ARTICLE IN PRESS

0168-9002/$ - se

doi:10.1016/j.ni

�Correspondi050-221-4317.

E-mail addre

(R. Bellazzini).

Nuclear Instruments and Methods in Physics Research A 535 (2004) 477–484

www.elsevier.com/locate/nima

Reading a GEM with a VLSI pixel ASIC used as a directcharge collecting anode

R. Bellazzinia,�, F. Angelinia, L. Baldinia, F. Bittia, A. Breza, M. Ceccantia,L. Latronicoa, M.M. Massaia, M. Minutia, N. Omodeia, M. Razzanoa,

C. Sgroa, G. Spandrea, E. Costa b, P. Soffittab

aINFN-Sezione di Pisa, Via Buonarroti 2, 56127 Pisa, ItalybIstituto di Astrofisica Spaziale del CNR, Area di Ricerca di Roma, V. Fosso del Cavaliere, 00131 Rome, Italy

Available online 22 September 2004

Abstract

In MicroPattern Gas Detectors (MPGD) when the pixel size is below 100mm and the number of pixels is large (above

1000) it is virtually impossible to use the conventional PCB read-out approach to bring the signal charge from the

individual pixel to the external electronics chain. For this reason a custom CMOS array of 2101 active pixels with 80mmpitch, directly used as the charge collecting anode of a GEM amplifying structure, has been developed and built. Each

charge collecting pad, hexagonally shaped, realized using the top metal layer of a deep submicron VLSI technology is

individually connected to a full electronics chain (pre-amplifier, shaping-amplifier, sample & hold, multiplexer) which is

built immediately below it by using the remaining five active layers. The GEM and the drift electrode window are

assembled directly over the chip so the ASIC itself becomes the pixelized anode of a MPGD. With this approach, for

the first time, gas detectors have reached the level of integration and resolution typical of solid-state pixel detectors.

Results from the first tests of this new read-out concept are presented. An Astronomical X-ray Polarimetry application

is also discussed.

r 2004 Elsevier B.V. All rights reserved.

1. Introduction

The most interesting feature of the Gas ElectronMultiplier (GEM) is the possibility of full decou-pling of the charge amplification structure from

e front matter r 2004 Elsevier B.V. All rights reserve

ma.2004.07.269

ng author. Tel.: +39-050-221-4367; fax: +39-

ss: ronaldo.bellazzini@pi.infn.it

the read-out structure. In this way both can beindependently optimized. Indeed, by organizingthe read-out plane in a multi-pixel pattern it ispossible to get a true 2D imaging capability. At thesame time a high granularity of the read-out planewould also allow to preserve the intrinsic resolvingpower of the device and its high-rate capabilitythat otherwise would be unavoidably lost by usinga conventional projective read-out approach.However, when the pixel size is small (below

d.

ARTICLE IN PRESS

R. Bellazzini et al. / Nuclear Instruments and Methods in Physics Research A 535 (2004) 477–484478

100mm) and the number of pixels is large (above1000) it is virtually impossible to bring the signalcharge from the individual pixel to a chain ofexternal read-out electronics even by using anadvanced, fine-line, multi-layer, PCB technology.The fan-out which connects the segmented anodescollecting the charge to the front-end electronics isthe real bottleneck. Technological constraints limitthe maximum number of independent electronicschannels that can be brought to the peripheralelectronics. Furthermore, the crosstalk betweenadjacent channels and the noise due to the highinput capacitance to the preamplifiers does notbecome negligible. In this case, it is the electronicschain that has to be brought to the individualpixel. We have implemented this concept bydeveloping and building a CMOS VLSI array of2101 pixels with 80mm pitch which is used directlyas the charge collecting anode of the GEM. Adescription of the read-out ASIC for a MPGD andof its advantages is given in the next section.Section 3 describes the coupling of the chip die tothe amplifying electrode, the assembly of the fulldetector and the results of laboratory testsobtained with a 5.9 keV X-ray source. The use ofthis new detection concept for Astronomical X-rayPolarimetry and other applications are discussedin the last section.

2. The CMOS VLSI chip

A drawing of the ASIC layout as seen from thetop metal layer is shown in Fig. 1. The activematrix in pink is surrounded by a passive guardring of 3–4 pixels set to the same potential as theactive pixels. The chip has been realized using a0:35mm 3.3V CMOS technology. No specific ESDprotection other than the parasitic capacitance ofthe drain-to-bulk junctions has been foreseen forthe pixel pads. Each microscopic pixel is fullycovered by a hexagonal metal electrode realizedusing the top layer of a 6 layers CMOS technol-ogy. Each pad is individually connected to a fullchain of nuclear type electronics (pre-amplifier,shaping-amplifier, sample & hold, multiplexer)which is built immediately below it by makinguse of the remaining five active layers. Fig. 2 shows

