Download - Charge-coupled-device charge-collection efficiency and the ...hosting.astro.cornell.edu/academics/courses/astro310/Books/Janesick_PhotonTransfer...Charge-coupled-device charge-collection

Transcript

Charge -coupled- device charge -collection efficiency and thephoton- transfer technique

James R. JanesickKenneth P. KlaasenTom ElliottJet Propulsion LaboratoryCalifornia Institute of Technology4800 Oak Grove DrivePasadena, California 91109

CONTENTS

1. Introduction2. Charge -collection efficiency (CCE)3. Photon- transfer technique

3.1. Ideal CCD camera3.2. Evaluation of constant K3.3. Evaluation of constant J3.4. Partial and split events included3.5. Photon -transfer curve3.6. Photon -transfer histogram

4. Photon -transfer use4.1. Frontside illumination (TI VPCCD)4.2. Backside illumination (TI 3PCCD)

5. Future improvements in CCE6. Acknowledgments7. References

Abstract. The charge -coupled device has shown unprecedented performanceas a photon detector in the areas of spectral response, charge transfer, andreadout noise. Recent experience indicates, however, that the full potential forthe CCD's charge -collection efficiency (CCE) lies well beyond that realized incurrently available devices. In this paper we present a definition of CCE perfor-mance and introduce a standard test tool (the photon- transfer technique) formeasuring and optimizing this important CCD parameter. We compare CCEcharacteristics for different types of CCDs, discuss the primary limitations inachieving high CCE performance, and outline the prospects for futureimprovement.

Subject terms: charge- coupled devices; charge diffusion; x -ray events; frontside illumina-tion; backside illumination.

Optical Engineering 26(10), 972 -980 (October 1987).

1. INTRODUCTIONCCDs in recent years have become the premier detector foruse in many spaceborne and ground -based astronomicalinstruments. They were selected for use in the Hubble SpaceTelescope Wide Field Planetary Camera (W F/ PC), the Gali-leo Jupiter Orbiter's Solid State Imager (SSI), and manyground -based imaging and spectroscopic applications. Pro-posed space applications include an x -ray imager on NASA'sAdvanced X -ray Astronomical Facility (AXAF), a SpaceTelescope Imaging Spectrometer (SIS), the Solar Optical Tele-scope (SOT), and the Comet Rendezvous/ Asteroid FlybyImaging Subsystem (CRAF ISS).

Invited Paper CH -102 received April 13, 1987; revised manuscript receivedMay 15, 1987; accepted for publication June 19, 1987; received by ManagingEditor July 6, 1987. This paper is a revision of Paper 570 -02, presented at theSPIE conference Solid State Imaging Arrays, Aug. 22 -23, 1985, San Diego,Calif. The paper presented there appears (unrefereed) in SPIE ProceedingsVol. 570.e 1987 Society of Photo -Optical Instrumentation Engineers.

972 / OPTICAL ENGINEERING / October 1987 / Vol. 26 No. 10

The fundamental parameters that ultimately limit CCDperformance are (1) read noise, (2) charge- transfer efficiency(CTE), (3) quantum efficiency (QE), and (4) charge -collectionefficiency (CCE). At their present stage of development, it ispossible to fabricate devices that have low read noise (in the 4to 15 a range), excellent CTE performance (<10 a deferredcharge), and unsurpassed QE performance over the entirespectral range from 1 to 11,000 A.1 -3 However, the full poten-tial of charge -collection efficiency lies well beyond that ofcurrently available devices. Optimization of this importantparameter represents a new challenge for the CCD manufac-turer and user. High CCE performance is required for manyapplications over all regions of the spectrum to which theCCD is sensitive. In the visible range, for example, CCDs areused in star trackers that demand high sensitivity (chargecollection without loss) in conjunction with high geometricaccuracy (collection without significant charge diffusion). Inthe x -ray and EUV regions of the spectrum, applicationsrequire confinement of signal charge to a single pixel withoutloss in order to accurately determine the energy of the incom-ing photon.

The means of measuring and achieving high CCE perfor-mance is the subject of this paper. In Sec. 2 we present a usefuldefinition for CCE performance in terms of parameters thatare readily found when testing the CCD. The definition isdivided into the two primary factors that are responsible forthe degradation of CCE performance, namely, charge loss andcharge diffusion. In Sec. 3 we introduce the concept of photontransfer, a technique used as a standard way of measuringCCE characteristics, and develop the theoretical foundationsupon which the photon- transfer method is based. We showthe strengths and limitations of the photon- transfer techniqueas it is used in measuring CCE characteristics of the CCD. InSec. 4 we apply the photon -transfer technique in measuring

Charge-coupled-device charge-collection efficiency and the photon-transfer technique

JamesR. Janesick Kenneth P. Klaasen Tom ElliottJet Propulsion Laboratory California Institute of Technology 4800 Oak Grove Drive Pasadena, California 91109

Abstract. The charge-coupled device has shown unprecedented performance as a photon detector in the areas of spectral response, charge transfer, and readout noise. Recent experience indicates, however, that the full potential for the CCD's charge-collection efficiency (CCE) lies well beyond that realized in currently available devices. In this paper we present a definition of CCE perfor­ mance and introduce a standard test tool (the photon-transfer technique) for measuring and optimizing this important CCD parameter. We compare CCE characteristics for different types of CCDs, discuss the primary limitations in achieving high CCE performance, and outline the prospects for future improvement.

