Investigation of Improvement of Pellet Tracking System

(1)

15002

Examensarbete 15 hp

Juni 2015

Investigation of Improvement

of Pellet Tracking System

(2)

Teknisk- naturvetenskaplig fakultet UTH-enheten Besöksadress: Ångströmlaboratoriet Lägerhyddsvägen 1 Hus 4, Plan 0 Postadress: Box 536 751 21 Uppsala Telefon: 018 – 471 30 03 Telefax: 018 – 471 30 00 Hemsida: http://www.teknat.uu.se/student

Abstract

Investigation of Improvement of Pellet Tracking

System

Sanne Torgersen, Adéle Wallin

A pellet target is an internal target system for accelerator experiments in nuclear and particle physics. The target consist of small spheres of frozen hydrogen, called pellets. The pellets interact with high energy accelerator-beam particles in a particle accelerator. The challenge is to track these pellets for good accuracy of interaction position in both time and space. The pellets are tracked with lasers and cameras. The main goal of this project was to develop a method to find the best time resolution and to optimize the efficiency of the pellet tracking system. This project addresses challenges with making trustworthy measurements were stability in the setup, difficulties with alignment and

optimizing of exposure cycles. Because of stability issues, a more stable and robust module that also will ease adjustment of alignment is under construction. A well-aligned setup can be confirmed in two ways. Firstly by confirming that the cameras detect pellet signals for a about 300 micron-height change of the laseror the camera and secondly that a focus interval is within 100 micron. A mathematical model that calculates class-distributions for varying exposure cycles and shifts can be used to predict the results from a measurement with pellets. The model can be used for testing an appropriate exposure cycle before running it with pellets. In addition, an appropriate laser power should be chosen since the effects of both too low or too high power gives misleading results. This project has contributed to progress in the field of pellet tracking.

(3)

1 Populärvetenskaplig sammanfattning

(5)

2 Introduction

2.1 Background

This project is about an optical tracking system for pellet targets. A pellet target generator is an internal target system for accelerator experiments in nuclear and particle physics. The target system consists of a generator where a jet of liquid hydrogen is broken up into droplets with a diameter of about 20 − 35 µm by a nozzle vibrating at a xed frequency in the range of 40 − 100 kHz. The pellets freeze by evaporation in a droplet chamber and form a beam of pellets that are injected in vacuum via a 7 cm long capillary. After collimation, the pellets are directed through a thin pipe (φ ∼ 5 mm) of several meters through the detector system of the experiment into the scattering chamber and further down to a pellet beam dump. In the scattering chamber they interact with a high energy accelerator beam of e.g. protons, anti-protons or deuterons.[1],[2] _{The technique is used in experiments in Germany}

and is planned to be used in the antiproton storage ring HESR. An optical tracking system for the pellets that would allow a precise three dimensional determination of the interaction-point between the pellet and an accelerator-beam particle is being developed within this project at IFA1 _{and TSL}2 _{in Uppsala.}

2.2 Problem and task

To achieve a precise three dimensional determination of the interaction point between a pellet and an accelerator beam particle the pellet tracking system requires a high resolution in both time and space. With the presently available line-scan (LS) cameras the repetition frequency of the exposure cycles is around 100 kHz which is almost enough to reach the desired resolution. There is also a dead time of about 3 µs between the exposures.

The main goal of this project was to develop a method to improve the time resolution of the pellet tracking system and to increase the eciency by making the same observation with two cameras with shifted exposures.

2.3 Theory

The pellet tracking system uses LS-cameras with a sub-millimeter thin line of pixels to measure position of individual pellets at a certain time. At one level the pellets are illuminated by two lasers from dierent directions and the reected light is detected by two cameras opposite each other. A set of cameras at dierent levels can be used to reconstruct direction and velocity of the pellets. This requires that the cameras are well aligned and work in a synchronized way so that time dierences can be measured eciently and precisely.

The read-out system is based on commercial frame-grabbers and external synchronization. In the read-out system each picture corresponds to one line. One line corresponds therefore to the time of one exposure cycle. The intensity of any pellet signal is analyzed as a light integral. This is the sum of the amplitudes of all the signals recognized as coming from one pellet.

