Review on Energy Resilience

D

R G C

h

�

�

�

�

�

a

A A

K B T N N L

1

d a a s c a e

0 h

Fusion Engineering and Design 88 (2013) 1253– 1256

Contents lists available at ScienceDirect

Fusion Engineering and Design

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / f u s e n g d e s

esign of a visible tomography diagnostic for negative ion RF source SPIDER

. Pasqualotto ∗, M. Agostini, M. Brombin, R.S. Delogu, N. Fonnesu, L. Lotto, F. Molon, C. Piron, . Serianni, C. Taliercio, M. Tollin, B. Zaniol

onsorzio RFX – Associazione Euratom-Enea sulla Fusione, corso Stati Uniti 4, I-35127 Padova, Italy

i g h l i g h t s

Prototype ITER neutral beam injector requires beam profile diagnostics. Visible tomography of H� emission can measure 2D beam intensity pro- file. A system with 15 linear CCD cameras and 3000 lines of sight is proposed. Camera specifications are discussed against requirements and expected signal. Test of prototype CCD camera vali- dates selected sensor.

g r a p h i c a l a b s t r a c t

r t i c l e i n f o

rticle history: vailable online 7 March 2013

eywords: eam diagnostic omography

a b s t r a c t

The ITER heating neutral beam injector, based on 1 MV D− ions, will be tested and optimized in the SPIDER source and MITICA full injector prototypes, using a set of diagnostics not available on ITER. Beam inten- sity uniformity is required to stay within ±10%, thus beam profile is measured with a complementary set of diagnostics. Among them, visible tomography measures the line of sight (LOS) integrated H� or D� radiation generated by the collisions between fast particles and neutral background molecules, and

eutral beam injector egative ion source inear CCD camera

emitted on a plane perpendicular to the beam. A sufficient number of well arranged LOSs allows a tomo- graphic reconstruction of the 2D beam emission profile, which is proportional to the beam density. On SPIDER the system is equipped with about 3000 LOSs, grouped in 15 fans, and the tomographic algorithm is based on the pixel method. The design of the diagnostic is presented, with description of layout and main components and test of the prototype linear CCD camera.

. Beam intensity profile diagnostics on SPIDER

The ITER heating neutral beam injector, based on a 40 A negative euterium ion beam produced by an RF source and accelerated t 1 MV [1], will be tested and optimized in the SPIDER source [2] nd MITICA full injector prototypes [3], using a comprehensive et of diagnostics [4]. The beam is composed of 1280 beamlets,

orresponding to the holes of the acceleration grids, spanning an rea of 1520 mm × 560 mm. SPIDER has to produce 100 kV accel- rated beam pulses, up to 1 h duration, with extracted D current

∗ Corresponding author. Tel.: +39 0498295849. E-mail address: [email protected] (R. Pasqualotto).

920-3796/$ – see front matter © 2013 Consorzio RFX Associazione Euratom ENEA sulla Fu ttp://dx.doi.org/10.1016/j.fusengdes.2013.02.009

© 2013 Consorzio RFX Associazione Euratom ENEA sulla Fusione. Published by Elsevier B.V. All rights reserved.

density > 285 A/m2, co-extracted electron fraction (e−/D−) < 1, beam uniformity within ±10% and beamlet divergence ≤7 mrad. These specifications have never been simultaneously achieved so far in a full scale, full performance device, thus source plasma and accelerated beam need to be thoroughly diagnosed.

A set of diagnostics measures the characteristics of the beam. Visible tomography estimates the 2D beam intensity profile with sufficient spatial resolution and its evolution during the entire pulse duration. It measures the line of sight (LOS) integrated H� or D� radiation (656 nm) generated after the collisions between

fast beam particles and neutral background molecules; a sufficient number of well arranged LOSs allows a tomographic reconstruc- tion of the 2D beam emission profile, which is proportional to the beam density. Tomography complements four other diagnostics.

sione. Published by Elsevier B.V. All rights reserved.