Subject terms: charge-coupled devices; charge diffusion; x-ray events; frontside illumina­ tion; backside illumination.

Optical Engineering 26(10), 972-980 (October 1987).

CONTENTS1. Introduction2. Charge-collection efficiency (CCE)3. Photon-transfer technique

3.1. Ideal CCD camera3.2. Evaluation of constant K3.3. Evaluation of constant J3.4. Partial and split events included3.5. Photon-transfer curve3.6. Photon-transfer histogram

4. Photon-transfer use4.1. Frontside illumination (TI VPCCD)4.2. Backside illumination (TI 3PCCD)

5. Future improvements in CCE6. Acknowledgments7. References

1. INTRODUCTIONCCDs in recent years have become the premier detector for use in many spaceborne and ground-based astronomical instruments. They were selected for use in the Hubble Space Telescope Wide Field Planetary Camera (WF/PC), the Gali­ leo Jupiter Orbiter's Solid State Imager (SSI), and many ground-based imaging and spectroscopic applications. Pro­ posed space applications include an x-ray imager on NASA's Advanced X-ray Astronomical Facility (AXAF), a Space Telescope Imaging Spectrometer (SIS), the Solar Optical Tele­ scope (SOT), and the Comet Rendezvous/Asteroid Flyby Imaging Subsystem (CRAF ISS).

Invited Paper CH-102 received April 13, 1987; revised manuscript received May 15, 1987; accepted for publication June 19, 1987; received by Managing Editor July 6,1987. This paper is a revision of Paper 570-02, presented at the SPIE conference Solid State Imaging Arrays, Aug. 22-23, 1985, San Diego, Calif. The paper presented there appears (unrefereed) in SPIE Proceedings Vol. 570. © 1987 Society of Photo-Optical Instrumentation Engineers.

The fundamental parameters that ultimately limit CCD performance are (1) read noise, (2) charge-transfer efficiency (CTE), (3) quantum efficiency (QE), and (4) charge-collection efficiency (CCE). At their present stage of development, it is possible to fabricate devices that have low read noise (in the 4 to 15 e~ range), excellent CTE performance (<10 e~ deferred charge), and unsurpassed QE performance over the entire spectral range from 1 to 11,000 A. 1 - 3 However, the full poten­ tial of charge-collection efficiency lies well beyond that of currently available devices. Optimization of this important parameter represents a new challenge for the CCD manufac­ turer and user. High CCE performance is required for many applications over all regions of the spectrum to which the CCD is sensitive. In the visible range, for example, CCDs are used in star trackers that demand high sensitivity (charge collection without loss) in conjunction with high geometric accuracy (collection without significant charge diffusion). In the x-ray and EUV regions of the spectrum, applications require confinement of signal charge to a single pixel without loss in order to accurately determine the energy of the incom­ ing photon.

The means of measuring and achieving high CCE perfor­ mance is the subject of this paper. In Sec. 2 we present a useful definition for CCE performance in terms of parameters that are readily found when testing the CCD. The definition is divided into the two primary factors that are responsible for the degradation of CCE performance, namely, charge loss and charge diffusion. In Sec. 3 we introduce the concept of photon transfer, a technique used as a standard way of measuring CCE characteristics, and develop the theoretical foundations upon which the photon-transfer method is based. We show the strengths and limitations of the photon-transfer technique as it is used in measuring CCE characteristics of the CCD. In Sec. 4 we apply the photon-transfer technique in measuring

972 / OPTICAL ENGINEERING / October 1987 / Vol. 26 No. 10

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/21/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

CHARGE -COUPLED -DEVICE CHARGE -COLLECTION EFFICIENCY AND THE PHOTON- TRANSFER TECHNIQUE

CCE performance for frontside- and backside -illuminatedCCDs and discuss the primary factors that ultimately limitCCE for each device. Finally, in Sec. 5 we discuss futureconsiderations for further improving CCE for the CCD.

2. CHARGE -COLLECTION EFFICIENCY

CCE is a relatively new CCD performance parameter that hasbeen defined, measured, and optimized at Jet PropulsionLaboratory (JPL) and elsewhere. CCE measures the ability ofthe CCD to collect all signal charge generated from a singlephoton event into a single pixel. High CCE performance isespecially critical for EUV and soft x -ray applications (e.g.,soft x -ray imaging spectrometers), where the ability of theCCD to accurately determine the energy of the photondepends upon collecting the photogenerated charge properly.Experience has shown that complete charge collection requiresthat two criteria be met: (1) There must be no trapping centerswithin the CCD to cause signal charge to be lost by recombi-nation, and (2) the charge of an individual photon must becollected within a single pixel and must not be allowed todivide among several pixels. Charge loss causes the photonenergy to be underestimated, while charge splitting degradesthe precision of charge measurement by requiring the summa-tion of several noisy pixels.

The degree of charge loss and charge splitting dependsupon where in the pixel the photon is absorbed. Photons thatare absorbed within the frontside depletion region (see Figs. 5and 8) of a given pixel are typically seen as the ideal event andare called "single -pixel events." Photons absorbed below thedepletion region, where the electric field is weaker, create acharge cloud that thermally diffuses outward until it reachesthe rapidly changing potential wells at the lower boundary ofthe pixel array. At that point, the charge cloud may split intotwo or more packets, which are collected in adjacent pixels.Events of this type are called "split events." Events in whichcharge is not conserved have been named simply "partialevents" and are usually generated in regions deep within theCCD, where loss of carriers through recombination occurs.

