ARTICLE IN PRESS

0168-9002/$ - se

doi:10.1016/j.ni

�CorrespondE-mail addr

Nuclear Instruments and Methods in Physics Research A 579 (2007) 844–847

www.elsevier.com/locate/nima

Serial powering of silicon strip detectors at SLHC

Marc Weber�, Giulio Villani, Mike Tyndel, Robert Apsimon

Rutherford Appleton Laboratory, Chilton Didcot, OX11 0QX, UK

Available online 31 May 2007

Abstract

Serial powering is a technique to power a large numbers of detector modules using a single power cable and a constant current source.

Module power is derived using local shunt regulators. Serial powering is of great interest for large-scale silicon tracking detectors for

particle physics with many thousands of densely packed detector modules. We have been operating six ATLAS Semiconductor Tracker

(SCT) modules using both serial powering and conventional independent powering. The modules operated stably with no indication of

any extra noise sources. We present selected results and discuss the system aspects of serial powering.

r 2007 Elsevier B.V. All rights reserved.

PACS: 29.40.Gx; 29.40.Wk

Keywords: Solid-state detectors; Silicon detectors

1. Introduction

The Large Hadron Collider (LHC) at CERN, Geneva,will be upgraded to the Super LHC (SLHC) to provide a10-fold increased luminosity by the year 2015. The silicontrackers of the LHC experiments will then be at the end oftheir life time and will need to be replaced. The futureATLAS tracker for SLHC will be an all-silicon trackerwith an inner pixel region surrounded by silicon micro-stripdetectors. It will consist of 30–60 million strips with a totalsensor area of between 150 and 200m2.

The new tracker will be an object of unprecedentedtechnical complexity. One key challenge is the distributionof power to the front-end electronics. In current silicontracking systems, each detector module is powered inde-pendently. The current ATLAS Semiconductor Tracker(SCT) [1], for example, requires several thousand powercables. The cable length exceeds 80m (one way) andresistance (including return) is several Ohms. This luxurycan no longer be afforded for ever larger and densertracking systems. Independent powering fails because

1.

There is no space for a 5- or 10-fold increased number ofpower cables.

e front matter r 2007 Elsevier B.V. All rights reserved.

ma.2007.05.290

ing author. Tel.: +441235 446061.

ess: m.m.weber@rl.ac.uk (M. Weber).

2.

The power efficiency is too low. More than 85% of thetotal power could be lost in the cables.

3.

The material represented by the power cables leads toexcessive multiple scattering and creation of secondaryparticles. Note that the current on detector powerflex circuits present 0.2% of a radiation length forone layer of SCT barrel modules at normal trackincidence.

4.

The number of cables causes severe packagingand interconnection constraints for detector modulesor supermodules.

Serial powering provides an elegant solution to each ofthese problems.

2. Serial powering

Serial powering has been reported for ATLAS pixelmodules with most encouraging results [2]. Its applicationto strip sensors is more critical due to the larger noisesensitivity and the size of silicon strip detector systems [3].A serial powering system for silicon detectors consists of

four elements: a current source; a shunt regulator andpower device (for digital power); a linear regulator(for analog power); and AC or opto-coupling of clock,command and data signals.

ARTICLE IN PRESS

Fig. 1. Example of a serial powering scheme. Analog and digital power are derived by regulators on each module. In this example, the module operation

voltage is assumed to be 4V as for SCT. (For SLHC, the module voltages will be reduced to approximately 1.5V.)

Fig. 2. Serial powering set-up with six SCT modules, serial powering

interface board, DAQ support cards and a constant current source in the

RAL clean room.

M. Weber et al. / Nuclear Instruments and Methods in Physics Research A 579 (2007) 844–847 845

The modules are all chained in series as sketched inFig. 1. The number of long cables is reduced by a factor of2n, if n modules are powered in series. (The factor 2 arisesfrom using a single power supply to derive analog anddigital voltage rather than providing them separately.)

Each module sits at a different potential and the totalvoltage across a series of n modules is n times the module

voltage. The current needed to power the total chain issimply the module current plus the (small amount of)current lost in the regulators. Note that analog ground,digital ground and sensor bias ground are tied together onthe module, as is common practice for independentpowering as well. Since the grounds of different modulesare different, floating HV power supplies must be used.

Serial powering can be much more efficient thanindependent powering since thermal losses in cables arereduced by a factor of n. The power efficiency for serialpowering is 1/(1+IR/Un), while for independent poweringit is 1/(1+IR/U), where I is the module current, R is thecable resistance, and U is the module voltage. For anSLHC tracker, typical values of R, I, and U would beR ¼ 4O, I ¼ 2.5A, U ¼ 1.5V. This implies that only 13%of the delivered power would reach the module forindependent powering. This is unacceptable for a trackerusing at least E100 kW of module power. Serial poweringgreatly improves efficiency. With the above assumptionsand powering 16 modules in series, power efficiency reaches70%, reducing the total power by a factor of five.

3. Set-up with SCT modules

In order to test serial powering with SCT modules [4], wehave built a serial powering interface board, whichconnects to an SCT module on one end and to the DAQ

(support card) on the other. The board is realized as a two-layer printed circuit board (PCB) and uses commercialpackaged ICs to implement the regulator and othercircuitry. Control, clock and data signals are AC coupled.A photograph of the set-up is shown in Fig. 2.The board was not optimized for size, but a smaller

38� 9mm2 large version was also built to implement serialpowering in a highly integrated supermodule (see Ref. [5]).The small board uses bare die ICs and is realized as a four-layer PCB. Increased miniaturization will be achieved ifserial powering circuitry is placed directly on the read-out hybrid and if custom (radiation-hard) ICs becomeavailable.