2.4 Method

2.4.1 The setup

The alignment of the equipment was investigated at the Uppsala Pellet Test Station for both pellet runs and with a bench test setup based on a pulsed LED/laser illuminating a shing line, see gure 1. In the pellet tracking system there were two cameras opposite each other, two lasers with a continuous light and a pellet beam with diusion in both position and speed. In the bench test setup a known pulse simulated the pellets. The setup using a LED was mainly used to develop an understanding and skill for handling the equipment. Later the LED was replaced by a pulsed laser with a suitable intensity and a more narrow beam of light.

The bench test setup was used study the read-out performance with respect to pulse length, focus of cameras and laser, laser power and electronic eects. Moreover the read-out performance was studied with dierent exposure cycles and shifts for both pellet runs and with the bench test setup, see gure 2.

(6)

Figure 1: The bench test setup with two LS-cameras, a LED and a shing line.

Figure 2: Visualization of two synchronized cameras with a time shift.

2.4.2 Exposure cycle

Dierent exposure cycles were investigated with dierent cycle time, exposure time and shift. The read-out system classies every signal according to when it was seen by the other camera, e.g. when a signal is seen by camera B in its nth_{cycle and by camera A in its n}th_{+ 1}_{cycle it's classied as seen by camera B when seen in next line}

by camera A and similarly seen by camera A when seen in the previous line by camera B. A mathematical model based on the geometric relationship between the exposure cycles and their relative shift was developed to calculate the distribution for each class. Measurements with varying shift were made at a pellet run for a series dierent exposure cycles. The cycles that were used were exposure cycles of 20 µs with exposure time of 15 µs and 10 µs and exposure cycles of 13 µs with exposure time of 8.5 µs, 7 µs and 5.5 µs. The model was then adjusted to t the measured values. Finally the same measurements were made for the bench test setup with the aim to reproduce the pellet data.

2.4.3 Pulse length

Dierent pulse lengths to the LED/laser input were used to investigate the read-out response to pellets of dierent size, speed and eects due to overlapping of dead time.

2.4.4 Focus

To investigate the dependence of alignment and focus the position of the laser and the cameras were varied. First the height of the laser was varied through the range in which any signal was observed by the viewing interface and the light signals were measured. Then measurements were made where the height of one camera was varied in the same way. The last measurements were made where the distance between one camera and the shing line was varied through the range in which any signal was observed.

2.4.5 Laser power

The laser power was varied to investigate the read-out response. 2.4.6 Multiple shing lines

(7)

be calculated how many micrometers corresponds to one pixel. The reason for this test was to investigate optical and electronic eects. Measurements were made with two shing lines with diameters of 120 µm and separation distance 400 µm, 900 µm and 1500 µm and three amplitudes of the LED input (3 V , 5 V and 7 V ). Additionally, one 120 µm-shing line was replaced with a shing line with a 200 µm diameter and measurement were made for the same distances and the greatest amplitude (7 V ).

(8)

3 Results

3.1 Exposure cycle

Measured values from pellet runs with dierent exposure cycles were plotted against the delay and compared to a model that calculates the distribution of entries in each class. The model was implemented in MATLAB and the code is given in appendix A. The model was developed in the following way. It was based on geometric assumptions. After being compared with measured data from pellet runs, the model was adjusted to better t the data with respect to assumed electronic eects. The model took the pulse length and a minimum time a signal needs to be detected into account. The minimun time, (min.signal) corresponds to the cut of the light integral. The read-out system treats a signal with too small light integral as noise and a longer signal gives a greater light integral than a shorter (with the same amplitude). An example with cycle time of 13 µs and an exposure time of 8.5 µs is shown in gure 4. The model was then compared to measurements from the bench test setup and a known pulse length to the laser input. The values from dierent exposure cycles with both pellets and the bench test setup were compared and are shown in gure 5 and 6. Graphs for some other exposure cycles are found in appendix B.

Figure 4: Measured values from a pellet run with cycle time 13 µs and exposure time 8.5 µs and corresponding calculated/ap-proximated values for a pulse length of 3 µs and a minimun signal time of 0.25 µs.

