Prepared for the U.S. Department of Energy under Contract Number DE-AC03-76SF00098 OCR Output

i- tttt tuué?§~ A f §ii— i.z"”¤ia

t

.t,`,. ..

W

*€era; §a " `=E·` **fg? aiixé

tt.

§W ag i my2*%xs F I %* §

‘ ~gi

th

1=== *

Peeaaease

CERN LIBRHRIES, G;r~{;v_@

January 1994

W.W. Moses, I. Kipnis, and M.H. H0

Photodiode Array Based Pet Detector ModuleA 16-Channel Charge Sensitive Amplifier IC for a Pin

Submitted to IEEE Transactions on Nucleax Science

*3 (JJ $ W1-E

U3 2, E ‘+?¥’+

UNIVERSITY OF CALIFORNIA

PreprintUC-406

LBL-24874W;) {

and R01 NS29655.by Public Health Service Grant Nos. P01 25840, R01 CA48002, side of the photodiode array substrate. OCR OutputEnergy under Contract N0. DE-AC03-76SF00098, and in part

the crystal of interaction) were also placed on the backThis work was supported in part by the U.S. Department oftion of the crystal of interaction, readout of the address ofplified if additional circuitry (e.g. calibration, identificaarea is desirable, as the tomograph design would be simphotodiode element coupled to that crystal. The photo

the photomultiplier tube and some is detected by the than a photodiode element ($0.09 cm, ). Even smallersite side. This requires that the each amplifier be smallerBGO crystal, some of the scintillation light is detected bysame substrate as the photodiode array, but on the oppoWhen a 511 keV annihilation photon interacts in a

An attractive solution is to place the amplifiers on therect side of the substrate for visibility.connecting them (each lead must be <2 pF).faces the BGO crystal array — it is displayed on the incorphotodiodes to minimize the capacitance from the leadssensitive side of the photodiode array in Figure 1 actuallybe practical. The amplifiers also must be close to theother 3x3 mm face to a photomultiplier tube. The photomagnitude size reduction is necessary for this design toon one 3x3 mm face to a silicon photodiode and on thewould have 512 detector modules, a several order of3 mm square by 30 mm deep BGO crystals, each coupledmodule requires 64 amplifiers and a typical PET cameraFigure 1, consists of an 8 by 8 array of optically isolated

The proposed PET detector module, shown in 55 cm! of printed circuit board area. As each detectorwith discrete electrical components [4] that occupiednoise charge sensitive amplifiers on a 2 mm square die.used a low noise charge sensitive amplifier constructedcircuit developed for this application containing 16 low

The design studies for the detector in Figure 1 [1-3]todiode array. In this paper we describe an integratednoise and the size of the amplifier that services the pho

2. DEVICE REQUIREMENTSelements (64 per square inch) place high demands on the511 keV photon interaction) and high density of heights to those observed in the photomultiplier tube.[3]. The low signal level (approximately 700 e‘ per 511 keV photon interacts by comparing their pulse511 keV photon interaction on an event by event basis

BGO crystal (in the 30 mm long dimension) at which theto use the photodiodes to measure the depth of the

module the photodiodes also measure the position in thecrystal of interaction [1, 2], and have extended the design identify the crystal of interaction. In the advancedthat uses a silicon PIN photodiode array to identify the

of the detector module, the photodiodes are used only toWe have previously proposed a PET detector module

energy threshold is 250-350 keV). In the simple versionis 2 ns fwhm) and initial energy discrimination (typical1. INTRoDucT1oNmultiplier tube provides a timing pulse (typical accuracy

photodiodes (3 pF, 200 pA per 3x3 mm element).fwhm can nearly be met with 300 um depletion thickness and may also be used to measure the depth of interaction.

to a photodiode, which identifies the crystal of interactiontime in us. With this amplifier, the target noise of 300 e`provides a timing pulse and initial energy discrimination, andtor dark current in pA and T is the amplifier shapingBGO crystal is attached to a photomultiplier tube, whichthe shot noise (e‘ fwhm) is 16.4a/IT , where I is the detec Figure 1: Exploded view of the proposed PET module. Each

