• Tracked external antenna noise factor, Fa
– Power received through a loss-free antenna Fa =
10*log10(Pn/KToB)
– Lists the median value hourly value for each time
block, Fam, at 1 MHz
– Lists the upper decile (90%) level Du
– Calculate noise E-field from Fa, BW, frequency
• Use normal or log-normal statistics and graphs
to adjust values
4
CCIR Stations
5
Fam for Summer 20-24h
6
Three Step Process for CCIR Data
Look up Expected
Median statistics
(Fam)
Adjust from Median to
appropriate Time
Availability Level
(Fa from Du, %T)
Adjust from Expected
Value (mean) to Service
Probability or
Confidence
(E from Fa, SNR, B, %S)
7
Computing E-field from CCIR
Loran Receiver Specifications
Frequency
BW of system
0.1 MHz
30000 Hz
From CCIR
Fam 1 MHz
93 dB
Fam 100 kHz
135 dB
Vdm 200 Hz
7.9 dB
Vd 30 kHz
12.8 dB
Du
11.2 dB
Delta=A-Arms
-7 dB
Variances of values
sigmaDelta
4.2 dB
sigmaVdm
1.5 dB
sigmaDu
2.5 dB
sigmaFam
3.9 dB
sigmaPowerError
2 dB
Assumed
sigmaSNRError
2 dB
Assumed
8
Computing E-field from CCIR (cont)
Specify Level
% Time D and Fa Exceeded
% Time D and Fa Not Exceeded
kappa (normal deviate)
95%
5%
1.64
Adjust Fam to Fa
D [dB]
sigmaD [dB]
Fa [dB]
Enoise (mean) for % Time [dBuV/m]
14 dB
3 dB
149 dB
64 dB uV/m
Adjust mean value to service prob
Service Probability
99%
kappaServiceProb
2.33
sigmaTotal
6.9 dB uV/m
kappaServiceProb*SigmaTotal
16 dB uV/m
Enoise will be below 95% Time and
with 99% Service Prob
80 dB uV/m
9
Computing CCIR Data
Look up MEDIAN
statistics
(Fam)
+14 dB
<Emedian> = 50 dB uV/m
Adjust for appropriate
Time Availability Level
(Fam, Du, 95%T)
<E95%> = 64 dB uV/m
+16 dB
Adjust for uncertainty in
statistics or Service
Availability or Confidence
(Fa, 99%S)
E95% |99%= 80 dB uV/m
10
Adjustment to Enoise at 95% Time
Availability for Service Probability
Service Probability
90%
95%
99%
99.9%
99.99999%
KServiceProb
1.28
1.64
2.33
3.09
5.20
sT for 95%
Time Availability [DB]
6.9
6.9
6.9
6.9
6.9
9
11
16
21
36
KServiceProb* sT
11
RMS Envelope vs Inst. Envelope
• Failure mode of receiver sometimes dictates
using Inst. Envelope values over RMS Envelope
values (eg FSK).
• Needed a way to measure the “impulsiveness”
of the noise.
• Came up with using Vd
– Vd = 20 * log10( Arms / Aavg)
– Vd for Rayleigh (Gaussian > 0) is 1.05
– Larger Vd mean the more impulsive or less Gaussian
the noise
12
CCIR APD Experimental Setup
• Thanks to Bob Matherson and CCIR
Instantaneous Noise, Vn
IF Inst. Noise Voltage, VIF
Atm Noise
Instantaneous Noise Envelope, A
RMS Noise Envelope, Arms
Average Noise Envelope, Aavg
13
VIF and Envelope Voltages
VIF Voltage and Noise Envelope Voltages
0.015
0.01
Voltage [V]
0.005
0
-0.005
VIF
A
Arms
Aavg
-0.01
-0.015
22.5
23
23.5
Time [msec]
24
24.5
14
25
Envelope RMS
Voltage (Arms) &
Instantaneous
Envelope Voltage (A)
• Arms is 3dB above rms
instantaneous noise
voltage VIFrms
• APD
– Amplitude Probability
Distribution or A
Posteriori Distribution
– Exceedance probability
P[A > Arms + D=A0]
0
• Parameterized by Vd Voltage deviation
• Plotted on Rayleigh
Probability Paper
15
CCIR FSK Example Results
Time availability as a function of service probability
Geneva, Switzerland
Summer season: 2000-2400 h
Frequency: 50 kHz Bandwidth: 100 Hz
Binary errors: 0.05%
Montgomery [1954] Model
16
CCIR Limitation (1983 §1)
• "The estimates for atmospheric noise levels
given in the Report are for the average
background noise level due to lightning in the
absence of other signals, whether intentionally
or unintentionally radiated. In addition, the noise
due to local thunderstorms has not been
included. In some areas of the world, the noise
from local thunderstorms can be important for a
significant percentage of the time."