(9)

Figure 6: Measured pellet and shing line run with cycle time of 13 µs, exposure time of 8.5 µs and pulse length 2.5 µs and calculated approximation with pulse length 3 µs and a minimun signal time of 0.25 µs.

3.2 Pulse length

Figure 7 shows the light integral versus the pulse lentgh and increases linearly with increasing pulse length. When a signal overlaps the dead time it is shown in two following exposure cycles (or lines). When a signal is detected in two following exposure cycles it is interpreted as two pellets and presented as one line between these pellets. The number of signals shown in two following lines is increasing with increasing pulse length. This is shown in gure 8 where there seems to be a threshold for a pulse length of 1.5 µs. The measurements were made for an exposure cycle of 12.5 µs with an exposure time of 10 µs.

(10)

Figure 8: Percentage of signals shown in two exposure cycles versus pulse length.

3.3 Focus

The setup was very delicate and dependent of being well-focused and well-aligned. To investigate the read-out response dependence of focus, rst the height of the laser was varied. Figure 9 shows how the camera signal intensity varies with the height of the laser. The width of the horizontal part is about 100 µm, which corresponds to the focus interval.

Then the height of camera A and the distance to the shing line was varied. Figure 10 shows how the intensity depend on the height of camera A and gure 11 shows how the intensity depend on a change in distance between camera A and the shing line. In both diagrams camera A detects signals within an interval of 300 µm in both height and distance when other parameters are optimized. There is also a peak in both diagrams corresponding to the setting for best focus or alignment. In addition, eects of the instability of the setup can also be noted in gure 11. Since the only parameter that was changed was the distance between camera A and the shing line and the intensity for camera B should therefore barely change we assume the decrease in intensity for camera B is due to soft bumps to the bench causing camera B to move enough for these eects.

(11)

Figure 10: Light integral versus the height of camera A.

Figure 11: Light integral versus the distance of camera A to the shing line

3.4 Laser power

The eect of varying the power of the laser input was that a power less than 3.3 mW was too weak and barely gave any signal. A power greater than 4.2 mW was too strong and gave an unreadable response. The light integral versus the power of the laser is shown in gure 12.

(12)

3.5 Multiple shing lines

The read-out of light signals gives position, signal cluster width and light integral for all entries. In tests with two shing lines the position occurs like to spikes in the histogram where the position is shown on the x-axis and the number of entries is shown on the y-axis. The known separation distance between the shing lines and the distance between the spikes where used to calculate how many micrometers corresponds to one pixel. Then for shing lines with a known diameter of 120 µm the expected signal cluster width was calculated and compared to the observed width.

The test was made for separation distance of 400 µm, 900 µm and 1500 µm and amplitudes of the LED input of 3 V, 5 V and 7 V . The measurement with a separation distance of 900 µm and the amplitude of the LED input of 7 V is shown in gure 13. In this case one pixel corresponded to about 30 µm, the expected width was 4 pixels and the observed width was 10 pixels. For the other measurements one pixel seems to correspond to 20 − 30 µm/pxl. This is not consistent resultsa any may be due to non-homogeneous illumination in the shing lines. In addition, other tests have been done earlier that show that one pixel corresponds to about 35 µm. However, for every measurement the observed signal cluster width is much greater than the expected. It was also found that the interpreted width increased with increased amplitude. The measured and expected width for dierent input-amplitudes are shown in gure 14. This is partly due to that leakage of signal to neighbouring pixels.

When one shing line was replaced for one with a diameter of 200 µm the signal cluster width also exhibits two spikes. One spike for each line with dierent diameter was expected. The result from the measurement with separation distance 400 µm and amplitude of the LED input of 7 V is shown in gure 15. The 200 µm-shing line was placed closer to the LED. With a 200 µm-shing line mentioned inconsistency is bigger (as expected).

(13)

Figure 14: Expected and measured mean of signal cluster width versus distance between shing lines. Measurements were made for dierent amplitudes of the LED input.

