Available on CMS information server CMS NOTE 1998/16 The Compact Muon Solenoid Experiment CMS Note Mailing address: CMS CERN, CH-1211 GENEVA 23, Switzerland January 1998 Performance test of the first prototype of 2 ways video camera for the Muon Barrel Position Monitor Laurent Brunel Kossuth University, Debrecen, Hungary and CERN, Geneva, Switzerland László Baksay, Tamás Bondár, Péter Raics, József Szabó, Kossuth University, Debrecen, Hungary Gy.L. Bencze KFKI Institute for Particle and Nuclear Physics, Budapest, Hungary and CERN,Geneva,Switzerland Abstract The CMS Barrel Position Monitor is based on 36 video cameras mounted on 36 very stable mechanical structures. One type of camera is used to observe optical sources mounted on the muon chambers. A first prototype was produced to test the main performances. This report gives the experimental results about stability, linearity and temperature effects. The work was supported by the Hungarian National Research Fund OTKA T 1794 and the Hungarian Higher Education Research Fund FKFP 31/1997.
Table of contents : 1. INTRODUCTION... 3 2. CAMERA BOX DESIGN... 3 1. OPTICAL DESIGN... 3 2. MECHANICAL DESIGN... 3 3. ELECTRONICS... 4 3. PERFORMANCE TESTS... 5 1. SET UP... 5 2. WHAT IS TESTED... 5 3. EXPECTED PERFORMANCE... 6 4. STABILITY... 6 5. SUB PIXEL LINEARITY... 6 6. LINEARITY ON A MOVEMENT OF FEW PIXELS... 7 7. FULL FIELD LINEARITY... 8 8. EFFECT OF THE TEMPERATURE... 9 4. CONCLUSION... 12 5. REFERENCES... 12 6. ANNEX... 13 2
1. Introduction The positions of the muon chambers of the CMS experiment are precisely monitored with respect to each other and also with respect to the central detector. This is the role of the so called Muon Barrel Position Monitor. It is based on video cameras fixed on very stable mechanical structures called MABs ([1] [2] [3]). See Figure 18 for the Muon Barrel Monitor principle (see also the Muon Technical Design Report [1]). See Figure 19 for the design of the MAB. The video cameras are used for 3 purposes : 1- Look at optical sources mounted on the muon chambers. 2- Look at other MABs in order to provide an optical connection between the MABs. 3- Look at a Z bar mounted along the coil. The Z bars support optical sources at the level of each concerned MAB. This gives the z co-ordinate of the MABs. Hence one has 3 types of video cameras. The first type must look at sources in the distance range of [5mm, 75mm]. The second at 2 distances : 32mm and 47mm. The third one distance : about 5mm. The optical design of each type is different in order to get the best compromise between the field and the precision. The test of a prototype of the first type is presented in this report. This prototype contains the final optical elements and the final video sensor. The mechanical design (best stability, cost,...) will be optimised with the next prototype. 2. Camera box design 1. Optical design The same sensor is shared for two lenses thanks to 2 mirrors (see Figure 1). The angular incidence of the light on the sensor is 17 degrees. The standard lens used for the first prototype is a Melles Griot 1LPX15 (plan convex 15mm diameter, thickness 3.1 mm, focal length 5mm). The aperture of the lens is limited to a diameter of 1mm by a circular diaphragm. The lens is just a singlet to minimise the cost. It is plan-convex to get a minimum aberration for far objects. The mirror is a Microcontrole 72BD.1 (diameter 19.5mm, thickness 4.57mm) with imaging quality. The sensor is a Peach matrix sensor (VVL). It is a low cost integrated image sensor. The number of pixels is 312x287, the size of one pixel 19,6 x 16 microns 2 and the size of the sensitive part is 6.12x4.59 mm 2. Figure 1: Optical design of the 2 ways camera prototype. 2. Mechanical design This first prototype was design in a way that it can be easily machined in a very small workshop. Six plates are screwed together by 2 mm screws. The thickness of the plates is 3mm. The material is aluminium for easy machining. Each optical element is mounted on a holder fixed to these plates. The outer size of the box is designed to fit in the first MAB prototype. The Figure 2 shows the technical drawing of the prototype which was machined in Debrecen group for machining. The photographs of the camera box are shown in figure 3. 3
Figure 2: Technical design of the 2 ways camera prototype. Figure 3: Photographs of the prototype. The mirrors are at the bottom. The sensor is at the top. 3. Electronics The output of the video sensor is the standard CCIR video signal. This signal is sent to a frame grabber (Primo board from the Euresys company) which digitises it (ADC 8 bits), resamples the lines and stores the image in its memory. A C library permits to do some operations on this image : object detection and calculation of its centroid. The centroid is the calculated position of the object (horizontal and vertical co-ordinate in pixel unit) taking into account the intensity of light in each pixel. A particularity of this sensor is that, due to the small number of physical lines, the same line is repeated twice in the video signal in order to reach the standard CCIR number of lines. Then, just one line over two is acquired in order to get the same number of lines as the physical sensor. 4