From this discussion, a definition for CCE for an individ-ual photon event I can be presented through the formula

CCEI =77; Pse-1

(1)

where CCEI represents the fraction of signal electrons, gener-ated by a particular interacting photon I, that is collected inany single affected pixel; bpe_I refers to the partial event andrepresents the number of signal carriers generated by a photonand collected by all pixels (the rest being lost to recombina-tion); PSe_1 refers to the split event and represents the numberof pixels that collect signal electrons generated by a photon;and m is defined as the ideal quantum yield, a quantity equalto the total number of electrons generated for an interactingphoton of energy EA (eV). The ideal quantum yield i isdirectly proportional to the photon energy and is foundaccording to the relationship

E,,

I)i 3.65(À<1000 A) . (2)

As an example of using Eq. (1), assume that an interactingphoton generates 1000 e (m), with 200 e lost to recombina-

tion and the 800 a remaining (rspe_1) split between and col-lected by two pixels (1 se_1). For this event, a CCE1 of 0.4 iscalculated no matter in what proportion the 800 a are splitbetween the two affected pixels.

To determine the average CCE performance of a CCD fora large number of interacting photons of the same energy,many events are measured for charge loss and splitting andthen averaged using the equation

CCE _N

i=

Spe - I

Pse -I(3)

where N is the number of photon events sampled.Equation (3) is used regularly in the laboratory in charac-

terizing the two mechanisms (the partial and split events)responsible for degrading CCE performance of the CCD.However, measuring CCE in the manner described by Eq. (3)requires a considerable amount of data reduction. since manyevents must be integrated. Also, Eq. (3) is usable over only alimited spectral region (typically, A < 30 A) because forlonger wavelengths the signal generated by an individualphoton becomes too small compared to the CCD read noisefloor to reliably resolve the individual event and determine theamount of charge lost and the number of pixels affected.

In this paper we describe another approach to evaluatingCCE performance for the CCD that is applicable to all wave-lengths of interest. The new technique (discussed in Sec. 3) isbased on the formula

ECCE=-' (4)

where CIE is called the effective quantum yield, a quantity thatmeasures the average number of electrons collected by anaffected pixel for an interacting photon of energy Ex. Theeffective quantum yield r%E is related to the partial and splitevents through

pe

EPse

where Spe/ PSe is the average value of the term ape -i /Pse

(5)

3. PHOTON -TRANSFER TECHNIQUEThe ideal CCD, which does not generate split or partial eventsbut exhibits perfect CCE performance, will deliver an effec-tive quantum yield equal to the ideal quantum yield (i.e.,rlE = rl;). Today's CCDs are rapidly progressing toward thisultimate goal; however, very strict conditions are placed onthe CCD in obtaining such performance, as we shall see inSec. 4. Because of the various CCD technologies and manu-facturers involved in fabricating CCDs, a standard "test tool"for evaluating CCE performance over a very large spectralrange is required.

In this section we discuss the technique of photon transfer,a test tool that was used in the past to evaluate CCD perfor-mance characteristics in absolute units.' It was realized onlyrecently that the photon -transfer technique also can be ap-plied as a standard method for evaluating the CCE perfor-mance of a CCD. In the discussion that follows, we firstdevelop the equations necessary to describe the technique,

OPTICAL ENGINEERING / October 1987 / Vol. 26 No. 10 / 973

CHARGE-COUPLED-DEVICE CHARGE-COLLECTION EFFICIENCY AND THE PHOTON-TRANSFER TECHNIQUE

CCE performance for frontside- and backside-illuminated CCDs and discuss the primary factors that ultimately limit CCE for each device. Finally, in Sec. 5 we discuss future considerations for further improving CCE for the CCD.

2. CHARGE-COLLECTION EFFICIENCY

CCE is a relatively new CCD performance parameter that has been defined, measured, and optimized at Jet Propulsion Laboratory (JPL) and elsewhere. CCE measures the ability of the CCD to collect all signal charge generated from a single photon event into a single pixel. High CCE performance is especially critical for EUV and soft x-ray applications (e.g., soft x-ray imaging spectrometers), where the ability of the CCD to accurately determine the energy of the photon depends upon collecting the photogenerated charge properly. Experience has shown that complete charge collection requires that two criteria be met: (1) There must be no trapping.centers within the CCD to cause signal charge to be lost by recombi­ nation, and (2) the charge of an individual photon must be collected within a single pixel and must not be allowed to divide among several pixels. Charge loss causes the photon energy to be underestimated, while charge splitting degrades the precision of charge measurement by requiring the summa­ tion of several noisy pixels.

The degree of charge loss and charge splitting depends upon where in the pixel the photon is absorbed. Photons that are absorbed within the frontside depletion region (see Figs. 5 and 8) of a given pixel are typically seen as the ideal event and are called "single-pixel events." Photons absorbed below the depletion region, where the electric field is weaker, create a charge cloud that thermally diffuses outward until it reaches the rapidly changing potential wells at the lower boundary of the pixel array. At that point, the charge cloud may split into two or more packets, which are collected in adjacent pixels. Events of this type are called "split events." Events in which charge is not conserved have been named simply "partial events" and are usually generated in regions deep within the CCD, where loss of carriers through recombination occurs.