(14)

4 Discussion

4.1 Exposure cycle

In theory, when the two cameras are perfectly synchronized they should detect the same number of signals and all entries should be seen in the class same. Additionally, if e.g. camera B is delayed no signals should be seen in the class A in next line until the delay is greater that the dead time and then increase linearly. This is not the case though. A reason for this may be that the signal has a signicant duration compared to the dead time and if a sucient big part of the signal appears in the exposure time it will be measured anyway. Another way to look at this is as if the dead time is shortened. The dead time is shortened in a way that can be thought of as on both sides, so that some of the signals that are centered in the dead time region partly appear in the same exposure cycle as the reference camera and some of the signals partly appear in the next exposure cycle. Furthermore, a signal needs to appear in the exposure time for a minimum time to be detected at all. The model we developed takes both this requirement of a minimum time a signal need to be seen, the eect of signals centered in the dead time appearing in the exposure time and a normalization factor into account when calculating the distribution. With a minimum time about 0.25 µs the model seemed to t the pellet-data quite well but not as well for the shing-line-measurements. The model t the shing-line-data much better with a 0.5 µs longer pulse as input. It is suspected that due to electronic eects a longer pulse is sent out from the pulse-generator in the latter case. After this realization the model was adjusted. The minimum time value of 0.25 µs is a value that t our data but since parameter values and cuts for the read-out interface can be changed this is not absolute. One cut is the lower limit of the light integral, all signals below this limit will be treated as noise and cut out from the analysis. In addition, neither the changeover between exposure and dead time nor the square wave is as sharp as the ideal case. These electronic eects also aect the results.

4.2 Pulse length

The dependence of pulse length was measured to investigate the read-out response to pellets of dierent size and eects due to signals overlapping the dead time. As seen in gure 7, the light integral increases linearly with increasing pulse length. Since the light integral is the sum of the amplitude of the signals in all pixels interpreted as one pellet it makes sense that a longer pulse results in a greater integral. Since the pellet beam is illuminated by a continuous laser beam a longer pulse is analogue to a bigger pellet. In this way it is plausible to calculate the size of a pellet.

As shown in gure 8, the percentage of signals shown in two following exposure cycles are shown. In the bench test setup we use a very low frequency of about 2 kHz corresponding to 40 lines between each signal. The result implies therefore that the same signal is shown in two exposure cycles and the only way this is possible is when the signal overlaps the dead time. Results like this could be used to study electronic eects such as how long the dead time really is when we set it to a controlled value.

Another observation that was made from the pulse length experiment was that camera B had a greater light integral in gure 7. This might be due to non-perfect alignment or if the laser beam was not perfectly orthogonal to the cameras line of view. If the laser was somewhat more directed from the same side of the setup as camera B a greater amount of light would be reected towards camera B than towards camera A and could explain this dierence.

4.3 Focus

From gure 9 and 10 we see that the focus interval for the camera is about 100 µm in height and the signal is visible in an interval of 300 µm. The shape of the curve is symmetric with maximum values on a plateau in the middle. To get the best alignment the camera should therefore be in the middle of the interval where the signals are visible. Since the LS-camera uses pictures of only one pixel in height and one pixel corresponds to about 30 µm it seems like the laser beam is about 250 µm wide to be shown for an interval of 300 µm, this also makes sense as a comparison to what we could see with our own eyes (using protection glasses).

(15)

instabilities of the setup which is a huge problem. One thing that could be done to overcome this challenge is to develop a new more sound and robust construction that would be easier to adjust and less sensitive to vibrations. A module just like this is under construction at TSL, Uppsala.

4.4 Laser power

For the shing-line setup, the most appropriate power of the laser is 4 mW . When using a very low power, signals were not seen at all or cut out as noise. When using higher but still low powers reections in the shing line was interpreted as two parallel shing lines. This was solved by using a higher power. A disadvantage with a too high power is that the light integral gets unreadable.

4.5 Measurements with two shing lines

Since we could not nd out how many micrometers one pixel corresponds to in the measurements, it is possible that there is a electronical leakage in pixels or that the camera for another reason interprets around the double amount of pixels than it should. To get better results with multiple shing lines several LEDs/lasers could be used so that every shing line get the same amount of light. One consequence with several LEDs/lasers is that the alignment of the setup gets more complicated.