the layout and the simplified equivalent scheme forone pixel. Upon activation of an external asyn-chronous trigger (in our case provided by amplify-ing and discriminating the fast signal obtainedfrom the top GEM electrode) and within a 10mswindow the automatic search of the maximumshaped signal starts. If the MaxHold signal is set,the maximum is held for subsequent read-outwhich is accomplished by sequentially connectingthe output of each pixel to a common analog bus(Fig. 3). A pixel is selected by introducing a tokeninto the shift register and can be electricallystimulated at the rising edge of the Write signal,injecting a charge �Qin (10 fC/V typical response)proportional to the voltage difference betweenVtest and Vss. Tokens are shifted one cell forwardat the falling edge of the input clock. If severaltokens are present in the shift register then theanalog output corresponds to the sum of theselected pixels, up to the saturation level of�30 fC. A useful feature of the chip is thepossibility to work both in Hold or Track mode.The shaped pulse from a pixel can be individuallyobserved at the analog output by keeping theMaxHold signal low. In Fig. 4 three differentshaped signals obtained by injecting a charge of1000, 6000 and 60 000 electrons, respectively, intothe calibration capacitance are shown. Requestedspecifications for the ASIC prototype are:

�

low noise (typical ENC �100 electrons at 0.1 pFinput capacitance),

�

�3:5 ms shaping time, � 60mW typical power consumption per pixel, � up to 5MHz system clock (i.e. serial analog

read-out at 200 ns/pixel corresponding to�400 ms total read-out time for 2100 pixels),

�

0.2–20 fC dynamic range, � 100mV/fC amplifier gain.

This read-out approach has the advantage com-pared to similar ones (TFT [1] or CCD [2] read-out) of being fully asynchronous and externallytriggerable. Furthermore, it supplies the completeanalog information of the collected charge allow-ing to image the energy deposition process of theabsorbed radiation. A photo of the actual ASICbonded to a ceramic CLCC68 package and a zoomover the hexagonal pixel pattern is shown in Fig. 5.

ARTICLE IN PRESS

Fig. 2. A drawing of the pixel layout with underlying electronics and its simplified equivalent electronic scheme.

Fig. 1. The actual ASIC layout as seen from the top (active pixels are pink, guard-ring and I/O pads are blue).

R. Bellazzini et al. / Nuclear Instruments and Methods in Physics Research A 535 (2004) 477–484 479

ARTICLE IN PRESS

Fig. 3. Serial read-out architecture.

Fig. 4. A photo of the shaped signals for three different values

of input charge at 100mV/fC amplifier gain.

R. Bellazzini et al. / Nuclear Instruments and Methods in Physics Research A 535 (2004) 477–484480

3. The MPGD assembly

A single GEM MPGD with an active gasvolume of less than 1 cm3 has been assembleddirectly over the chip die, so the ASIC itself hasbecome the pixelized collecting anode of thedetector. With this approach, for the first time,gas detectors have reached the level of integrationand resolution typical of solid state pixel detectors.Different phases of the assembly are shown inFig. 6. In the actual prototype a drift region(absorption gap) of 6mm above the GEM foil hasbeen chosen, while a 1mm spacer defines thecollection gap between the bottom GEM andthe pixel matrix of the read-out chip. The GEMhas a standard thickness of 50mm and holesof 50 mm at 90mm pitch on a triangular pattern.

ARTICLE IN PRESS

Fig. 5. A photo of the chip bonded to the CLCC68 ceramic package and a zoomed view of the pixel matrix.

Fig. 6. Assembly phases of the MPGD over the chip: (a) All the mechanical details of the top section of the detector are glued together

while the chip is still protected by a metallic cover, (b) the chip is exposed and the mechanics glued upon it, (c) the MPGD is ready for

test.

R. Bellazzini et al. / Nuclear Instruments and Methods in Physics Research A 535 (2004) 477–484 481

The entrance window is a 25mm Mylar foil,aluminized on one side. The gas mixture used tofill the detector is 80% neon, 20% DME. Such alow Z gas mixture has been chosen for the highstopping power/scattering ratio and a still reason-able detection efficiency at low X-ray energy.Typical voltages applied to the drift electrode andto the GEM are, respectively: �1000V, �500V(Top GEM), �100V (Bottom GEM), the collect-ing electrodes being at �0 voltage. In thiscondition the detector operates at a typical gainof 1000. Thanks to the very low pixel capaci-tance at the preamplifier input, a noise level of1.8mV corresponding to �100 electrons has been

measured. The rms value of the pedestals distribu-tion for each read-out channel is reported in Fig. 7.With a gas gain of 1000 and the measured noiselevel the detector has significant sensitivity to asingle primary electron. Strobing each pixel witha 1V signal (�1000ADC counts) a uniformityof response of 3% rms for all the 2101 channelshas been observed. As all the processing occurswithin the pixel, negligible crosstalk has beenmeasured in the channels adjacent to the onespulsed with the 1V signal. The addressingcapability of each individual pixel has beenchecked with the internal calibration system. Thefirst application of this new MPGD concept is

ARTICLE IN PRESS

Fig. 7. Noise measurement: rms value of the pedestals

distribution for each electronics channel.