From this discussion, a definition for CCE for an individ­ ual photon event I can be presented through the formula

CCE, = (1)

where CCEj represents the fraction of signal electrons, gener­ ated by a particular interacting photon I, that is collected in any single affected pixel; ^e _j refers to the partial event and represents the number of signal carriers generated by a photon and collected by all pixels (the rest being lost to recombina­ tion); Pse _, refers to the split event and represents the number of pixels that collect signal electrons generated by a photon; and Tjj is defined as the ideal quantum yield, a quantity equal to the total number of electrons generated for an interacting photon of energy Ex (eV). The ideal quantum yield r/j is directly proportional to the photon energy and is found according to the relationship

3.65(A<1000A) . (2)

tion and the 800 e remaining Upe-j) split between and col­ lected by two pixels (Pse _,). For this event, a CCEj of 0.4 is calculated no matter in what proportion the 800 e~ are split between the two affected pixels.

To determine the average CCE performance of a CCD for a large number of interacting photons of the same energy, many events are measured for charge loss and splitting and then averaged using the equation

(3)

where N is the number of photon events sampled.Equation (3) is used regularly in the laboratory in charac­

terizing the two mechanisms (the partial and split events) responsible for degrading CCE performance of the CCD. However, measuring CCE in the manner described by Eq. (3) requires a considerable amount of data reduction-since many events must be integrated. Also, Eq. (3) is usable over only a limited spectral region (typically, X < 30 A) because for longer wavelengths the signal generated by an individual photon becomes too small compared to the CCD read noise floor to reliably resolve the individual event and determine the amount of charge lost and the number of pixels affected.

In this paper we describe another approach to evaluating CCE performance for the CCD that is applicable to all wave­ lengths of interest. The new technique (discussed in Sec. 3) is based on the formula

(4)

where r/E is called the effective quantum yield, a quantity that measures the average number of electrons collected by an affected pixel for an interacting photon of energy Ex . The effective quantum yield rjE is related to the partial and split events through

As an example of using Eq. (1), assume that an interacting photon generates 1000 e~ (17}), with 200 e~ lost to recombina­

(5)

where £pe /Pse is the average value of the term £ !/Pse _j.

3. PHOTON-TRANSFER TECHNIQUE

The ideal CCD, which does not generate split or partial events but exhibits perfect CCE performance, will deliver an effec­ tive quantum yield equal to the ideal quantum yield (i.e., r;E = rjj). Today's CCDs are rapidly progressing toward this ultimate goal; however, very strict conditions are placed on the CCD in obtaining such performance, as we shall see in Sec. 4. Because of the various CCD technologies and manu­ facturers involved in fabricating CCDs, a standard "test tool" for evaluating CCE performance over a very large spectral range is required.

In this section we discuss the technique of photon transfer, a test tool that was used in the past to evaluate CCD perfor­ mance characteristics in absolute units. 1 It was realized only recently that the photon-transfer technique also can be ap­ plied as a standard method for evaluating the CCE perfor­ mance of a CCD. In the discussion that follows, we first develop the equations necessary to describe the technique,

OPTICAL ENGINEERING / October 1987 / Vol. 26 No. 10 / 973

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/21/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

INCIDENTI

PHOTONSI

PI

QE1 NE

INTERACTING

PHOTONS

INCIDENTPHOTONS

QE

ELECTRONS

COLLECTED

INTERACTING

PHOTONS

L R

JANESICK, KLAASEN, ELLIOTT

SV

7

I

_J

Al A2

DIGITALNUMBERVOLTS VOLTS

VOLT

DIGITALNUMBER

ELECTRONS

TRANSFERREDVOLT

SIDNI

CCD

SENSITIVITYSIGNAL A/DC

CHAIN GAINGAIN

CC D

Fig. 1. Schematic showing individual transfer functions of an idealCCD camera.

assuming that we have an ideal CCD camera with no partialor split event generation. We show that the ideal quantumyield i can be determined through the photon- transferapproach. We next examine a typical CCD camera, whichincludes partial and split event generation, and show that thephoton- transfer technique gives a reasonable approximationfor the effective quantum yield 77E, defined in Eq. (5), which inturn is used to calculate the CCE performance of the CCD[Eq. (4)], at least in a relative sense.

3.1. Ideal CCD cameraFigure 1 is a schematic representation of the overall transferfunction of an ideal CCD camera. The camera can be de-scribed in terms of five transfer functions, three that arerelated to the CCD and two that are related to the externalCCD signal processing circuitry. The input to the camera isgiven in units of incident photons, and the final output of thecamera is achieved by encoding each pixel's signal into adigital number (DN), typically using 12 to 16 bits. The outputsignal S (DN) resulting from a given exposure of the CCDcamera shown in Fig. 1 is given by

S (DN) = PQE0);SvAlA2 , (6)

where S (DN) represents the average signal (DN) over allaffected pixels, P is the mean number of incident photons perpixel on the CCD, QED is defined as the interacting quantumefficiency (interacting photons/ incident photons), rl is theideal quantum yield defined by Eq. (2), Sv is the sensitivity ofthe CCD on -chip circuitry (V/ el, A, is the electronic gain ofthe camera (V/ V), and A2 is the transfer function of theanalog -to- digital converter (DN/ V).