5 Conclusion

The main goal of this project was to develop a method to improve the time resolution and eciency of the pellet tracking system. Challenges with making trustworthy measurements were stability in the setup, diculties with alignment and optimizing of exposure cycles.

(16)

6 References

1. Personal communication with Calén, H., Supervisor, Uppsala University, Department of Physics and Astron-omy, division for Nuclear Physics

(17)

Appendix A

MATLAB-code

MATLAB was used to develop the matematical model for the distribution of entries in the classes ABsm, BnxA and BnoA.

matdata.m

This le loads and plots the values from a chosen measurement (pellet och bench test setup) against the delay. The loaded text les are built up in columns with the delay, the classes ABsm/B, BnxA/B, BnoA/B in the rst, second, third and fourth column respectively.

c l e a r a l l

% f i l e = '150506_C13_E10 . txt ' ; % Sparse measurements % f i l e = '150506_C20_E15 . txt ' ; % Sparse measurements % f i l e = '150506_C20_E10 . txt ' ; % Sparse measurements % f i l e = '150508_C13_E5 . 5 . txt ' ; % dead>exp % f i l e = '150508_C13_E7. txt ' ; % f i l e = '150508_C13_E8 . 5 . txt ' ; % f i l e = '150518_C13_E10_W2 . 5 . txt ' ; % f i l e = '150519_C13_E8. 5_W4. 5 . txt ' ; f i l e = '150526_C13_E7_W4. 5 . txt ' ; comma2point_overwrite ( f i l e ) A=load ( f i l e ) ; cyc = 1 3 ; %c y c l e time exp = 7 ; %exposure pulse = 4 . 5 ; %pulse length f i g = 1 ; u = 200; %r e s o l u t i o n Delay = A( : , 1 ) ; ABsm = A( : , 2 ) ; BnxA = A( : , 3 ) ; BnoA = A( : , 4 ) ; norm = ABsm(1)+BnxA(1);%+BnoA ( 1 ) ; % Divided by t o t a l e n t r i e s in B delay=l i n s p a c e (0 , cyc , u ) ' ; sm=z e r o s (u , 1 ) ; nx=z e r o s (u , 1 ) ; no=z e r o s (u , 1 ) ; f o r k=1:u

[ sm( k ) , nx ( k ) , no ( k ) ] = e n t r i e s ( cyc , exp , pulse , delay ( k ) ) ; end

sm=norm . * sm ; nx=norm . * nx ; no=norm . * no ; s e t ( 0 , ' DefaultAxesFontSize ' , 2 4 )

s e t ( 0 , ' DefaultAxesFontName ' , ' Times New Roman ' ) s e t ( 0 , ' defaultLineLineWidth ' , 2 . 5 ) ;

(18)

p l o t ( Delay ,ABsm, ' b * ' , Delay , BnxA, ' g * ' , Delay , BnoA, ' r * ' , . . . delay , sm , 'b− ' , delay , nx , ' g − ' , delay , no , ' r − '); legend ( 'ABsm − p e l l e t ' , 'BnxA − p e l l e t ' , 'BnoA − p e l l e t ' , . . .

'ABsm − theory ' , 'BnxA − theory ' , 'BnoA − theory ' , . . . ' Location ' , ' e a s t o u t s i d e ' )

g l o b a l min

t i t l e ( [ ' Cycle : ' , num2str ( cyc ) , ' \mus ' . . . ' Exposure : ' , num2str ( exp ) , ' \mus ' . . . ' Pulse : ' , num2str ( pulse ) , ' \mus ' . . . 'Min . s i g n a l : ' , num2str ( min ) , ' \mus ' ] ) ; y l a b e l ( ' Percentage ' )

x l a b e l ( ' Delay [ \ mus ] ' ) xlim ( [ 0 , cyc ] ) ;

ylim ( [ − 0 . 0 0 5 , 1 ] ) ; entries.m

The function entries.m calculates the distribution of entries in each class. f u n c t i o n [ sm , nx , no ] = e n t r i e s ( cyc , exp , pulse , delay ) g l o b a l min

(19)

no = dead/exp ; e l s e i f delay<=exp+y sm = 0 ; nx = ( delay−dead )/ exp ; no = ( cyc−delay )/ exp ; e l s e i f delay>exp+y sm = 0 ; nx = ( delay−dead )/ exp ; no = ( cyc−delay )/ exp ; end end end pellet_vs_shingline.m

This le plots the measured values from a pellet run, the measured values from the bench test up and calculated values against the delay.