1 eering and Design 88 (2013) 1253– 1256

T t c r w s m m t D 9 s b b t p p t p p n b

i t r u a s t t a t c s d g t i v l d a d

2

v S p L i a v u o i t c l t s e

a

254 R. Pasqualotto et al. / Fusion Engin

he instrumented calorimeter STRIKE measures with IR cameras he thermal pattern on the back surface of a set of 1D carbon fiber omposite tiles, used as beam dump [5]. It provides the best spatial esolution (about 2 mm) at 50 Hz frame rate, but it is incompatible ith the heat load of long pulses (>10 s). Installed only on SPIDER, it

erves as a benchmark for other beam profile diagnostics to be used ore confidently on MITICA. Beam emission spectroscopy (BES) easures the LOS integrated, spectrally dispersed H� line [6]. The

wo corresponding line components are spectrally resolved by the oppler effect when the beam is observed at an angle different from 0◦. Line intensity is proportional to the average beam particle den- ity along the LOS: BES can be used to evaluate the beam uniformity, y comparison of LOS intensities, but with low spatial resolution, ecause of the limited number of LOSs. Its main role is to estimate he beam divergence from the width of the high energy line com- onent. Beam intensity profile is also measured on the beam dump anels: thermocouples measure cooling water and bulk material emperature, providing respectively an estimate of the vertical ower load profile (30 points) and a 2D temperature map (60 oints) of the dump panels front surface; a neutron imaging diag- ostic measures the neutron 2D profile generated by reaction of eam deuterons with those implanted in the dump surface [7].

The advantages of tomography are that it can provide a full 2D ntensity profile with high combined spatial and temporal resolu- ion, down to the order of some beamlets and tens of milliseconds espectively, worse only to STRIKE, for arbitrary pulse duration, sing no in-vacuum instrumentation. The two main limitations re profile shape constraints required for the tomographic inver- ion and the need of a large number of viewports to accommodate he fans of LOSs. After validation on SPIDER, tomography will be he main beam profile monitor in MITICA at intermediate stages long the accelerator, complemented by thermocouples and neu- ron imaging at the dump and by thermocouples on beam line omponents. The design was driven by the requirement to mea- ure the uniformity of the neutral beam, with maximum acceptable eviation from uniformity of ±10%, thus the deviation of the tomo- raphic reconstruction from the real emissivity of the beam has o be sufficiently lower than this value, i.e. few percent. This task s somewhat simplified because the uniformity is expected to be iolated only over long distances across the beam, not over a beam- et to beamlet distance, so that a 2D linear approximation of the ependence of the flux across the beam has been adopted to design nd test the expected capabilities of the system, as explained in etail in [8].

. Diagnostic layout

Design of the layout was constrained by the geometry of beam, acuum vessel and in-vacuum mechanical components, especially TRIKE. Two approaches were initially considered. Ideally each oint of a circle around the beam should be the apex of a fan of OSs covering the cross section of the beam. This could be achieved n the best approximation by an in-vacuum linear detector or fiber rray gyrating all around the beam, on a rail attached to the vacuum essel, similarly to medical tomography. The alternative solution is sing a finite set of linear cameras, mounted at fixed positions just utside the vacuum vessel, each looking at the beam through a ded- cated viewport and associated to a fan of LOSs. The in-vacuum solu- ion would be much richer in LOSs, but complicated by the remotely ontrolled in-vacuum motion associated with the cable or fiber ength motion with the camera or optical head and the much lower ime resolution (several seconds) compared to a multi-detector

ystem, because based on sequential acquisitions. As a result, the x-vacuum approach was preferred and further developed.

As illustrated in Fig. 1, the tomographic diagnostic comprises set of 15 dedicated viewports, arranged all around the vacuum

Fig. 1. Tomographic diagnostic layout: fans of LOSs through viewports around SPI- DER vessel; CCD cameras optically connected to PCs in diagnostic room.

vessel, about evenly distributed, except for the bottom part of the vessel, occupied by the BES viewports. From each viewport a fan of LOSs originates, crossing the beam at an angle that can range between 90 and 105◦ (looking downstream) to the beam axis. Clearly only one angle can be arranged with the beam at a time and standard operation will be at 90◦. LOSs are accommodated just downstream the exit of the source case, in order to avoid any interference with the STRIKE panels and their supporting frame.

The definition and arrangement of the 15 fans (vertex position, direction of the fan axis, angular aperture and number of LOSs within each fan) and of the corresponding viewports have been optimized within the development of the inversion algorithm [8].