The quantities QED and m are related through

QE = ThQEI , (7)

where QE is the average quantum efficiency (electrons collec-ted/ incident photon).

To convert the output signal S (DN) into fundamentalphysical units, it is necessary to find the appropriate factors toconvert DN units into either interacting photons or signalelectrons. The constants that do this conversion are defined bythe equations

K = (SvAIA2)-I ,

J = (r);SvAIA2)-1 ,

(8)

(9)

where the units of K and J are e / DN and interacting pho-tons/ DN, respectively. Note that Eqs. (8) and (9) are related

974 / OPTICAL ENGINEERING / October 1987 / Vol. 26 No. 10

through 77; by

(10)

It is possible to determine the factors K and J by measuringeach transfer function in Fig. 1 separately and then combiningthese results as in Eqs. (8) and (9). However, because of theuncertainty in a number of parameters of the CCD (whichprevents us from knowing QED, rl and Sv independently), wecannot in practice directly determine K or J to any greataccuracy. Instead, we have developed a simple technique thatrequires no knowledge of the individual transfer functions todetermine the factors K and J.

3.2. Evaluation of constant K

For the CCD stimulated with photons that generate only oneelectron -hole (e -h) pair for each interaction (i.e., rl =1; X > 3000 A), Eq. (6) reduces to the form

S (DN) = PIK -I , (I l)

where PI = PQE, represents the number of interactingphotons per pixel.

The constant K can be determined by relating the outputsignal S (DN) to its variance, as (DN). The variance as (DN)of Eq. (11) is found using the propagation of errors, whichyields the following equation for the ideal CCD (i.e., perfectcharge collection and charge transfer):

-( a S (DN)]2 r 3S (DN)122

as (DN) L ô PI aPI +[

a K J a1

+ aR (DN) , (12)

where we have added in quadrature the read noise floorvariance aR (DN) [see Fig. 1; aÁ (DN) = aR (e-) K -2].

Performing the required differentiation on Eq. (12) andassuming that the constant K has negligible variance (i.e.,aK = 0), we find the following expression for the variance inS (DN):

2

as (DN) = (-K I (DN) . (13)

Since aPl = PI because of photon statistics, the followingequation for the constant K in terms of S (DN) and as (DN)results:

S (DN)K as (DN) -4 (DN) (À>3000 A) (14)

Equation (14) is a useful expression and can be used, with nofurther calibration, to convert output measurements in DNdirectly into units of electrons.

3.3. Evaluation of constant JFor wavelengths longer than 3000 A, the constants K and Jare equivalent [Eq. (10), rl = 1]. However, as we move intothe UV, EUV, and x -ray regions of the spectrum, multiple e -hpairs are generated by each interacting photon, resulting in

> 1 and a decrease in the value J. For these conditions, the

JANESICK, KLAASEN, ELLIOTT

QE,

INTERACTING PHOTONS

INCIDENT PHOTONS

QE

ELECTRONS COLLECTED

INTERACTING PHOTONS

'?'*R2

VOLTS

ELECTRONS TRANSFERRED

CCD SENSITIVITY

J

VOLTS

VOLT

SIGNAL CHAIN GAIN

DIGITAL NUMBER

VOLT

A/DC GAIN

DIGITAL NUMBER

S(DN)

Fig. 1. Schematic showing individual transfer functions of an ideal CCD camera.

assuming that we have an ideal CCD camera with no partial or split event generation. We show that the ideal quantum yield TJ can be determined through the photon-transfer approach. We next examine a typical CCD camera, which includes partial and split event generation, and show that the photon-transfer technique gives a reasonable approximation for the effective quantum yield rjE , defined in Eq. (5), which in turn is used to calculate the CCE performance of the CCD [Eq. (4)], at least in a relative sense.

3.1. Ideal CCD cameraFigure 1 is a schematic representation of the overall transfer function of an ideal CCD camera. The camera can be de­ scribed in terms of five transfer functions, three that are related to the CCD and two that are related to the external CCD signal processing circuitry. The input to the camera is given in units of incident photons, and the final output of the camera is achieved by encoding each pixel's signal into a digital number (DN), typically using 12 to 16 bits. The output signal S (DN) resulting from a given exposure of the CCD camera shown in Fig. 1 is given by

S(DN) = PQE,r7i SvA,A2 , (6)

where S (DN) represents the average signal (DN) over all affected pixels, P is the mean number of incident photons per pixel on the CCD, QEj is defined as the interacting quantum efficiency (interacting photons/incident photons), rj{ is the ideal quantum yield defined by Eq. (2), Sv is the sensitivity of the CCD on-chip circuitry (V/e~), A, is the electronic gain of the camera (V/V), and A2 is the transfer function of the analog-to-digital converter (DN/ V).

The quantities QE, and 77} are related through

QE = (7)

where QE is the average quantum efficiency (electrons collec­ ted/incident photon).

To convert the output signal S (DN) into fundamental physical units, it is necessary to find the appropriate factors to convert DN units into either interacting photons or signal electrons. The constants that do this conversion are defined by the equations

,A2r ! , (8) ^ A2r 1 » w

where the units of K and J are e~/DN and interacting pho­ tons/ DN, respectively. Note that Eqs. (8) and (9) are related

through Tft by

_ _K

1 J(10)

It is possible to determine the factors K and J by measuring each transfer function in Fig. 1 separately and then combining these results as in Eqs. (8) and (9). However, because of the uncertainty in a number of parameters of the CCD (which prevents us from knowing QEj, rj{ , and Sv independently), we cannot in practice directly determine K or J to any great accuracy. Instead, we have developed a simple technique that requires no knowledge of the individual transfer functions to determine the factors K and J.