3. Performance tests During one week, this camera box was tested on a small bench (see Figure 4). Only relative tests were performed like stability, linearity and temperature effect. Previous studies ([5] [6] [7]) using a standard camera box and a standard video lens where done, involving an absolute reference frame and studying the calibration of the camera box in this frame. 1. Set up The set-up is as follows. The camera box is rigidly fixed on a 1m long concrete table. The source is mounted on a precise translation stage. The stage is moved manually and the position is checked optically : a microscope permits to observe a precise ruler. The light source is a diffused LED mounted behind a diaphragm of 2 mm diameter (see Figure 5). The spot obtained on the video sensor has a size of about 2 pixels diameter. 8 mm Light source Camera prototype Optical table Translation Stage Figure 4: Set-up used for the test of performance : a precise translation stage (Karl Zeiss) displaces the source along an horizontal line. Figure 5 : Drawing of the source used : a diffused LED is placed behind a diaphragm of 2 mm diameter. 2. What is tested Here are the performance results that will be presented in this report : Stability : The difference between different measurements of an optical source which stays at the same place. Linearity : The optical source is displaced very precisely along a line. The camera set-up gives different centroid coordinates of the spot on the sensor. If the camera and the electronics are perfect, the curve showing the co- 5
ordinate measured by the camera with respect to the real displacement of the source is a perfect line. Three different tests will estimate the difference between the real measurements and the perfect line. Temperature effect : For a stable optical source the measurement consists in looking at the variation of the calculated co-ordinate when the temperature of the box changes. 3. Expected performance Simulation studies of the full barrel position monitor were made [4]. This simulation assumes that the chain : source, lens, sensor, electronics and processing brings in the sensor space an error of 1 micron sigma or 5% of a pixel. In a first approximation, this means that the error must stay in a range of +/- 2 microns or +/- 1% of a pixel. 4. Stability This test evaluates the short term (few minutes) stability. It focuses on the pure electronic noise. The Figure 6 shows the variation of centroid measurement for an untouched mechanical set-up. The calculated centroid varies randomly around an average value with a standard deviation of 2% of a pixel. By averaging a sequence of measurements done in a short time (few seconds), one can efficiently reduce this noise..1.8 Co-ordinate variation in pixels.6.4.2 -.2 -.4 -.6 -.8 -.1 CentroidX CentroidY 1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 Image number (time scale) Figure 6: Stability. Vertical unit: pixel. Horizontal unit : number of the centroid calculation. During a short period of time (few minutes) the same spot is acquired and its centroid is evaluated. The diameter of the spot is around 2 pixels. The sigma is around 2% of a pixel. This noise is random and can be reduced by averaging. 5. Sub-pixel linearity In this test, the optical source is moved transversally so that the spot moves on the sensor by less than one pixel. It is not obvious that the centroid calculation reflects accurately a motion smaller than one pixel. The Figure 7 shows the result of this test. A displacement of.2mm at 8mm distance was with steps of 5 microns. An experimental curve is plotted expressing the calculated centroid with respect to the displacement read with the translation stage. The vertical unit is the pixel. The best line is fitted to this experimental curve. The Figure 8 shows the difference between the calculated co-ordinates and this best line or residuals. The result of this test is that the very small range linearity is correct, with a maximum error of 1.5% of a pixel. Remark : The centroid co-ordinate given here is averaged over 5 images acquired sequentially. This reduces the random noise. The error bars give an estimation of the resulting random noise (+/-.2 pixel). 6