173+13C, where C is the input capacitance in pF, andJohnson noise (e` fwhm at 1 us peaking time) isdc current applied to the 2 mm square chip. The

3 mm s uaredetector current compensation are each controlled by a 7<— 64 BGO Crystalsrise time, fall time, bias current through the first FET, and

mance is optimized for 10 pF detector capacitance. Theoutput of 100 mV per 1000 e‘ input, and its noise performade using 1.2 um CMOS technology, produces anidentify the crystal of interaction. The amplifier, which ismodule that uses an 8x8 array of PIN photodiodes to

Photodlodespreamplifier has been developed for a PET detector

Array of 64A 16-channel integrated circuit charge sensitive

Abstract 1" Square Photomultipller Tube

Lawrence Berkeley Laboratory, University of California, Berkeley, CA 94720W. W. Moses, I. Kipnis, and M. H. H0

PIN PHOTODIODE ARRAY BASED PET DETECTOR MODULE

A 16-CHANNEL CHARGE SENSITIVE AMPLIFIER IC FOR A

LBL-34874Submitted t0 IEEE Transactions 0n Nuclear Science

shaping amplifier channel. OCR OutputFigure 3: Simplified circuit diagram of a single preamplifier /external current source via a current mirror (IM] ).

this input transistor, so this current is controlled by an iM1 Vsspower dissipation depend on the bias current through

i) 0) 0) IU)0for a detector capacitance of 10 pF. Both noise and

ireset I2] resetto n-channel transistors, and its capacitance is optimized TestM] because of its reduced white and 1/f noise comparedload. A p-channel device is used for the input transistor C =a·| | M2] ]M3|_E‘. T i icommon—source cascode amplifier with a cascode active v * y——terz o ~[.Jl1T¤~F* Ip rz Htchannel. The preamplifier input stage is a single stage

Figure 3 shows a simplified circuit diagram of oneOutW ti q ]C lv

of the device is shown in Figure 2. ifa12* raullnsa111:* 0) I 0) 0) U) 0) (00fills approximately half of a 2 x 2 mm die. A photograph

fabricated with a standard 1.2 um CMOS process, andtest pulse capacitor coupled to its input. The circuit is(without shaping) of 100 mV per 1000 e", and a 0.16 pFing amplifier capable of driving loads >10 kQ, a gain providing power, the four control currents (rise time, falllow noise, charge sensitive preamplifier, a RC—CR shap module controls the integrated circuit amplifier array by

The integrated circuit has 16 channels, each with a larger components. A custom single width CAMACdetector system [8]. and wire bonded to traces that allow connection tohas subsequently been modified for a synchrotron x-ray circuit board to avoid 1/f noise due to lossy dielectrics [9]adjustable via external dc current sources. This design The integrated circuit is mounted on a Teflon printedthe first FET, and detector dark current sink) are

4. AMPLIFIER PERFoRMANcEfour parameters (rise time, fall time, bias current throughmeet the requirements specified above. To ease testing,

this shaping amplifier, which is 250 fF /20 fF = 12.5.detector module, one such design [6, 7] was modified toratio CS2/Cf; determines the unfiltered voltage gain ofcapacitance) are similar to the requirements for this PETcurrent sources via current mirrors Imc and Ira];. Thepower dissipation, detector dark current, and detectorconstants are controlled from 500 ns to 50 us by externalparameters for these amplifiers (shaping time, gain,tor is used because of its higher gain. The shaping timedeveloped for silicon microstrip detectors [5]. As thethe main difference being that an n-channel input transisMany integrated circuit amplifier arrays have beenshaped by a circuit of the same type as the preamplifier,