3.2. Evaluation of constant K

For the CCD stimulated with photons that generate only one electron-hole (e-h) pair for each interaction (i.e., rj{ = 1; A > 3000 A), Eq. (6) reduces to the form

S(DN) = PIK~ (11)

where PI = PQE, represents the number of interacting photons per pixel.

The constant K can be determined by relating the output signal S (DN) to its variance, a| (DN). The variance a§ (DN) of Eq. (11) is found using the propagation of errors, which yields the following equation for the ideal CCD (i.e., perfect charge collection and charge transfer):

(12)

where we have added in quadrature the read noise floor variance a£ (DN) [see Fig. 1; a£ (DN) - a* (e~) If2].

Performing the required differentiation on Eq. (12) and assuming that the constant K has negligible variance (i.e., cjjt = 0), we find the following expression for the variance in S (DN):

(13)

Since ajsj = PI because of photon statistics, the following equation for the constant K in terms of S (DN) and a§ (DN) results:

K = S(DN)(DN) - a 2 (DN)

(A>3000 A) . (14)

Equation (14) is a useful expression and can be used, with no further calibration, to convert output measurements in DN directly into units of electrons.

3.3. Evaluation of constant J

For wavelengths longer than 3000 A, the constants K and J are equivalent [Eq. (10), rj{ = 1]. However, as we move into the UV, EUV, and x-ray regions of the spectrum, multiple e-h pairs are generated by each interacting photon, resulting in rji > 1 and a decrease in the value J. For these conditions, the

974 / OPTICAL ENGINEERING / October 1987 / Vol. 26 No. 10

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/21/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

CHARGE -COUPLED -DEVICE CHARGE -COLLECTION EFFICIENCY AND THE PHOTON -TRANSFER TECHNIQUE

constant J also can be found by relating the output signalS (DN), given by Eq. (6), to its variance as (DN). Throughpropagation of errors, the variance in the signal for the idealCCD can be expressed by

[âS(DN)2z

[âS(DN)2

as (DN)â PI avi + a ,

an;

[ âS(DN)j2a+

a xJ K + aR (DN) . (15)

Differentiating Eq. (15) and assuming that the quantum yieldrli has negligible variance (i.e., no partial or split event;a2. = 0), we find

S (DN)J

= aS (DN) - aR (DN)(X<3000 A) (16)

Equations (14) and (16) form the basis for the photon- transfertechnique. By simply measuring the mean signal and its vari-ance for both visible photons and photons at any other spe-cific wavelength of illumination, we can determine the valuesK and J. Once the constants K and J are known, the idealquantum yield for photons at the wavelength under consider-ation can be calculated through Eq. (10).

3.4. Partial and split events included

Up to this point, we have assumed no partial or split eventgeneration within the CCD (i.e., i E = rl;). We now show thatthe ratio K/J with partial and split events included gives anupper limit for the effective quantum yield rjE, which in turngives an upper limit for CCE performance for the CCD asdefined by Eq. (4). We also show that as the number of partialand split events within the CCD decreases, the ratio K/Japproaches the real value of rlE and in the limit 77E equals rl;,when perfect CCE is achieved.

To analytically solve for the constant J under these condi-tions, we must give Eq. (6) a new form so that the variances ofthe partial and split events are included when the overall vari-ance of the signal is calculated. Such an equation for the averagesignal in any given pixel can be written in the form

S (DN) = P171;M(CPe

K -I (17)Psell'

where M is the average number of interacting photons perpixel.

It is informative to compare the behavior of the signalgiven by Eq. (17) to the signal given in Eq. (6) for the idealCCD camera without partial or split events. The signal de-scribed in Eq. (6) is proportional to il; and is not influenced byCCE characteristics since CCE is assumed to be perfect. In thecase of Eq. (17), we find that the signal is proportional toape/ PSe) when the number of interacting photons per pixel issmall (i.e., M ~ 1) and interactions are not adjacent to eachother in the CCD array. In this case, the amount of signalmeasured is dependent on both partial and split event behav-ior. However, when the number of interacting photons perpixel is large (i.e., M ~ PSe), the signal is dependent only onthe partial event [i.e., S (DN) = PIC,peK -l] and the effects ofthe split event are averaged out.

It can be shown, again using propagation of errors, that thesignal S (DN), given by Eq. (17), and its variance as (DN) arerelated to the effective quantum yield by

S (DN) / 2(4 Pseat \-1-1

1e - K [as (DN) - aR (DN) (\P5e + P5e + epe'

(18)

where ap is the event -to -event variance in the number of pixelsthat collect signal electrons per interacting photon and al isthe event -to -event variance for the total number of electronscollected. Here, we have assumed that the average of `ape -IIPse_I is equal to the average of Cpe1 divided by the average of"se _1, which becomes nearly correct for large numbers ofphoton events.

Equation (18) also can be written in the form

(19)

where E is PSe 2(4/ Pse) + PSe(GV ape) and K and J are asdefined in Eqs. (14) and (16), which we normally measureusing the photon- transfer technique.