3. Detection system

The detection system comprises one linear CCD camera with its objective lens, for each fan. A set of 15 cameras is required. Detector and lens have been designed according to the following requirements: detector imaged at about 2.2 m from the lens, i.e. in the center of the vessel; depth of focus about ±75 cm, i.e. the maximum intersection length between LOSs and beam; aperture of the fan up to 42◦ (full aperture); same CCD for all fans and min- imum number of different types of lenses, for design convenience, with the constraint of using most of the available CCD length; long enough CCD with a suitable focal length lens, to cover the largest aperture of the fans; optical system with a spatial resolution suf- ficiently smaller than the aperture of the single LOS and not much larger than the pixel size.

A wide bandpass interference filter could be used, installed on the lens, to select the H� line, rejecting other spectral contributions; however, especially outside the source, H� is expected to be by far the largest contribution to the intensity integrated over the entire spectrum and a filter might be unnecessary.

The CCD on its own must be selected with the following prop- erties: sufficiently high quantum efficiency at the H� wavelength; good sensitivity, to be compatible with the small signals resulting from the thin LOS (high F/#) and the time integration resulting from a frame rate in the range 10–100 Hz; high dynamic range, larger than effective 10 bits, in order to simultaneously measure with sufficient accuracy signals from different LOSs with intensity ratio 1:100; controllable from the central acquisition and control system of SPIDER, with the controller located outside the bio-shield, sev- eral meters from the camera; acquisition of single frames controlled

by external trigger pulse, in order to guarantee clear reference to central timing system.

Clearly the choice of the CCD-optics combination is also con- strained by selecting components easily available on the market, in

R. Pasqualotto et al. / Fusion Engineering and Design 88 (2013) 1253– 1256 1255

Table 1 Fan parameters, from left: fan aperture, fan apex angular position, number of LOSs, lens focal length, number of pixels used out of 2048 available and number of pixels/LOS.

̨ [deg] Pos. [deg] # LOS f [mm] # pixels used #/LOS

34.5 225 241 35 1612 7 31.3 200 218 50 1998 9 42.0 180 294 35 1994 7 22.3 161 156 50 1408 9 38.7 135 270 35 1824 7 33.1 117 231 35 1546 7 22.7 98 158 50 1431 9 23.0 90 161 50 1456 9 11.1 82 78 50 693 9 20.9 46 146 50 1320 9 23.0 18 160 50 1451 9 37.8 0 264 35 1780 7

o t i s o f d t a t i a L

l f t t r

f c C t i i w ( r u b t e

t t v p i r i p c

d i a

D

H

H

D

filter edge

653.5 654 654.5 655 655.5 656 656.5 nm

41.2 341 287 35 1955 7 39.8 332 278 35 1883 7 26.6 315 185 50 1686 9

rder to keep a reasonably low cost and spares availability. A sys- em that can satisfy all the listed requirements has been identified n the combination of (a) a linear CCD, 28 mm long, made of 2048 quare pixels of 14 �m length (size and number of pixels in the ther dimension is relatively flexible); (b) a commercial objective or a film camera, with a focal length of 35 or 50 mm, to accommo- ate the variety of aperture angles of the fans. In order to achieve he required depth of field, the 35 mm lens must be used with F/11 perture, while the 50 mm must be set at F/16. With this detec- ion system and with 2.5 mrad fixed aperture contiguous LOSs (the nversion algorithm actually considers a 50% overlap, with double perture), the LOS height at 2 m is 5.1 mm and the total number of OSs is 3127. Table 1 reports the 15 fans characteristics.

In principle a shorter CCD, e.g. with 1024 pixels and 14 mm ength, could be considered; in this case a cheaper lens suitable or 1′′ CCD (12.8 mm width) might probably be just compatible and he two focal lengths would be halved (18 and 25 mm). However he number of pixels/LOS and the signal would be halved: for this eason it has not been preferred.

A suitable CCD was selected and tested, chosen among the ew models available on the market used in spectroscopy appli- ations. A wider choice would be available of faster but noisier CDs, tailored for industrial applications. The chosen model was he linear CCD S10420-1106-01 from Hamamatsu, outstanding for ts dynamic range and low noise at the expense of a low speed. It s a back-thinned, high quantum efficiency, low noise CCD sensor

ith 2048 × 64 pixels, 14 �m × 14 �m pixel size, high sensitivity 6.5 �V/e−), high full well capacity (300 ke−) and wide dynamic ange (50,000) and maximum readout speed of 0.5 MHz. The man- facturer does not market a self-contained camera with this CCD, ut only a driver circuit to control it and manage the acquisition hrough a USB 2.0 port (board model C11287), with 14 bit ADC, xternal trigger, 10 ms frame readout time.