Fig. 8. Reconstructed track of a 5 keV photoelectron. Track direction

step. (Read-out frequency 5MHz).

R. Bellazzini et al. / Nuclear Instruments and Methods in Physics Research A 535 (2004) 477–484482

for an Astronomical X-ray polarimeter in the lowenergy band 1–10 keV. Information on the degreeand angle of polarization of astronomical sourcescan be derived from the angular distribution of theinitial part of the photoelectron tracks whenprojected onto a finely segmented 2D imagingdetector. As reported in previous papers ([3,4]) thealgorithm for the reconstruction of the photoelec-tron path starts from the evaluation of thebarycenter of the charge distribution on the read-out pixels and the maximization of the secondmoment (M2) of the charge distribution to definethe principal axis of the track. In a further stepthe asymmetry of the charge release along the

reconstruction algorithm: red line, first step; blue line, second

ARTICLE IN PRESS

Fig. 9. The raw signals relative to the event shown in Fig. 8.

Fig. 10. Real reconstructed tracks.

R. Bellazzini et al. / Nuclear Instruments and Methods in Physics Research A 535 (2004) 477–484 483

principal axis (third moment M3) is computed andthe conversion point derived by moving along thisaxis in the direction of negative M3; where thereleased charge is smaller, by a length � M2: Thereconstruction of the direction of emission is thendone by taking into account only the pixels in aregion weighted according to the distance from theestimated conversion point. The morphology of areal track obtained by illuminating the device witha low-energy radioactive source (5.9 keV X-rayfrom 55Fe) is shown in Fig. 8. The small cluster

due to the Auger electron and the initial part of thetrack are well distinguishable from the largerBragg peak. The projection of the charge distribu-tion along the principal axis is also shown. Theplot of the raw signals of all the channels forthe same event shows the optimal signal to noiseratio obtained with this detector (Fig. 9). Around50 000 electrons from the gas-amplified primaryphotoelectrons are subdivided over 53 pixels. Tworeal events, including a double track, are shown inFig. 10.

ARTICLE IN PRESS

R. Bellazzini et al. / Nuclear Instruments and Methods in Physics Research A 535 (2004) 477–484484

4. Conclusions

A system in which the GEM foil, the absorptiongap and the entrance window are assembleddirectly over a custom CMOS chip die has beendeveloped. The transfer of charge from theamplifying region to the collection and read-outregion occurs via electric fields. The ASIC itselfbecomes at the same time the charge collectinganode and the pixelized read-out of a MicroPat-tern Gas Detector. For the first time the fullelectronics chain and the detector are completelyintegrated without the need of complicated bump-bonding. At a gain of 1000 a high sensitivity tosingle primary electron detection is reached. Noproblems have been found up to now in operatingthe system under HV and in a gas environment.An astronomical X-ray Polarimeter applicationhas been presented. The final design will have16–32 k channels and 60–70mm pixel size (’ 1 cm2

active area). Depending on pixel and die size,electronic shaping time, analog vs. digital read-out, counting vs. integrating mode, and gas filling,many others applications can be envisaged. For

example, at this conference a similar approachusing an existing CMOS chip (MediPixz) coupledto a triple GEM Micromegas gas amplifyingstructure for a TCP application has been presented[5]. This would open new directions in gas detectorread-out, bringing the field to the same level ofintegration as in solid-state detectors.

References

[1] J.K. Black, P. Deines-Jones, S.E. Ready, R.A. Street, Nucl.

Instr. and Meth. A 513 (Issue 3) (2003) 639.

[2] M. Bogner, G. Buschhorn, R. Kotthaus, R. Oberhuber,

M. Rzepka, K.H. Schmidt, Nucl. Instr. and Meth. A 377

(1996) 529.

[3] E. Costa, P. Soffitta, R. Bellazzini, A. Brez, N. Lumb,

G. Spandre, Nature 411 (2001) 662.

[4] R. Bellazzini, F. Angelini, L. Baldini, A. Brez, E. Costa,

L. Latronico, N. Lumb, M.M. Massai, N. Omodei, P.

Soffitta, G. Spandre, IEEE Trans. Nucl. Sci. 49 (3)

(2002).

[5] P. Colas, A.P. Colijn, A. Fornaini, Y. Giomataris, H. van

der Graaf, E.H.M. Heijne, X. Llopart, J. Schmitz,

J. Timmermans, J.L. Vischhers, Proceedings of the 10th

Vienna Conference on Instrumentation, Vienna, 2004, Nucl.

Instr. and Meth. A, these proceedings.