3. AMPL11=1ER DEs1GN The output of the preamplifier stage is amplified andsource via a current mirror (lrcsct).

tion decay time of BGO) in order to minimize dead time. time constant that is controlled by an external current(subject to a 0.5 ps minimum due to the 0.3 us scintilla plifier to a reference voltage Vmf, differentiating with anoise minimization, but it will also be as short as possible baseline restoration by forcing the output of the preamThe main consideration for selecting the shaping time is 50 mV/fC. A differential amplifier (M2, M3) provides dcmodules has not yet been determined, but will be =1 us. back capacitor C; gives the preamplifier a gain ofof 200 pA are expected. The shaping time for the detector "active" cascode circuits, but insertion of a 20 fF feedthese conditions, a capacitance of 3 pF and dark current Very high open-loop gain (>100 dB) is achieved usingof 100 V, and an operating temperature of +25° C. Underdepletion thickness of 300 um, a maximum bias voltage

circuitry can be added if necessary.active area of each element is 2.8 mm square, with acircuit amplifier chip. The chip is roughly half filled, so more

its electrical properties are not known. The specified Figure 2: Photograph of the 2 mm wide, 16 channel integratedPIN photodiode array has not yet been manufactured, so

The design specifications are uncertain because theE on

todiode combination must be $300 e‘ fwhm.700 e", implying that the noise from the amplifier / phofrom a 511 keV energy deposit in the BGO crystal isnoise fluctuations [2]. The average photodiode signaltively eliminate events that are mis—identified because ofnoise ratio of 22.25:1 (where noise is in fwhm) will effecwork showed that for a 64 element detector, a signal tocorrectly identify the crystal of interaction. Previousfier combination are set by the ability 0f the module t0

The noise requirements for the photodiode / ampli· $*5 *%*5%

LBL-34874

4.0 127 + 7.3 C (DF reset current source. The total noise at 2 us shaping time OCR Output2.0 142 + 9.2 C (DF with a large feedback resistor because of the noise in the

__ 1.0 173+13C(DF is larger than the l1.7~/TT expected from an amplifierL_ 0.75 188 + 15 C (DF and T is the shaping time in ps. As explained earlier, this_ 0.5 220 + 18 C (DF l6.4~/F , where I is photodiode dark current Ipd in pA

further. The shot noise (in e‘ fwhm) is expected to bePeaking Time | Amplifier Noise (e" fwhm)current until the amplifier saturates if Ircsct is reducedthe photodiode dark current Ipd by reducing the resetfunction of capacitance (in pF) for several peaking times.

The reset current Inset is adjusted to compensate forTable 1: Straight line fit to amplifier noise (in e‘ fwhm) as aelement of the photodiode array).opposed to the 3 pF and ==200 pA expected for eachfrom the printed circuit board dielectric.capacitance is 9 pF and the dark current is =100 pA (as

not included in this fit as it does not contain the noiseroom temperature (+25° C). Under these conditions, thethe results displayed in Table 1. The data point at 0 pF isdiode is biased with +30 V and the assembly operated atThe data at each shaping time are fit to a straight line, and300 um expected for the photodiode array). The photoplotted for several amplifier peaking times in Figure 4.and the depletion thickness is 100 um (as opposed to theinput to the trace). The results of these measurements arecrystal. The active area of this device is 2.77 mm square,by removing the bond wire connecting the amplifierspecial package to allow close coupling to the scintillatortached to this trace (measurements at 0.0 pF are obtainedPIN photodiode is a Hamamatsu S-2506 mounted in atrace and the measured value of the CK—05 capacitor atthe opposite end coupled to a photomultiplier tube. Themeasured 0.9 pF capacitance of the printed circuit boardone end coupled to a 3 mm square PIN photodiode andnal capacitor, where the capacitance quoted includes themodule with a single 3x3x30 mm BGO crystal having