% p e l l e t = '150508_C13_E7. txt ' ; % f i s h i n g l i n e = '150526_C13_E7_W2. 5 . txt ' ; p e l l e t = '150508_C13_E8 . 5 . txt ' ; f i s h i n g l i n e = '150519_C13_E8. 5_W2. 5 . txt ' ; % comma2point_overwrite ( p e l l e t ) % comma2point_overwrite ( f i s h l i n e ) A=load ( p e l l e t ) ; B=load ( f i s h i n g l i n e ) ; cyc = 13 ; %c y c l e time exp = 8 . 5 ; %exposure pulse_las = 2 . 5 ;

(20)

[ sm( k ) , nx ( k ) , no ( k ) ] = ent2 ( cyc , exp , pulse , delay ( k ) ) ; end

sm=norm . * sm ; nx=norm . * nx ; no=norm . * no ; s e t ( 0 , ' DefaultAxesFontSize ' , 2 4 )

s e t ( 0 , ' DefaultAxesFontName ' , ' Times New Roman ' ) s e t ( 0 , ' defaultLineLineWidth ' , 2 . 5 ) ;

s e t ( f i g u r e ( 1 ) , ' Position ' , [ 50 50 800 4 0 0 ] ) f i g u r e (1 )

p l o t ( DelayP ,ABsmP, ' b * ' , DelayP ,BnxAP, ' g * ' , DelayP , BnoAP, ' r * ' , . . . DelayF ,ABsmF, ' bo ' , DelayF , BnxAF, ' go ' , DelayF , BnoAF, ' ro ' , . . . delay , sm , 'b− ' , delay , nx , ' g − ' , delay , no , ' r − ');

legend ( 'ABsm − p e l l e t ' , 'BnxA − p e l l e t ' , 'BnoA − p e l l e t ' , . . .

'ABsm − f i s h i n g l i n e ' , 'BnxA − f i s h i n g l i n e ' , 'BnoA − f i s h i n g l i n e ' , . . . ' Location ' , ' e a s t o u t s i d e ' , . . .

'ABsm − theory ' , 'BnxA − theory ' , 'BnoA − theory ' ) g l o b a l min

t i t l e ( [ ' Cycle : ' , num2str ( cyc ) , ' \mus ' . . . ' Exposure : ' , num2str ( exp ) , ' \mus ' . . . ' Pulse l a s e r : ' , num2str ( pulse_las ) , ' \mus ' . . . ' Pulse c a l c : ' , num2str ( pulse ) , ' \mus ' . . . 'Min . s i g n a l : ' , num2str ( min ) , ' \mus ' ] ) ;

y l a b e l ( ' Percentage ' ) x l a b e l ( ' Delay [ \ mus ] ' ) xlim ( [ 0 , cyc ] ) ;

(21)

Appendix B

Diagrams

Measured data and model expectations. Pellet-data

Measured data from pellet-runs compared to the model using pulse length 2.5 µs and 3 µs. By comparing the graphs the eects of dierent exposure time and dierent choice of pulse length are shown.

Figure 16: Measured values from a pellet run with cycle time 13 µs and exposure time 5.5 µs and corresponding calculated/ap-proximated values for a pulse length of 2.5 µs to the left and3 µs to the right.

Figure 17: Measured values from a pellet run with cycle time 13 µs and exposure time 7 µs and corresponding calculated/ap-proximated values for a pulse length of 2.5 µs to the left and3 µs to the right.

(22)

Fishing-line-data

Measured data from bench test setup-runs compared to the model for exposure time 7 µs and 8 µs and pulse length 1.5 µs, 2.5 µs, 3.5 µs and 4.5 µs and the corresponding calculated values for a 0.5 µs longer pulse. By comparing the graphs the eects of dierent exposure time and dierent pulse length are shown.