centroid displacement in pixel.9.8.7.6.5.4.3.2.1.2.3.5.6.8.9.11.12.14.15.17.18.2 Displacement at 8mm in mm. Figure 7: Sub pixel linearity test. Analyse of the behaviour of the centroid for a very small displacement (<1pixel). Fifty centroids are averaged for each point. At the right, the sketch shows the initial and final positions of the spot..2.15.1 residuals Pixel.5 -.5 -.1 -.15 -.2.1.2.3.4.5.6.7.8.9 Phi displacement at 8mm distance in mm.1.11.12.13.14.15.16.17.18.19 Figure 8: Difference with the best line fitting for a displacement of the source corresponding to a movement of the spot of less than one pixel (.2 mm at 8mm distance corresponding to.8 pixel). The error bars represent the estimated random error after averaging a sequence of 5 acquisitions..2 6. Linearity on a movement of few pixels Here, the same exercise is made with a slightly bigger range : 4 pixels which corresponds to a movement of 1mm at 8 mm distance. On Figure 9, one can see the calculated co-ordinate with respect to the transverse displacement of the source. As for the last part, this curve is compared to a fitted line. The residuals are shown in Figure 1. A periodicity of one pixel period can be observed. The result of this test is that the few pixels range linearity is correct, with a maximum error of 2% of a pixel. 7
Y centroid in pixels 4.5 4 3.5 3 2.5 2 1.5 1.5.1.2.3.4.5.6.7.8.9 1 "Phi" displacement at 8mm distance in mm Figure 9: Linearity test for a displacement of 4 pixels which corresponds to 1 mm at 8mm distance from the camera. At the right, the sketch shows the initial and final position of the spot..4.3 1 pixel.2 Residual in pixel.1 -.1.5.1.15.2.25.3.35.4.45.5.55.6.65.7.75.8.85.9.95 1 -.2 -.3 Displacement at 8mm in mm Figure 1: Residual from the best line fitting. The error bars represent the estimated random error after averaging a sequence of 5 acquisitions. A periodicity of 1 pixel can be observed. 7. Full field linearity Here, one checks the linearity over the whole field. The source is moved horizontally (which corresponds to the CMS ri co-ordinate). For the sensor, a vertical movement is observed. The source is displaced by 5mm which corresponds to 2 pixels on the sensor. The residuals (see Figure 11) from the fitted line have a maximum of 7% of a pixel (1.1 micron). This result must be taken with a lot of care since, for this large movement, the source imperfections, the optical aberrations of the lens and the sensor matrix imperfection might bring errors. In the future, specific tests will be done in order to produce a good design for the optical source. The source used here is a LED diffused and put behind a small diaphragm of 2mm diameter. The aberrations of the plano-convex lens can be precisely calculated providing a good knowledge of the source. The optical aberration might bring a very regular error for 8
is not the case with the curve presented here. Therefore, other effects than just the optical aberrations are present. 25 Y centroid in pixels 2 15 1 5-25 -22-19 -16-13 -1-7 -4-1 2 5 8 11 14 17 2 23 Full Sensor Phi displacement at 8mm in mm Figure 11: Full field linearity test over 5mm at 8mm distance. 2 pixels are covered..1.8 Residual from the best line in pixel.6.4.2 -.2 -.4 -.6-25 -22-19 -16-13 -1-7 -4-1 2 5 8 11 14 17 2 23 -.8 -.1 Phi displacement at 8mm in mm Figure 12: Residuals of a line fitting of a full field linearity test (over 2 pixels vertically for the sensor). The error bars represent the estimated random error after averaging a sequence of 5 acquisitions. 8. Effect of the temperature Since the typical size of the camera box is 5mm and the material is aluminium (thermal expansion coefficient: 24x1-6 degree -1 ), one can expect relative displacements of the optical elements in the box of the order of 1.2 microns per degree. This is more than half a pixel (12 microns) for a temperature change of 1 degrees. Fortunately, not all the deformations affect the precision of the camera box. For example, a displacement along the optical axis changes the magnification only but doesn t change the position of the spot a the centre of the field and provokes a very little displacement at the border of the field (since the numerical aperture is low). The test is done by sending a hot air flow on the box. Three temperature sensors are mounted on the box : 2 are connected to the metal and one measures the temperature of the air inside the box. See Figure 13. The duration of this experiment is about 1 hour. The warming up phase is faster than the cooling down phase. So one can estimate that the thermal state of the box is better represented by the numbers given by the sensors during the cooling down period. 9
