fast camera image analysis - columbia...
TRANSCRIPT
CTX
Fast Camera Image Analysis
B. A. Grierson9.28.2008
CTX
Introduction
A high speed camera is used to image the visible light in CTX. The camera is begins recording at the specified trigger time, and records until the data buffer is filled. The images are mapped to device coordinates. Various ‘movies’ of light fluctuations can be made by simple signal analysis in space and time.
Statistics of the light fluctuations are calculated, and correlations with other diagnostics are performed.
In real device coordinates, the azimuthal wavenumber spectrum is calculated. It is found that very little power exists for small scale fluctuations.
CTX
frame 3frame 2
Camera
• Phantom v7.1 High Speed Camera
• Frame rate used was 10,000 fps. 10kHz Sampling Rate. 5kHz Nyquist Frequency.
• Images stored as ‘.tif ’ files.
frame 1
x
y
t
Each image is a grid, and filtering can be done in x, y or t.
CTX
Total Light in Time
Single Frame
Fast C amera Light Intensity
0.25 0.30 0.35T ime (s)
5 .0•10 5
1 .0•10 6
1 .5•10 6
2 .0•10 6
Tot
al L
ight
•Total light recorded in time.•The large amplitude, low frequency fluctuations are attributed to magnet ripple.
•The higher frequency fluctuations peak near 1.5=2.0 kHz and are plasma fluctuations associated with ExB rotation of m=1,2 density perturbations.
CTX
Light and Density Low Frequency Fluctuations are Correlated
S ignal (AU )
0.25 0.30 0.35T ime (s)
0 .0
0 .1
0 .2
0 .3
0 .4
dens
ity:r
aw
Light Intensity
0.25 0.30 0.35T ime (s)
5 .0•10 5
1 .0•10 6
1 .5•10 6
2 .0•10 6
!(li
ght)
S moothed density and intensity
0.25 0.30 0.35-1 .0
-0 .5
0 .0
0 .5
1 .0
Density
CTX
What About the ‘High’ Frequency?
• We know that density and polar loss are well correlated, with zero lag time.
• We will see that light is well correlated with density with zero lag time, as well as polar loss ‘in the neighborhood’ of the camera.
CTX
High Frequency Density and Light are Well Correlated
• This confirms two things.
• One: density fluctuations and light fluctuations are well correlated.
• Two: The zero-lag of the two signals means we can ‘sync’ the camera with other diagnostics.
-0 .70
-0 .37
-0 .034
0.30
0.63
0 .25 0 .30 0.35T ime (s)
-4-2024
! (m
s)
density:raw-Light C ross-C orre la tion Function
<D ensity-Light C ross-C orre la tion Function> t
-4 -2 0 2 4! (ms)
-0 .4
-0 .2
0 .0
0 .2
0 .4
Cor
rela
tion
-0 .75
-0 .40
-0 .037
0.32
0.68
0 .25 0 .30 0.35T ime (s)
-4-2024
! (m
s)A14 52:R AW -Light C ross-C orre la tion Function
<D ensity-Light C ross-C orre la tion Function> t
-4 -2 0 2 4! (ms)
-0 .4
-0 .2
0 .0
0 .2
0 .4
Cor
rela
tion
CTX
Mapping to Real CTX Coordinates
• A view of the magnet cap at L=25cm is used to ‘match’ the image to real coordinates.
• Certain features are not plasma fluctuations. -50 -40 -30 -20 -10
x (cm)
10
20
30
40
50
60
y (c
m)
Angle
1520
2530
35
35
40
40
45
45
49
49
54
54
59
59
64
64
69
69
747984
View obstructed by port-hole screen to shield from microwave interference.
Hot tungsten filament.
Horizontal noise line: didn’t do a session
reference for the camera.
CTX
Light Amplitude at a Single Radius
• The light amplitude as a function of angle is calculated by selecting pixels which exist on a arc through the grid of pixels.
• Large scale features and trends as well as small scale fluctuations are clearly seen in the light intensity.
• This light intensity is calculated for each frame of the movie, and at multiple radial locations.
Angle
2025
3035
4045
4954
54
59
59
64
64
69
69
7479
Light Amplitude At 35 cm
100 120 140 160Azimuthal Angle ! (deg)
020
40
60
80
100
Ligh
t Am
plitu
de (
AU
)
CTX
Mean Light Amplitude at a Single Radius
• The mean light as a function of angle for 35cm radius is calculated for 500 frames.
• The individual amplitudes are also stored.
• The deviation from this mean should reveal the fluctuations which we care about!
M ean Light Amplitude At 35 cm
100 120 140 160Azimuthal Angle ! (deg)
10
20
30
40
50
Ligh
t Apl
itude
(A
U)
CTX
Light Fluctuations at a Single Radius From One Frame
• Because the angle axis is non-uniformly gridded, it is interpolated to a regularly spaced grid, allowing the FFT to be used.
• It is found that there is always a strong low wavenumber component, and a broad higher wavenumber spectrum.
• To watch the light fluctuations in an animation typically shows a large fraction of one period of a sinusoid, or a ‘see-saw’ type motion.
F luctuation Light Amplitude
100 120 140 160Azimuthal Angle ! (deg)
-6-4-202468
Ligh
t Am
plitu
de (
AU
)
k-S pectrum
0 50 100 150 200 250 300k (m -1)
0 .1
0 .2
0 .3
0 .4
0 .5
Spe
ctra
l Apl
itude
(A
U)
CTX
Wavenumber Spectrum in Time
• Because the angle axis is non uniformly gridded, it is interpolated to a regularly spaced grid, allowing the FFT to be used.
• It is found that there is always a strong low wavenumber component, and a broad higher wavenumber spectrum.
• To watch the light fluctuations in an animation typically shows a large fraction of one period of a sinusoid, or a ‘see-saw’ type motion.
0.306 0.308 0.310 0.312 0.314T ime (s)
0
50
100
150
200
250
300
k (m
-1)
k-spectrum in time
0.306 0.308 0.310 0.312 0.314T ime (s)
0
10
20
30
40
k (m
-1)
k-spectrum in time
CTX
Summary
• These images were recorded with a ‘threshold’ frame rate, i.e. poor signal/noise ratio.
• We didn’t do a session reference.
• We had physical objects obscuring the view.
• No filtering in time or space has been done for this data.
• However, we can still get useful information from the images.
CTX
Summary II
• With time-filtering the images by high-pass filtering the time signature of each pixel, we can more easily see the plasma dynamics (already done).
• With bench testing, we can determine the spatial noise fluctuation level. This can justify any spatial filtering.
• With improved microwave interference reduction, eliminating the obstructions in the views and justifiable spatial filtering of the images, we should be able to go to higher frame rates.
• At higher frame rates, we should be able to see structures as they move through the view.