The performance of the optical system has been simulated with he Zemax code, using a standard 35 mm film camera lens. The spa- ial resolution of the system was found uniform over the field of iew and limited only by the diffraction limit. The image of a 14 �m ixel is about 0.8 mm at 2 m, while for comparison an emitting point

n the image plane produces on the CCD a spot diagram of ∼7.5 �m ms radius, and in turn a point on the CCD produces a spot diagram n the image plane of ∼0.33 mm rms radius, which implies that a ixel collects light from an about 1.5 mm wide area and there is a ross talk of about half pixel between adjacent LOSs.

The photon flux density of the 100 keV beam with an average ensity of about 2 × 1014 m−3 traveling at velocity of order 106 m/s

n a background gas with density of order 1018 m−3 (resulting from residual gas pressure of 0.05 Pa and assuming a gas temperature

Fig. 2. H� and D� lines with sketched 0.5 nm transition length of an interference filter.

of 2000 K), is about 5 × 1017 ph/s/m3 [6]. From this, the expected signal on a CCD pixel can be calculated, considering a 1 m long LOS and the throughput of the system given by the 3 mm diam- eter aperture of the collection lens, a 0.8 mm diameter source at 2 m from the lens, corresponding to the pixel image: about 2 × 108 photons/s. Assuming 70% quantum efficiency, 5% optics transmis- sion, 12.2 electrons/ADC-count and 1 ms integration, the signal per pixel is 1 × 104 electrons or 800 counts. The detector saturates at 3 × 105 electrons with full line binning, the acquisition board at 1.6 × 104 counts and has a readout noise of 3 counts. The binned signal is then expected just above the saturation level. Marginal LOSs intersecting only one beamlet can have 100 times less signal, which can still be measured with ∼2% readout noise.

Setting the fans not normal to the beam, e.g. at 15◦, maximum allowed to still avoid any mechanical interference, would allow to spectrally separate the beam and background gas H� components, opening the possibility to detect only one of them if the other is properly filtered out. At 15◦, the two components are 1.5 nm apart for D and 2.4 nm for H (Fig. 2). One component could be selected using a sharp interference filter, with minimum realistically man- ufacturable transition width of 0.5 nm. This would leave only 1 nm to accommodate the blue-shift spread caused by the finite range of angles of incidence onto the filter, limiting it to ±5◦, narrower than most fan apertures. A possible solution could be an eyepiece instead of the standard photographic objective or after it, operated with the light propagating in opposite direction than it normally is. This approach is currently under investigation.

4. Prototype tests

A prototype camera has been home made with the CCD, the readout board and a 50 mm focal length film camera lens. A setup simulating the operational layout on SPIDER was assembled: cam- era and lens were installed on a rotation stage, on a rail where a target was positioned at different distances from it, corresponding to the different points along the LOS. Camera rotation was used to simulate the fan of LOSs, each rotation position representing a specific LOS. The required depth of field, ranging between 1.5 and 3 m from the lens, was best achieved by setting the objective aper- ture at F/16. Finding the best focusing across the depth of field was performed by positioning a target with sets of parallel black and white line pairs (lp) and looking for an optimum contrast simul- taneously at the two extremes and in the center of the depth of field. The modulation transfer function (MTF) has been measured to quantitatively characterize the optical performance, calculating the relative contrast by using the target introduced above: the MTFs have been measured for the central LOS at the three main positions

along the depth of field and at the middle position (225 cm) for the central and two extreme LOSs (±15◦). The curve is the same for all angles and positions and 50% contrast is achieved at 0.25 lp/mm (Fig. 3), i.e. the spatial resolution is 2 mm.

1256 R. Pasqualotto et al. / Fusion Engineering

Fig. 3. MTF measured for the central LOS at the three main positions along the depth of field.

0

2

4

6

8

10

12

100 1000 10000

re la

tiv e

no is

e (%

)

counts

experimental noise

Poisson noise

readout noise: 3 counts

F t

g s l m a p t d o s

c g T t v c m

e u a f w

[

[

[

[

[

[

ig. 4. Measured relative standard deviation and theoretical Poisson noise of pho- oelectrons vs signal level.