Temperature face 2 Temperature face 1 Temperature inside the camera Temperature (C) 7 6 5 4 3 2 1 Temperature air inside the camera Temperature face 1 Temperature face 2 Figure 13: The temperature is taken on 3 points of the camera: 2 connected to metallic faces and one in the air inside the box. 1 5 9 13 17 21 25 29 33 Number of the temperature measurement Figure 14: There is coincidence between these 3 temperatures particularly during the slow cooling down phase (end part). 37 41 Result: The change of the position of the light spot with respect to the temperature is shown in the Figure 16. No thermal effect is detected for the vertical co-ordinate of the sensor which corresponds to the CMS I coordinate(see Figure 15). The main effect is observed for the CMS r co-ordinate which corresponds to the horizontal co-ordinate of the sensor. There is a displacement of the spot on the sensor of about 5% of a pixel (1 microns) per degree. Video line Y (vertical) X (horizontal) r ri Sensor frame CMS frame Lens Mirror + mirror holder Figure 15: Scheme in 3D of the optical arrangement of one side of the camera box. The sensor is mounted such that the video lines correspond to the CMS r axis. 1
.5 Centroid coordinate variation in pixel 25 3 35 4 45 5 55 6 65 -.5-1 X centroid (R) -1.5 Y centroid (Phi) -2 Temperature inside the box Warming up phase Cooling down phase Figure 16: Effect of the temperature on the centroid stability. Only the air temperature inside the box is plotted in this graph. There are two parts : warming up and cooling down part. The warming up part is the less sampled part. This effect can be explained by 2 main displacements : 1- The mirror is fixed on a prismatic holder. The dilatation of this holder make the mirror turning (about 15 microradian per degree which brings a displacement of the spot on the sensor of 1.62 microns per degree). 2- The full box dilates. Then the sensor moves up (by.86 micron per degree). Due to the incidence angle of 17 degrees, it implies a displacement of.26 microns per degree of the spot along the horizontal axis of the sensor (X axis). These 2 effects are compensating each other but the mirror rotation brings much bigger displacement of the light spot on the sensor (6 times more) than the box dilatation. If one chooses another material for the mirror holder such that the thermal expansion coefficient is 6 times lower than that of the aluminium (for example : glass), one can have 2 exactly compensating effects at the centre of the field. Nevertheless, the change of magnification would stay the same : a displacement of the spot of.7 micron per degree at the border of the field. warm position cold position Video sensor Mirror Mirror holder Figure 17: Schematic view of the 2 main effects of dilation. The mirror turns slightly and the sensor moves up. These 2 effects are compensating each other but the dilation of this holder brings bigger light spot displacement (6 times more) than the sensor translation. 11
4. Conclusion As explained in part 3.3, the error is considered to be satisfactory if it stays within the range +/-1% pixel or +/- 2 microns. This video camera is very satisfactory for these aspects : short term stability, sub-pixel linearity and small range linearity. The full field linearity is quite satisfactory. Nevertheless, more tests are needed particularly to understand the role of the source and the optical aberrations. The thermal stability is satisfactory in the CMS ri direction. For the CMS r direction, a non stability of 1 micron per degree has been detected. If the material of the mirror holder is adapted this non stability might go down to at the centre of the field and.7 micron per degree at the border of the field. The next table summarises these results : Source of error Maximum error measured Remark pixel micron Electronic noise 6% 1.2 Random (can be reduced efficiently) Sub pixel linearity 1.5%.3 Few pixels linearity 2.5%.5 Large field linearity 8% 1.4 Further studies needed (source) Temperature CMSIdirection Too small to be measured Too small to be measured In principle:.7 micron/degree at the border of the field. CMS r direction 5%/degree 1 micron/degree Can be in principle reduced to.7 micron/degree at the border of the field. 5. References [1.] Muon project Technical Design Report. [2.] L. Brunel CMS TN/94-239 Muon Barrel Alignment. [3.] L. Brunel CMS TN/96-123 Muon barrel alignment. Basic ideas. Third report. [4.] L. Brunel CMS-NOTE-96/18. Simulation of the muon barrel position monitor. [5.] L. Brunel CMS-TN-96/124. First manipulation for a video Camera Calibration [6.] L. Brunel CMS-TN-96/125. Second experiment for a video Camera Calibration [7.] L. Brunel CMS-NOTE-96/19. Video Camera Calibration. 12
6. Annex Figure 18: General view and principle of the Muon barrel position monitor. 13
Figure 19: an example of MAB (the position 3/-/1). There are 8 cameras of type 1 ( muon chamber cameras ) and 4 cameras of type 2 ( diagonal cameras ). 14