The true value of the effective quantum yield i E given byEq. (19) is less than K/J by the factor c I. As the number ofpartial and split events decreases, the accuracy of K/Jimproves and in the limit is exact when E = 1 (i.e., rlE _Therefore, when measuring the effective quantum yield usingthe photon- transfer technique in the presence of partial andsplit events, the ratio K/J gives an upper limit for flE. Forexample, for the Texas Instruments (TI) 3PCCD (a CCD typediscussed in Sec. 4), we find experimentally that for individual5.9 keV (Fe55) photon events, one out of 11 events splitsbetween 2 pixels, with only a few partial events observed. Forthis CCD we calculate an average P5e of 1.09 pixels withvariances aP = 0.166 and a2 - 0. Assuming these values, wefind that c I = 0.72. Therefore, the true value of rlE is actu-ally smaller by 0.72 than the value of ,lE measured using thephoton- transfer method (i.e., K/J).

Even though the ratio K/J does not give an exact value forthe effective quantum yield, this quantity still is useful inevaluating and optimizing CCE performance of the CCD, aswe shall see in Sec. 4. Therefore, unless otherwise indicated,we use the ratio K/J as found through the photon- transfermethod as our standard measuring tool in comparing rlE andCCE performance for different CCDs under different operat-ing conditions, while keeping in mind that the absolute valuesof these quantities are lower (by c I).

3.5. Photon -transfer curve

The constants K and J can be found either graphically ordirectly through Eqs. (14) and (16). We examine the graphicalapproach first because the method gives insight into themechanics of the photon- transfer technique.

The constants K and J can be found graphically by plottinga curve (called the "photon- transfer curve ") of noise as (DN)as a function of signal S (DN), typically for a 20X20 pixelarray on the CCD. One such photon- transfer curve is shownin Fig. 2. For this curve we use 7000 A illumination, whichguarantees that TIE = rl; = 1 and therefore can be used infinding the constant K. The abscissa, S (DN), is the averagesignal level of the 400 pixels with the array uniformly illumi-nated at some level. (Here, we assume that electrical offset and

OPTICAL ENGINEERING / October 1987 / Vol. 26 No. 10 / 975

CHARGE-COUPLED-DEVICE CHARGE-COLLECTION EFFICIENCY AND THE PHOTON-TRANSFER TECHNIQUE

constant J also can be found by relating the output signal S (DN), given by Eq. (6), to its variance o$ (DN). Through propagation of errors, the variance in the signal for the ideal CCD can be expressed by

°l\ 'as(DN)T. 3*. J

(15)

Differentiating Eq. (15) and assuming that the quantum yield r/j has negligible variance (i.e., no partial or split event; 61. — 0), we find

J = S(DN)| (DN) - a£ (DN)

(XOOOOA) (16)

Equations (14) and (16) form the basis for the photon-transfer technique. By simply measuring the mean signal and its vari­ ance for both visible photons and photons at any other spe­ cific wavelength of illumination, we can determine the values K and J. Once the constants K and J are known, the ideal quantum yield for photons at the wavelength under consider­ ation can be calculated through Eq. (10).

3.4. Partial and split events included

Up to this point, we have assumed no partial or split event generation within the CCD (i.e., rjE = rj{). We now show that the ratio K/J with partial and split events included gives an upper limit for the effective quantum yield rjE , which in turn gives an upper limit for CCE performance for the CCD as defined by Eq. (4). We also show that as the number of partial and split events within the CCD decreases, the ratio K/J approaches the real value of rjE and in the limit rjE equals rj^ when perfect CCE is achieved.

To analytically solve for the constant J under these condi­ tions, we must give Eq. (6) a new form so that the variances of the partial and split events are included when the overall vari­ ance of the signal is calculated. Such an equation for the average signal in any given pixel can be written in the form

S(DN) = (17)

where M is the average number of interacting photons per pixel.

It is informative to compare the behavior of the signal given by Eq. (17) to the signal given in Eq. (6) for the ideal CCD camera without partial or split events. The signal de­ scribed in Eq. (6) is proportional to rj{ and is not influenced by CCE characteristics since CCE is assumed to be perfect. In the case of Eq. (17), we find that the signal is proportional to (£pe/ PSC) when the number of interacting photons per pixel is small (i.e., M « 1) and interactions are not adjacent to each other in the CCD array. In this case, the amount of signal measured is dependent on both partial and split event behav­ ior. However, when the number of interacting photons per pixel is large (i.e., M « Pse ), the signal is dependent only on the partial event [i.e., S (DN) = PI^K" 1 ] and the effects of the split event are averaged out.

It can be shown, again using propagation of errors, that the signal S (DN), given by Eq. (17), and its variance a| (DN) are related to the effective quantum yield by

= K S(DN)(DN) - a£ (DN)

P., + +

(18)

where a£ is the event-to-event variance in the number of pixels that collect signal electrons per interacting photon and a^ is the event-to-event variance for the total number of electrons collected. Here, we have assumed that the average of £pe _j/ Pse -i is equal to the average of £pe _j divided by the average of Pse -i» which becomes nearly correct for large numbers of photon events.

Equation (18) also can be written in the form

K_ Je

(19)

where e is Pse + 2(a£/Pse ) + Pse (a£/a£e ) and K and J are as defined in Eqs. (14) and (16), which we normally measure using the photon-transfer technique.