This measurement is made for several values of exterThe amplifier array is evaluated by constructing a testunit conversion from pulse height bins to e'

design using a feedback resistor.the position of the centroid of the test pulse peak allowsshot noise is increased by a factor of =~/iover thecomputer. The fwhm of the distribution is calculated andcurrent Ipd and the nulling current Ircsct = I d) so the panalog to digital converter (ADC) read out by a personalindependent sources of shot noise (the photodiode darklating the resulting pulse height spectrum using andisadvantage of this method is that there are two2 mV step to the 0.16 pF test capacitor Ccal, and accumuof Ircsct is adjusted to be slightly greater than Ipd. The(<30 pA), injecting a 2000 e- test pulse by applying ais able to supply (but not sink) current, so the amplitudeload by setting the detector current null to a small valueImsct that is controlled externally. The preamplifier inputThe noise is measured as a function of the capacitivewith a negative dc current source (i.e. a current sink)current through this device (M1) is approximately 1 mA.amplifier nulls the dark current as shown in Figure 5(b)dissipates most of the power in the circuit, the biasfabricate in an integrated circuit, so this integrated circuitconsumption is 5 mW per channel. As the first transistorJohnson noise. Resistor values this large are difficult tohere, the operating voltage is 5 V and the poweras large as possible (1 GQ typically) to minimize theeach of the 16 outputs. For all measurements quotedJohnson noise is amplified, so the resistor value must becurrent null), as well as buffering and amplifying (x10)back resistor is connected to the amplifier input, itstime, bias current through the first FET, and detectorand T is the amplifier shaping time in us. Since this feedof ll.7m, where I is the detector dark current in pA

tance for several values of amplifier peaking time. Ipd as shown in Figure 5(a), yielding a noise (in e` fwhm)Figure 4: Amplifier noise (in c` fwhm) versus input capaci

amplifier feedback resistor Rfb sinks the dark currentCapacitance (pF) dark current. In the discrete component amplifier, the

O 2 4 6 8 10 12 14 dark current depends on the method used to sink theThe shot noise contribution from the photodiode

100 with a constant current source Imc,(a) through a large (1 GQ) feedback resistor Rn, or (b) nullingFigure 5: Methods for sinking the photodiode dark current:

200 -.,.4;.9.%2.0 ps (8) __V!: (b) @_ reset

300 pd l/Ft jtb lp? I4 j1.0 Us0:75 ps

4000.5 ps

biasYbias500

LBL-34874Submitted t0 IEEE Transactions 0n Nuclear Science

photodiodes (expected capacitance of 3 pF and darka fwhm) can be met with 300 um depletion thickness

IEEE Trans. Nucl. Sci. NS-20: pp. 182-187, 1973.of 2.25:1 signal to noise ratio (where noise is measured as[9] Radeka V. Field effect transistors for charge amplifiers.is sufficiently low at 1 us shaping time that the design goal

Nucl. Sci. NS-41: (submitted for publication), 1994.module excited by 511 keV photons. The predicted noisenoise, x-ray silicon strip detector system. IEEE Trans.pulsing with capacitive loads and with a test detector

[8] Ludewigt B, Jaklevic J, Kipnis I, et al. A high rate, lowdiode array. The amplifier is characterized both by test

1120-1126, 1990.be mounted on the back of the 64 element, 1 ini photoCDF silicon vertex detector. IEEE Trans. Nucl. Sci. pp.array and additional readout and control electronics can prototyping of the front end readout system for the

[7]channel is small enough (0.00125 cm2) that the amplifier Haber C, Carithers WC, Ely RP, et al. Design andcurrents applied to the 2 mm square chip. Each amplifier Nucl. Sci. NS-35: pp. 171-175, 1988.detector current compensation) are controlled by dc grated circuit with sparse data readout. IEEE Trans.time, fall time, bias current through the first FET, and 128 channel silicon strip detector instrumentation inte

[61 Kleinfelder SA, Carithers WC, Ely RP, et al. A flexibledepth of interaction. Four operating parameters (rise