ARTICLE IN PRESSM. Weber et al. / Nuclear Instruments and Methods in Physics Research A 579 (2007) 844–847846

4. Results

Standard SCT detector evaluation software was run todetermine noise and gain of six SCT modules poweredindependently or in series. The average noise performanceof the modules powered in series is similar if not betterthan that of the same modules powered independently(see Fig. 3). The same conclusions are valid for eachchannel. The chip gains for independent and serial power-ing are consistent.

More sophisticated tests were also performed. Switchingoff the bias voltage of one of the six modules does not

Distribution

board 2

Digital

PS 2

Digital

PS n

Analog

PS 1

Digital

PS 1

Analog

PS 2

Analog

PS n

Distribution

board 1

Distribution

Board 2

Power

supply

Distribution

Board 1

Power

supplies

c c b b a

Fig. 4. Sketch of power connections for n modules powered independently (t

connections is indicated schematically by using different line widths. The conne

connections would typically be commercial connectors.

1350

1400

1450

1500

1550

1600

755 663 159 628 662 006

Module #

<E

NC

>

IP

SP

Fig. 3. Average noise (ENC) for six SCT modules powered independently

(IP) or in series (SP). The modules were run for more than 24 h before

collecting the data shown. The statistical precision of the data points

varies between 1.3 and 5 e.

increase the noise levels of the other modules. Similarly,operating one module at a low (fixed) discriminatorthreshold, which increases its digital power consumptionand occupancy, has no effect on the other modules. Inanother test, a sinusoidal current modulation of 15mAamplitude was superimposed to the serial powering currentof 1.6A. The modulation frequency was varied over arange of 1Hz–80MHz and data from noise runs werecollected. We performed a similar scan with 2.5MHzfrequency steps for a single hybrid. No increase of noisebeyond a few percent is observed.Serial powering improves the overall power efficiency by

large factors, as shown in Section 2, by minimizing thepower losses in the cables. However, there will be anincrease in the local power consumption because of theadditional circuitry required. We compared the power con-sumption of the six modules for the two powering schemes.The power consumption of the serial powering circuitry isE18% of the module power and is dominated by thermallosses in the power transistor. The losses arise because themodule current is not constant but varies depending onthe readout chip activity. With the constant current sourcecurrent being set to the maximum module current, theselosses are roughly given by the module operation voltagetimes the difference in average module current andconstant current source current. The power consumptionof the regulator and LVDS buffer chips is small incomparison. For future readout chip generations withlower operation voltages and reduced digital currentfluctuations, the power consumption of the serial powercircuitry will be significantly reduced.

Module 1 Module 2 Module n

Module 1 Module 2 Module n

a aa

op) and serially (bottom). The different wire gauge of the a, b and c type

ctions to the modules could be realized with wire bonds, whereas the other

ARTICLE IN PRESSM. Weber et al. / Nuclear Instruments and Methods in Physics Research A 579 (2007) 844–847 847

5. Risk analysis

For normal independent powering, a broken connection(or a short!) would lose only one module. For serialpowering, a single-point failure leading to an open on theserial current loop will leave all modules in a seriesunpowered. It is possible to quantify this risk, which wedefine as the probability to break a given connection timesthe number of modules lost, summed over all connections.

For ATLAS SCT there are four different cable andconnector types. Near the power supply, large wire gaugecables are used to minimize ohmic losses. Closer to thedetector thinner and shorter cables are used; on thedetector flexible power tapes are routed to the detectormodules. In Fig. 4, a typical arrangement of cables isshown. To determine the risk we assign a failureprobability a, b, c, etc. to each connection/connector typeand count the number of connections. The risk of serialpowering normalized to that of independent powering isgiven by [aSP(n+1)+2bSP+2cSP]/4(aIP+bIP+cIP), whereaSP is the probability of failure for the power connection tothe module for serial powering, aIP is the correspondingprobability for independent powering, bSP relates to theconnection between the distribution boards 1 and 2, cSPrelates to the connection at the power supply, etc. Clearly,serial powering becomes more risky with increasing modulenumber, but the risk is strongly ‘‘damped’’ by a factorof four. If we assume that the larger connectors aresignificantly more reliable than wire bonds, the risk ratiocan be approximated by aSP(n+1)/4aIP.

A serial system with 16 modules in series would beapproximately four times more risky than an independentpowering system if aSP equals aIP. In practice, the largereal-estate gains of serial powering systems would beexploited to engineer very robust module connections.

6. Conclusions

We designed and built circuitry to power six SCT siliconmicro-strip modules in series. Noise performance wasstudied in detail and is found to be excellent. The powerconsumption of the serial powering regulators was foundto be E18% of the module power and is dominated by thecurrent variations of the readout chip. A first risk analysissuggests that multi-module serial powering systems are lessrisky than naively anticipated.

References

[1] ATLAS Collaboration, Inner Detector Technical Design Report,

CERN/LHCC/97-16 and CERN/LHCC/97-17, 1997.

[2] D.B. Ta, T. Stockmanns, F. Hugging, P. Fischer, J. Grosse-Knetter,

O. Runolfsson, N. Wermes, Nucl. Instr. and Meth. A 557 (2006)

445.

[3] M. Weber, G. Villani, M. Lammentausta, Serial powering for silicon

strip detectors at SLHC, CERN-LHCC-2005-038, in: Proceedings of

the 11th Workshop on Electronics for LHC and Future Experiments,

pp. 214–217.

[4] J. Carter, et al., Nucl. Instr. and Meth. A 568 (2006) 642.

[5] C. Haber, et al., Development of Large Area Integrated Silicon

Tracking Elements for the LHC Luminosity Upgrade, these proceedings.