Other characteristics of the detector have been tested: electronic ain linearity, by changing the gain value, and linearity with input ignal, by varying the duration of a LED pulse, recorded with a onger integration time. Detector noise has been characterized by

easuring the standard deviation at different signal levels, using n integration time of 20 ms, and comparing it with the Poisson hotoelectron noise, showing that this is the dominant contribu- ion down to 100 counts, under which the 3 counts readout noise ominates (Fig. 4). A relative noise <10%, required from the the- retical prediction and the inversion results of [8], is achieved at ignal levels >30 counts, leaving a useful dynamic range of 1:500.

The driver circuit has a trigger input, so that each single frame an be externally triggered, and an output synchronized with trig- er input, to drive external devices like a shutter or a light source. he prototype camera is not equipped with a shutter, meaning hat acquisition is performed also during the readout, causing ertical blooming, which is however not worrying in this appli- ation, as tomography will always use the CCD in full binning ode. A custom acquisition program compatible with the MDSPlus

nvironment of the central SPIDER acquisition system was built,

sing the drivers provided with the prototype board, to test the cquisition of a sequence of contiguous frames and the maximum rame rate. Frame sequences were acquired in full binning mode, ith 10 ms integration time and externally triggering every frame.

[

[

and Design 88 (2013) 1253– 1256

Without downloading any data during the triggered operation, 50 Hz maximum frame rate was obtained: in this case acquired frames are just stored in the board memory and are overwritten during acquisition. When instead data are transferred to the PC through the USB port during the acquisition, the maximum achiev- able frame rate is 42 Hz, which is still acceptable for this application. If in addition data are also stored to disk during the acquisition, the frame rate is not affected. A still unsolved issue is that the avail- able drivers can control only one device, meaning that a dedicated PC is required for each CCD camera. Attempts to use more virtual machines running on the same PC and each controlling one CCD did not work, failing to guarantee a constant frame rate, because of the excessive amount of resources required for the simultaneous acqui- sition. To solve this problem, other camera integrators are being contacted to develop a new custom driver board using the same CCD.

5. Conclusions

A tomographic D� diagnostic has been designed to measure the SPIDER beam profile and uniformity, based on 15 linear CCD cameras arranged around the beam. The system layout complies with the requirements set in ref. [8], providing the necessary 3000 LOSs with the proper distribution for the inversion algo- rithm. A suitable CCD has been evaluated and tested, coupled to a photographic lens, providing the dynamic range and sensitiv- ity required by expected signal level, and the spatial resolution and depth of focus needed for the LOSs. Some limitations in operability of the associated electronics are driving the devel- opment of a new camera based on the same CCD, while the applicability of an interference filter is being further investi- gated.

Acknowledgments

This work was set up in collaboration and financial support of Fusion for Energy.

References

1] R. Hemsworth, H. Decamps, J. Graceffa, B. Schunke, M. Tanaka, M. Dremel, et al., Nuclear Fusion 49 (2009) 045006.

2] P. Sonato, P. Agostinetti, G. Anaclerio, V. Antoni, O. Barana, M. Bigi, et al., Fusion Engineering and Design 84 (2009) 269.

3] A. Masiello, G. Agarici, T. Bonicelli, M. Simon, J. Alonso, M. Bigi, et al., Fusion Engineering and Design 86 (2011) 860.

4] R. Pasqualotto, G. Serianni, P. Sonato, M. Agostini, M. Brombin, G. Croci, et al., Review of Scientific Instruments 83 02 (2012) B103.

5] G. Serianni, M. Dalla Palma, M. De Muri, D. Fasolo, R. Pasqualotto, N. Pomaro, et al., Review of Scientific Instruments 83 02 (2012) B725.

6] B. Zaniol, R. Pasqualotto, M. Barbisan, Review of Scientific Instruments 83 (2012)

043117.

7] G. Croci, M. Rebai, G. Claps, M. Cavenago, M. Dalla Palma, G. Gervasini, et al., Journal of Instrumentation 7 (2012) C03010.

8] M. Agostini, M. Brombin, G. Serianni, R. Pasqualotto, Physical review special topics – accelerators and beams 14 (2011) 102801.