17
CCIR Limitations (cont)
• “Atmospheric radio noise is characterized
by large, rapid fluctuations, but if the noise
power is averaged over a period of several
minutes, the average values are found to
be nearly constant during a given hour
variations rarely exceeding ±2dB except
near sunrise or sunset, or when there are
local thunderstorms.”
18
CCIR Questions
• What should we specify as a service
probability 95%?
• What is an appropriate level of availability?
• How do the values change when we are
near lightning?
19
dB(V/m/Hz)
Electric/Magnetic Field Data (cont)
20
Preta 1984
Obtaining E-field from Lightning
Return Stroke Data
Lightning E-field noise density at
100kHz and 50km (Preta)
Add conversion for 30kHz BW
Add conversion from dB V/m to dB uV/m
Lightning E-field at 100kHz, 30kHz BW, 50km
-100
45
120
65
dB V/m/Hz
dB Hz
dB
dB uV/m
21
Comparing E-fields and Distance
Lightning E-field at 100kHz, 30kHz BW, 50km
65 dB uV/m
Enoise for 95% Time and
99% Service Prob
80 dB uV/m
• Lightning E-field data varies with 1/D
• E2-E1 = 20*log10(D1/D2) = 20*log10(50km/D2)
• For 100 kHz, an Enoise of 80 dB uV/m is equivalent to being
8.7 km from continuous repetitive lightning return strokes!
• Can do this for all levels of % Time availability for any Service Probability
22
Conclusion from CCIR Data
• Calculate Noise Level
• Specify % Time and Confidence
• Alternatively, as a distance to a lightning
return stroke. I.e. define this as a ATC
constraint
23
Summary
• Reviewed CCIR-322 and current lightning
research and found good correlation between
both data sets.
• Need to set both a not-to-exceed level and a
confidence to make a reasonable estimate of
noise level or can define a distance to a lightning
return stroke. e.g. define this as a ATC
constraint.
• Need to use nonlinear processing to increase
SNR.
• Have pieces for a time-domain model.
24
Backup Slides
25
Processing Gain
26
Mitigation Techniques
• Clipping or Blanking based on Vd
– Spaulding/Middleton
– Modestino/Enge
• Using the Loran antenna as a lightning
detector/locator
• Low frequency signals propagate faster
than higher frequency signals
• Magnetic antenna / magnetometer
27
Spaulding/Middleton (Parametric)
• Divide all noise into three canonical forms
– Class A (3 parameters)
• (duration of noise)(receiver bandwidth) >> 1
• Negligible transients are produced in linear front-end of receiver.
• Steady state series of waves produced.
– Class B (3 or 6 parameters)
• (duration of noise)(receiver bandwidth) << 1
• All transients are produced in linear front-end of receiver.
• Overlapping impulse response.
– Class C (8 parameters)
• (duration of noise)(receiver bandwidth) << 1
• Additive mixture of Class A and B
• Use optimal adaptive non-linear filtering for improvement
• Use suboptimal non-linear filtering (e.g. hole punch, hard
limiter, etc)
28
Spaulding 1986
l x ( x j | P( ) ) 
d
log f ( x j | H 0 )
dx j
29
Spaulding 1986
~30dB
Improvement!
~30dB
30
Spaulding 1986
Summary
• Reviewed some signal processing with nonGaussian Noise
• Much to be learned from the already
published work regarding the processing of
signals in non-Gaussian noise.
• Large gains may be made by utilizing
adaptive filters and even sub-optimal ones.
