Beruflich Dokumente
Kultur Dokumente
Ryan Said
November 5, 2008
Amplitude
~ d~s
B
(1)
S
(Recall that the induced voltage is given by Faraday's law:)
Vind =
The denition of self inductance is
dt
I
= BN A.
(2)
L=
For a loop antenna with N windings and area A,
L=
If we know the current
Ical
BN A
I
L, N ,
and
A,
we can calculate
Bcal .
(4)
Bcal =
eld
So (3) becomes
(3)
Ical
LIcal
NA
(5)
acal
to the magnetic
The calibration number, the factor that converts measured amplitudes to the incident magnetic
Bcal
acal . The next two subsections describe the calibration circuits in the HUMSNIFFER
and AWESOME preamps.
eld strength, is simply
1.1
The HUMSNIFFER box contains a calibration circuit which generates a 1V peak-to-peak square wave from
the Test Point to ground. This signal passes through a voltage divider to become
we can readily nd
Vcal .
Using
Rcal = 10k,
The voltage divider scales the (one-sided) voltage at the Cal test point
by 22.1/(22.1+49900).
Vcal =
1.2
The AWESOME preamp card contains a microprocessor which generates a 10-bit psuedo-random number
sequence (PRNS). The bit width is
T2 = 1023 T1
T1 = 1/256e3
seconds. Measuring the voltage at +Cal Test Point with respect to ground, a 1 bit is
e
A
w.r.t. ground). If the dierential voltage between the + and - Cal Test points is measured, the PRNS
should alternate between
1V ,
e=
2A
2V.
As with the HUMSNIFFER calibration circuit, this signal passes through a voltage divider to become
Vcal .
Again we have
Rcal = 10k,
and we nd
e*100/(100+2200)
Vcal amplitude of A = A
Vcal
1
T2 . Noting that the Fourier Transform of a square pulse
centered at 0 with width T1 is proportional to sinc(f T1 ) and that the frequency domain is sampled every
1
, we nd that the Fourier Series of the PRNS x(t) is given by
df = T12 = 1023T
1
multiplied by a delta train with impulses every
F.S.{x(t)}[k] fk =
A
2
X
2 sinc
k
1023
k=0
k 6= 0
k/(1023 T1 ) = k 250.244
1
T
+A
Z
T
X sinc(k/1023)
is the one-sided
|f (t)|2 dt =
(6)
|fk |2
(7)
k=
A2
A2
=
+
2
4
X
2
2 X
sinc
k6=0
k
1023
(8)
Using
k=
sinc2 (kb) = 1b ,
b1
(9)
(which is readily derived by applying Parseval's theorem to the inverse Fourier Series of
sinc(kb)),
and
therefore
k6=0
sinc2 (kb) =
1
b
1,
b1
(10)
we have
2 2
A
X
=
(1023 1)
2
2
(11)
X=
sinc envelope of the Fourier series. At 50 kHz, the amplitude of each tone is lower by
0.55dB
20log10 (sinc(50e3/256e3)) =
NOTE: in Calibrate_VLF,
e was
A
Points was assumed to be 1V), yielding calibration values that are 6dB too low.
Cal Test
1.3
acal ),
amplitude at specic frequencies. To measure the single-sided peak amplitude at a particular frequency:
Take a DFT of the result (for speed, use an FFT with a power of 2)
Specically, windowing
x[n]
with a window
w[n]
l samples, we normalize
1,
k = 0, N2
F =
2, otherwise
oset by
P
,
A[k] = F DF T {x[n]w[nl]}[k]
w[n]
n
kfs
N , where
A[k] at frequency
(12)
fs is the sampling
frequency.
Using a Gaussian window increases the accuracy of quadratic interpolation in the (dB) amplitude.
1.4
x[n]
at frequency
kfs
N is given
by
and so
kfs
N
2
F |DF T {x[n]w[nl]}[k]|
P
2
fs
|w[n]|
n
pT 2
Hz
,
F =
1,
2,
k = 0, N2
otherwise
(13)
pT
carries units of
.
must be
An estimate of the Power Spectral Density (PSD) is obtained by taking the average of many successive
periodograms, where each
Pl
Pl
L
P SD =
(14)
where L is the total number of sample periodograms used. This technique for estimating the PSD is known
as Welch's method.
1.5
Receiver Sensitivity
Figure 1 shows the measured, uncalibrated Power Spectral Density from several hardware setups using the
AWESOME receiver. With the minimum value set to
16-bit sampling, covering the range
215
to
215 1
=1
oor (QNF) is
QN F = 10log10
shown by the dotted black line.
