Casio EXILIM cameras
Theory
Basic parameters
The Casio EX-F1 is a commercially available camera with high speed imaging capabilities.
The camera is equipped with a variable objective with an anti-reflexive coating. Entrance pupil size is ø4.5 cm, the viewing angle is variable from 9.2° to 88° with the corresponding maximal numerical aperture from 0.11 to 0.18.
The camera contains a 1/1.8” CMOS imaging chip IMX017CGE with a maximum 6.2 Mpix resolution and 2.5µm square pixels.
- Operation regimes: High speed video mode - 336×96@1200 fps, 432×192@600 fps, 512×384@300 fps
- Shutter speed: up to 1/40000 s
- Sensitivity: ISO 100 - 1600
Minimal sensitivity is 4200e− at color 3200 K, flux 700 cd/m2 , exposure 1/50 and F-number 5.6
Rolling shutter mode:
it is possible to use the so-called rolling shutter effect to increase the time resolution up to the effective 40 000 fps with a resolution of 336×1 pixels (one row of CMOS chip). The principle of the rolling shutter1 is shown in Fig. 1. Each pixel raw from the CMOS sensor is read out separately by parallel readout circuits. The time shift of the electronic shutter in consecutive rows is given by the readout time. The readout times of individual rows cannot be overlapped because they are read by the same controller. Furthermore, when the last row of CMOS chip is reached, there is a time lag equal to 16% of the frame time, which is needed for processing of each full image (see Fig. 1). Then, the readout cycle can begin again starting from the first row.
Fig 1: Principle of readout sequence for CMOS chips
Data format
Data are saved as 8bit movie with high compression level with mp4 algorithm. Intensity is saved nonlinearly in order to extent dynamic range. \[ y = x / (bx + a) \] where \(x\) is stored intensity and \(y\) is original pixel intensity and \(a = 43000 \pm 600\), \(b = 140\pm3\). The script for correction is available here
Fig 2: Measure nonlinearity of saved intensity
Spectral sensitivity
Color sensitivity is approximately corresponding to human eye in the case of “sunny day” color setting. Measured relative sensitivity in in Fig. 3.
Fig 3: Measured relative spectral sensitivity of camera
Experimental setting
Two cameras are placed on perpendicular diagnostic ports as it is shown in Fig. 4. Orientation of the cameras is set so that the horizontal rows of the CMOS chip were rotated perpendicularly to the toroidal magnetic field direction in order to allow exploitation of the rolling shutter effect.
Fig 4: Experimental setting for tomographic reconstruction
Remote control
It is important to have remote control of camera for most of the advance applications. The camera EX-F1 can by remotely controlled in 3 different ways. The simplest way is a remote trigger, but only single or burst shooting can be triggered and camera settings cannot be set remotely. The second way is via free official remote control software [5] released by Casio. It is user friendly software with graphical interface, hence unsuitable for automated measurement controlled by scripts. And the last way is unofficial open source program [10] based on reverse engineering of Casio official control software. All of the camera parameters can be set remotely before the data acquisition and after the recording data are transferred over a USB to the computer, here are post-processed.
Calibration and plasma position
The focal length of the camera on LFS was set to 144 mm and focused at distance 42 cm in the center of the chamber (Fig. 4). The upper camera was set to focal length was 120 mm and objective was focused at 32 cm. The light intensity of the plasma observed from both cameras was compared and upper camera had 19% higher intensity. The aperture of the cameras were fully opened on both cameras, signal gain was set to ISO 800, exposure to 1/40000 s, white balance to daylight and HS movie mode 1200 fps. Cameras were triggered by signal from a computer, but accuracy of this triggering is worse than 10 ms.
For exact synchronization with the global trigger should be used 0.5 ms long rectangular pulse from high power LED diodes, because LED diode will achieve full brightness in under a microsecond.
Fig 5: An example of tomographic reconstruction
Other improvements
Before further analysis of the images taken by the HS camera they have to be significantly preprocessed. Three basic issues had to be solved:
- Reflections from background.
- Gap time during the image readout.