1988.tify the crystal of interaction and possibly measure thephysics. IEEE Trans. Nucl. Sci. NS-35: pp. 146-150,module that uses an array of silicon photodiodes to iden

[5] Williams HH. Fast analog integrated circuits for particlefier array has been developed as part of a PET detector1084-1089, 1989.A low noise, charge sensitive integrated circuit amplidetector for PET. IEEE Trans. Nucl. Sci. NS-36: pp.characterization of a position-sensitive photodiode/BGO5. CoNc1.Us1oNs

[4] Derenzo SE, Moses WW, Jackson HG, et al. Initial

this additional source of noise. (submitted for publication), 1994.depth of interaction. IEEE Trans. Nucl. Sci. NS-41:current. More work is necessary to identify and eliminatedetector module using a PIN photodiode to measurepredicted from the photodiode capacitance and dark

[3] Moses WW and Derenzo SE. Design studies for a PETrated from the noise, the noise is significantly larger than

Trans. Nucl. Sci. NS-40: pp. 1036-1040, 1993.the amplifier output. Although the signal is well sepaphotodiodes to identify the crystal of interaction. IEEE

removing the excitation source and randomly digitizing a PET detector module utilizing an array of siliconshown in Figure 5 is the noise distribution, obtained by [2] Moses WW, Derenzo SE, Nutt R, et al. Performance ofclear peak centered at 650 e‘ with a fwhm of 470 e`. Also tion. J. Nucl. Med. 33: pp. 862, 1992.resulting pulse height spectrum, shown in Figure 5, has a temperature silicon photodiodes for crystal identificacorresponding to an energy deposit 2250 keV. The a high-rate, high-resolution PET module using room

[1] Moses WW, Derenzo SE and Budinger TF. Design forwhenever the photomultiplier tube detects a signalfrom a Ge source, and the amplifier output digitized68

RE1=ERENcEsThe BGO crystal is irradiated with 511 keV photonswhen test pulsing with this photodiode connected.

P01-HL25840, No. R01-CA48002, and No. R01—NS29655.which is significantly less than the 425 e‘ fwhm measured

Neurological Disorders and Stroke under grants No.noise from 100 pA dark current added in quadrature),

National Cancer Institute, and National Institute of9 pF capacitive load and 164 e‘ fwhm due to the shotHealth, National Heart, Lung, and Blood Institute,

is expected to be 280 e* fwhm (225 e’ fwhm due to theAC03-76SF00098, in part by the National Institutes ofU.S. Department of Energy under contract No. DE

511 keV photons and when triggered randomly. Applications and Biophysical Research Division of thephotodiode combination when the detector was excited with

Health and Environmental Research, MedicalFigure 5: Pulse height spectrum measured by the amplifier /part by the Director, Office of Energy Research, Office of

Pulse Height (e`) that this work was possible. This work was supported in0 500 1000 1500 2000 of CTI, Inc. and Lawrence Berkeley Lab for suggesting

support, and Dr. D. Binkley, Dr. R. Nutt, and Dr. N. Roe

500 Lawrence Berkeley Lab for invaluable technical

We thank Mr. T. Vuletich and Mr. T. Merrick of0 1¤<>¤ F ·—---——-- ii -—-- #470e* e--» Ht -~—~ I MGB ExcitationACKNOWLEDGMENTS5 t5¤¤ i ---~-—

U, 2000 signal to noise ratio of 2.25:1 to be achieved.E 2500 source of noise that must be eliminated for the design

0t.t.t,, background, these photodiodes have an additionalphoton interactions in BGO is clearly separated from

3500 indicate that while the signal resulting from 511 keV?[?EF??T??T'ti$,...§ __,t.t....____.____, t .,.tt.._.,.._...,,_ ¤¤<>¤·—E—E»~—-——Et——» EM-~L22%;$°¥”P»i2$§?J?$iZ4000 current of 200 pA). Measurements with photodiodes

Submitted to IEEE Transactions 0n Nuclear Science

LBL-34874