• Currently trying to fit CCIR APD to 3Parameter Class B Model to various Vd in
order to bound gains
31
Negative Leader CG Lightning
32
National Lightning Detection Network
• Privately run
• Real-Time data
collection
• Mostly detection and
high level data
• Some waveform data
• Can use newer analytical
field models
33
Sample Data
Date
Time
Lat
Lon
Amplitude
Range
Az
# Strk
5/9/2001
19:12:08
36.065
-106.483
-14.7 kA
18.9 mi
327.7 deg
1
5/9/2001
19:14:07
36.095
-106.431
-13.0 kA
19.4 mi
338.3 deg
1
5/9/2001
19:19:34
36.088
-106.432
-13.4 kA
19.0 mi
337.6 deg
1
5/9/2001
19:22:13
36.07
-106.471
-14.6 kA
18.9 mi
330.1 deg
1
5/9/2001
19:27:12
36.043
-106.404
-14.2 kA
15.5 mi
338.5 deg
2
5/9/2001
19:28:18
35.983
-106.369
-22.9 kA
11.0 mi
340.2 deg
4
5/9/2001
19:28:22
36.067
-106.491
-16.0 kA
19.2 mi
326.8 deg
4
5/9/2001
19:29:13
35.956
-106.314
-12.3 kA
8.5 mi
355.5 deg
1
5/9/2001
19:29:31
36.05
-106.392
-12.3 kA
15.7 mi
341.5 deg
1
5/9/2001
19:29:57
35.864
-105.964
+11.2 kA
19.0 mi
83.9 deg
1
5/9/2001
19:30:04
36.055
-106.472
-32.5 kA
18.0 mi
328.1 deg
2
5/9/2001
19:31:53
36.037
-106.443
-19.2 kA
16.1 mi
330.8 deg
3
5/9/2001
19:32:14
36.066
-106.337
-11.7 kA
16.1 mi
353.1 deg
1
5/9/2001
19:32:42
36.043
-106.454
-10.0 kA
16.7 mi
329.6 deg
1
5/9/2001
19:33:13
36.065
-106.458
-15.4 kA
18.2 mi
331.4 deg
1
5/9/2001
19:33:24
35.978
-106.291
-16.2 kA
10.0 mi
3.7 deg
2
5/9/2001
19:34:47
36.035
-106.419
-15.5 kA
15.4 mi
334.9 deg
3
5/9/2001
19:35:50
36.053
-106.433
-12.6 kA
16.8 mi
334.3 deg
1
5/9/2001
19:37:14
36.057
-106.421
-10.8 kA
16.8 mi
336.8 deg
1
5/9/2001
19:38:14
36.05
-106.401
-29.6 kA
15.9 mi
339.7 deg
2
5/9/2001
19:38:54
36.06
-106.413
-13.9 kA
16.8 mi
338.5 deg
342
… for another hour and a half !
Lightning Characteristics
Type
Stepped Leader
Return Stroke
Intracloud
Event Time [usec]
Duration [msec]
4
0-550
200
0-1100
2
0-10
35
Electric/Magnetic Field Data
Intracloud
Stepped Leader
Return Pulse
Weidman 1981
Intracloud
36
Krider 1975
Lightning Spectrum
Return
Positive Intracloud
Stepped Leader
Negative Intracloud
37
Comparing E-fields and Distance
Lightning E-field at 100kHz, 30kHz BW, 50km
65 dB uV/m
Enoise for 90% Time and
99.99999% Service Prob
96 dB uV/m
• Lightning E-field data varies with 1/D
• 20*log10(E2/E1) = 20*log10(D1/D2) = 20*log10(50km/D2)
• For 100 kHz, an Enoise of 96 dB uV/m is equivalent to Lightning Return Stroke
at 1.4 km!
• Can do this for all levels of % Time availability for any Service Probability
38
Equivalent Power between % Time Fa Exceed and Lightning at a Given Distance
With Given Service Probability
Equivalent Distance to Continuous Lightning Strike [m]
100000
10000
1000
100
10
Median Service Prob
99.99999% Service Prob
1
1e-005
0.0001
0.001
0.01
0.1
Percentage of Time Fa Exceeded [%]
1
5 39
10
CCIR Median Data for 99% Level
Atmospheric Noise 99% Worst Case of All Time and Seasons [dB V/m]
50
65
45
Latitude [deg]
85
75
40
70
35
80
30
25
50
-120
-110
55
-100
60
65
-90
Longitude [deg]
70
-80
75
-70
80
-60
85
90
40
Time Domain Model
• Thunder days for this hemisphere
– ~60 in N. Am / ~100-200 Central/South Am
• Lightning stroke data from NLDN
• Electric/magnetic field models from Uman
• CCIR alternative based on APD
– Gaussian portion - background
– Impulsive portion – strikes
41
Envelope APD
0
42
Plot of APD on CCIR Probability Paper
40
30
3 Parameter Fit
Amplitude [dB]
20
10
0
-10
-20
-30
.0001.001 .01
.1
.5
1
5
10
20
30
40
50
60
70
80
85
Percent Time Ordinate Exceeded
90
95
98
43
99