ESCENARIO 1: TRANSICIÓN DE RIESGO NORMAL A RIESGO ALTO

n s O 8 0 2 4 6 N um ber o f filte rs

Figure 5.7. Output power from the integrating sphere as a function of the number of optical filters used to attenuate the diffuse light distribution.

5.2.2. Experimental validation of the digitisation param eters

5.2.2.1. N on-uniform ity o f the video signal

In order to verify the values for the digitisation param eters three cam eras, the Photon EEV cam era and two other norm al video cam eras, all conform ing to the CC IR standard and from different m anufacturers, were com pared. The video signals were first m easured using an oscilloscope and subsequently a series o f im ages w ere taken. All A D conversion param eters apart from the digitisation rate and form at m em ory

were kept to their default values = 63, = 1). At low levels of illum ination

the EEV cam era was observed to give non-uniform output as opposed to the other two cam eras, w hich show ed good uniform ity.

Figure 5.8 show s a plot o f the horizontal profiles corresponding to zero input (dark) im ages for the three cam eras tested. An arbitrary offset w as added to the data to avoid overlapping betw een the profiles. Since all the im ages w ere taken w ithout illum inating the sensor and all cam eras produce the sam e (standard) output, the data indicated that the non-uniform ity observed in the EEV cam era was due to the readout electronics, and was not an effect o f the digitisation procedure. C onsequently any quantitative m easurem ents done with this cam era had to take into account any non­ uniform ity effects.

Chapter 5 A CCD-based x-ray imaging system

EEV

Camera 2

Camera 3

0 64 128 192 256 320 384

H o rizo n ta l p o s itio n [pixels]

Figure 5.8. Horizontal profiles o f dark im ages taken with 3 different CCIR cameras

under identical digitisation conditions. The data indicate a non-uniformity in the output o f the EEV camera due to the readout electronics.

5.2.2.2. Analysis o f non-uniformities by Fourier methods

The presence of spatial non-uniformities in the image data was investigated using Fourier techniques. It is well known that some artefacts present in the digitised image can occur due to the interaction between the readout electronics of the camera and the signal acquisition electronics (Boreman 1987).

The conversion of continuous scene data into an array of digitised values involves a complex sequence of sample-and-hold operations. Differences between the pixel readout clock and the ADC sampling rate can give rise to spurious frequency content on the digitised image. Figure 5.9a shows an image of a uniformly illuminated field that was taken using a sampling rate of 16 MHz. A beating frequency arising from the asynchronicity between the pixel readout clock and the input clock that controls the digitisation process can be clearly seen. It is interesting to note that non-uniformities are present only in the raster (horizontal) direction, while the image has good uniformity in the vertical direction.

The information given by the spatial frequency content of the digitised images can be used to determine the correct value for the sampling frequency of the ADC without knowing the exact time format of the video waveform. A series of images was taken under the same uniform illumination conditions using different sampling frequencies.

Chapter 5 A CCD-based x-ray imaging system

F igure 5.9b show s the corresponding horizontal profiles for im ages taken at 15, 16, 17 and 18 M H z. An arbitrary offset o f 20, 40 and 60 AD U was added to the profiles taken at 16, 17 and 18 M H z in order to avoid overlapping. It can be seen that an increase on the digitisation frequency also increases the frequency o f the beats.

200 180 C3 160 140 120 100 0 100 200 300 400 (a) H o r iz o n ta l p o s itio n [ p i x e l s ] (b)

Figure 5.9. (a) Image of a uniformly illuminated field using a digitisation frequency of 16 MHz; notice the beating frequency in the raster direction, (b) Horizontal profiles o f images taken at different digitisation frequencies.

Estim ates o f the spatial-frequency content o f the video signal in the raster direction were obtained by taking the D iscrete Fourier Transform (DFT) of the profiles show n in figure 5.9b. The results o f the calculation are shown in figure 5.10a, where the m odulus squared o f the Fourier transform s are plotted as a function o f spatial frequency for all four im age profiles. The spikes in the spectra correspond to the harm onic com ponent o f the beating frequency present at different sam pling rates, and are indicated with an arrow for each o f the four different sam pling frequencies used. Figure 5.10b show s a plot o f the spatial frequency o f the beats, determ ined from the pow er spectrum estim ates, as a function o f the fram e grabber sam pling frequency. L inear regression analysis o f the data gives

A ,., [m m " ] = 2.38 ■ f ,s a m p l. [M hz] - 35.06 (5.2)

The correct sam pling rate is given by the frequency for w hich = 0, that is, by the

interception o f the regression line with the sam pling frequency axis. Therefore, using the result from the linear fit given by equation 5.2 the correct sam pling frequency is 14.73 M Hz. This value agrees to within 0.3% with the sam pling frequency calculated

Chapter 5 A CCD-based x-ray imaging system

in §5.1.3. It was observed that variations o f ±0.1 M H z around this value had no effect in the output signal. Furtherm ore, it was noticed that dark signal and background n on­ uniform ity depend on the particular CCD being used.

3 3 2 5 0 -10 -8 -6 -4 -2 0 4 6 S patial freq u en cy [mm ] 8 10 14 15 16 17 18 19

Sam pling frequ en cy [MHz] Figure 5.10. (a) Modulus of the DFT of the profiles shown in figure 5.9b. The

spikes correspond to the harmonic component of the beating frequency present at different sampling rates, (b) Beat frequency as a function of digitisation frequency in the ADC.

T hese results indicate that non-uniform ities in the raster direction are related to the video circuitry o f the cam era, and that they are not artefacts introduced by the digitisation process. A lthough it is possible, in principle, to adjust the cam era electronics to m inim ise these effects, a calibration procedure o f this kind requires specialised equipm ent. Besides, the nature of this assessm ent in which several CC D s w ere being tested using a single cam era m ade this procedure im practicable.

5.2.3. Frame grabber linearity and ADC resolution

5.2.3.1. Gain and A D -reference linearity

The experim ental setup described in §5.2 was used to investigate the effects that changes in the fram e grabber gain and/or A D -reference w ould have on the m easured signal. For the gain linearity m easurem ents the A D -reference value was set to 63, which corresponds to the default setup configuration after fram e grabber initialisation, and the gain was set to its m inim um value (0.5). Starting from an illum ination level sufficient to saturate the CC D the output pow er was reduced until the m easured signal

Chapter 5 A CCD-based x-ray imaging system