2 1
12 fs /2
pT
= 57.8 dB
Hz
(15)
The blue curve shows the measured noise oor, obtained using Welch's
method, of the receiver with no lter card. This noise oor is generated from a noise source after the lter
card. The green curve shows the noise oor with the lter card inserted into the preamp, but with no preamp
connection. This noise oor is generated by a noise source introduced in the anti-aliasing lter card which
eectively reduces the dynamic range of the system by
overall dynamic range to
80 dB.
Preamp: 30 dB Gain
dB-raw/sqrt(Hz)
10
Preamp: 20 dB Gain
20
Preamp: 10 dB Gain
30
Preamp: 0 dB Gain
40
50
Quantization Noise Floor
60
0
10
20
30
Frequency [kHz]
40
50
The next four plots (red, cyan, magenta, and yellow) show the uncalibrated (raw) noise oor with the
preamp connected at the four possible gain settings: 0, 10, 20, and 30 dB, respectively. At the 0 dB gain
setting, the noise oor is dominated by the lter card above 20 kHz and by the preamp card below 20 kHz.
Increasing the gain in the preamp by 10 dB, we see that the noise oor increases by
and
7 dB below 20 kHz
3 dB above 20 kHz. Since the input signal is amplied by 10 dB, we will have an improvement of 3
dB (7 dB) below (above) 20 kHz in the signal to noise ratio (SNR). The dynamic range will also be reduced
by 7 dB (3 dB) below (above) 20 kHz.
increased by 10 dB, and so the SNR will remain the same while the dynamic range is reduced by 10 dB.
This happens again going to the 30 dB gain setting.
The optimal gain setting will therefore be application dependent.
portance, then the 0dB gain setting should be used. If sensitivity is more important, than the 10 dB gain
setting should be used. Little is gained by going to the 20 dB and 30 dB gain settings.
The PSD curves drawn in Figure 1 correspond to raw uncalibrated data. The input-referred sensitivity is
Bcal
acal , where Bcal depends on the
antenna A area and number of windings N , as in equation (5). Figure 2 shows the input-referred sensitivity
2
using the 0dB gain setting for two antenna congurations: T1 (A = 1.69 m , N = 12) and T2 (A = 17.64
17.646
2
m , N = 6). The sensitivity of T2 is 20log10 1.6912 = 14.4dB better than T1. For reference, the red curve
shows the natural noise oor, and the dotted blue and green curves show the theoretical sensitivities for the
calculated by scaling each frequency component by the calibration number
two antenna sizes. The frequency axis is plotted on a log scale to highlight the features.
Increasing
N A
of the antenna improves sensitivity and SNR. If the local measured noise oor is close
to the receiver sensitivity, then that site will benet from an antenna with a larger
that increasing
N A
N A.
Note, too,
increases the gain, thereby reducing the maximum peak amplitude before receiver
saturation.
Figures 3 and 4 plot the measured PSD in blue, together with the natural noise oor for reference and
the measured receiver sensitivity in magenta, at Tucson, AZ and Juneau, AK, respectively. At Tucson, a
smaller T1 antenna is used, but the measured background noise is still higher than the receiver sensitivity.
T1
T2
T1
T2
0.5
10
Frequency [kHz]
50
At Juneau, the larger T2 antenna is used, allowing for measuring a natural noise oor which is
15 dB lower
1.6
A Note on Windowing
The measured spectral dynamic range will be limited by the relative side lobe height of the windowing
function
w.
For example, if a rectangular window is used, the rst side lobe is only 13 dB down from the
main-lobe peak. While the side lobe level rolls o at approximately 6 dB per octave, the measured spectrum
will still have a much lower maximum dynamic range than the receiver. In general, the windowing function
should be chosen to ensure that the side lobe level does not limit the measured dynamic range. A Kaiser
window with a high
provides a convenient trade-o between main lobe width and side lobe height.
System Response
raw
pT , for one channel. The Broadband
Analysis Tutorial goes through the mechanics of the amplitude and phase calibrations in greater detail.
The attenuation below 1 kHz and above 48 kHz are reected in the receiver sensitivity in Figure 2.
Figure 6 shows the corresponding inverse system response, which may be convolved with the raw data
record to produce calibrated data. The lter taps in Figure 6 carry units of
pT
raw .
Unwrapped Phase
[deg]
Magnitude
[dB-raw/pT]
Time [usec]
Frequency [Hz]
Time [usec]
Unwrapped Phase
[deg]
Magnitude
[dB-pT/raw]
Pre-filter
Frequency [Hz]