- Missing time synchronization with the global trigger
The spatial profile of plasma light reflected from the tokamak chamber is strongly non-uniform as can be seen in the top image of Fig. 4. Increase of the camera signal at the beginning and end of each frame is caused by the fact that reflectivity of tokamak chamber segments along the lines of sight tangential to the port edges is significantly stronger compared to the section viewed by the central lines of sight. We employed Singular Value Decomposition (SVD) to remove these artifacts in the signal due to non-homogeneous reflections. Each column in the decomposed matrix corresponds to one frame. After the decomposition, any time evolution is removed from the first two “topos“ vectors and also these two vectors are smoothed in the spatial direction. The projection of these vectors is subtracted from the original image and the resulting set of images is once more decomposed. This time, the first SVD vectors contain a majority of the reflection patterns, and therefore, the first few vectors are smoothed and the original image without the reflections is recovered.
Figure 6: An example of original (top) and preprocessed (bottom) image sequence from discharge #7251 (discharge in Helium) obtained by the LFS camera (Fig. 4). The image is in real color.
In the second step, the missing gaps have to be calculated. Therefore, the images are once more decomposed using the SVD. However, this time the vectors of the decomposed matrix are columns of each image. The topos vectors and their time evolutions (chronos) are obtained from the SVD. The fast Fourier transformation for incomplete data3 is used to interpolate the quazi-periodic chronos in the missing gaps. The Fourier transformation is applied on 0.1 ms windows around each gap and then the ordinary inverse Fourier transformation is computed and the missing time evolution (chronos) is estimated. The gaps have to be filled in order to increase the number of time slices suitable for the reconstruction by up to 32% in the case of the two cameras with unsynchronized video (gaps).
The Matlab script is available here
EXF1 Control software
Modified version of open source remote control software - exf1ctrl. Added small fix for 64bit architecture, linux support, high speed video improvements, better remote zoom/focus.
Simple example of use:
Firmware rev. 2.00 is required and the camera must be put in remote control mode before being connected to the host. Update camera firmware and try original software at first link (firmware + drivers).
Use of LInux command line version:
Compile:
“Manual” USB control :
where is 0 for 1th camera in your USB port …
Scriptable “automatic” calling
N_cam=`usb-devices | grep "Vendor=07cf ProdID=1023" | wc -l`
touch "command_file"
tail -f "command_file" | ./ExF1Ctrl 0 >> LOG &
echo "S 0.2 0.1" >> "command_file" # setup camera focus to 20% and zoom to 10% of maximum
echo "m 10 <video file name>" >> "command_file" # "trigger", start 10s record of HS video to <video file name>.MOV file
echo -e "q\n" >> "command_file" # correctly disconnect camera Convert the HS video:
Other commands:
Hint: a [x] sets aperture (x = 1-10). 1: F2.7 (default). 2: F3.0. 3: F3.3. 4: F3.8. 5: F4.2. 6: F4.7. 7: F5.3. 8: F6.0. 9: F6.7. 10: F7.5.
Hint: c [x] sets mode / movie mode (x = 1-9). 1: Single shot (default). 2: Continuous shutter. 3: Prerecord still image. 4: Movie (STD). 5: Prerecord movie (STD). 6: Movie (HD). 7: Prerecord movie (HD). 8: Movie (HS). 9: Prerecord movie (HS).
Hint: e [x] sets exposure (x = 1-4). 1: M. 2: Auto (default). 3: A. 4: S.
Hint: f [x] sets focus (x = 1-4). 1: Auto (default). 2: Macro. 3: Infinity. 4: Manual.
Hint: h activates half press.
Hint: i [x] sets iso (x = 1-6). 1: Auto (default). 2: 100. 3: 200. 4: 400. 5: 800. 6: 1600.
Hint: q quits this program.
Hint: m [x [y]] records a x second long movie called y.
Hint: s [x [y [z]]] activates shutter and stores a picture called x and a thumbnail called y. If the continuous shutter is enabled (modes 2 and 3), z determines the shutter duration.
Hint: v [x [y]] focus y steps. x=in focuses in and x=out focuses out. Continuous focus is used if y is not defined.
Hint: z [x [y]] zooms y steps. x=in zooms in and x=out zooms out. Continuous zoom is used if y is not defined.
Hint: S [x [y]] focus to x*100% of maximum and zoom to y*100% of maximum + setup the HS video to the preferred settings
You can change S mode settings in exf1api.c, line 175, in function setup_HS_movie
Other utilities used on GOLEM tokamak:
- CAMERA_INIT.sh - auto initialization code
- CAMERA_START.sh - bash control software
- CAMERA_TOMO.sh - initialization of tomographic reconstruction