The true value of the effective quantum yield rjE given by Eq. (19) is less than K/J by the factor e" 1 . As the number of partial and split events decreases, the accuracy of K/J improves and in the limit is exact when e = 1 (i.e., rjE = rjj). Therefore, when measuring the effective quantum yield using the photon-transfer technique in the presence of partial and split events, the ratio K/J gives an upper limit for rjE . For example, for the Texas Instruments (TI) 3PCCD (a CCD type discussed in Sec. 4), we find experimentally that for individual 5.9 keV (Fe55 ) photon events, one out of 11 events splits between 2 pixels, with only a few partial events observed. For this CCD we calculate an average Pse of 1.09 pixels with variancesap = 0.166 and a^ « 0. Assuming these values, we find that e" 1 = 0.72. Therefore, the true value of rjE is actu­ ally smaller by 0.72 than the value of rjE measured using the photon-transfer method (i.e., K/J).

Even though the ratio K/ J does not give an exact value for the effective quantum yield, this quantity still is useful in evaluating and optimizing CCE performance of the CCD, as we shall see in Sec. 4. Therefore, unless otherwise indicated, we use the ratio K/J as found through the photon-transfer method as our standard measuring tool in comparing rjE and CCE performance for different CCDs under different operat­ ing conditions, while keeping in mind that the absolute values of these quantities are lower (by e" 1 ).

3.5. Photon-transfer curve

The constants K and J can be found either graphically or directly through Eqs. (14) and (16). We examine the graphical approach first because the method gives insight into the mechanics of the photon-transfer technique.

The constants K and J can be found graphically by plotting a curve (called the "photon-transfer curve") of noise as (DN) as a function of signal S (DN), typically for a 20X20 pixel array on the CCD. One such photon-transfer curve is shown in Fig. 2. For this curve we use 7000 A illumination, which guarantees that rjE = rj{ = I and therefore can be used in finding the constant K. The abscissa, S (DN), is the average signal level of the 400 pixels with the array uniformly illumi­ nated at some level. (Here, we assume that electrical offset and

OPTICAL ENGINEERING / October 1987 / Vol. 26 No. 10 / 975

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/21/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

JANESICK, KLAASEN, ELLIOTT

T= 7000 A

0 10tóz READ NOISE, oR

SLOPE = 0101

K = 2 e-/DN

SIGNALSHOT NOISESLOPE = 1/2

100 I I I i

100 101 102 103 104 105

SIGNAL (S), DN

Fig. 2. Photon -transfer curve using 7000 A illumination (*1; = 1).

10000Ea)

rq '' ,"'i , .,,.i "1

nE12.1Á1=1420 _

1000 71E11216Á1=3

77E(700011=1ó

100 .Z 2. 1A `a.

10 1216 Á

7000 Á

A

0.001 0.01 0.1 1 10 100 1000 10000 100000

SIGNAL, DN

Fig. 3. Photon -transfer curves taken at three photon energies, eachyielding a different effective quantum yield.

dark current were subtracted from the data before the signallevel was determined.) The ordinate, as (DN), is the standarddeviation of the signal of those 400 pixels at each exposure.The standard deviation is found after the CCD pixel -to -pixelnonuniformity has been removed. This can be accomplishedby differencing (pixel by pixel) two frames taken at the samelight level, calculating the standard deviation of the resultantdifference, and dividing by 2, which yields the desired as(DN).

The read noise aR (DN), indicated in Fig. 2, represents theintrinsic noise associated with the readout circuitry, i.e., theCCD on -chip amplifier and any other noise sources that areindependent of the signal level. As the signal is increased, thenoise eventually becomes dominated by the shot noise of thesignal and is characterized by a line of slope 1/ 2. From Eq.(14) we note that the intersection of the slope 1/ 2 line of thesignal axis [i.e., as (DN) = 1] represents the desired conver-sion constant K.

The same graphical approach can be used in determiningthe constant J when i > 1 [Eq. (16)]. For example, in Fig. 3we show three photon- transfer curves (taken with the sameCCD camera and CCD) generated from flat fields at wave-lengths of 7000 A, 1216 A, and 2.1 A. The correspondingintersections on the signal axis at each of these wavelengthsare 2.3, 0.77, and 1.62 X 10 -3. Since 77E = m = 1 for 7000 Aillumination, the signal at as (DN) = 1 for this photon -transfer curve represents the value of constant K (i.e.,K = 2.3 e -/ DN). The other two intersections, for the wave-lengths 1216 A and 2.1. A, represent values for J that can beused in conjunction with K to find rlE (= K /J), yielding an

976 / OPTICAL ENGINEERING / October 1987 / Vol. 26 No. 10

30

20

10

TI3PCCD, _X = 4000 A -

J = K = 1.5

0.50 0.75 1.0 1.25 1.50 1.75 2.00 2.25 2.50

INTERACTING PHOTONS /DN

Fig. 4. Photon -transfer histogram using 4000 A illumination(+F, =1).

average of 3 e and 1420 e per affected pixel per interactingphoton, respectively.

In the case of2.1 A(Ex = 5.9keV; = 1610 el ),thephoton- transfer technique yields a K/J of 1420 e-. An actual7k of 1215 a is readily determined by measuring individualphoton events [Eq. (5)], which gives a value for CI in Eq. (19)of 0.85. From Eq. (4), an upper limit of 0.88 for CCE perfor-mance is calculated using K/ J found from the photon- transfercurve, while a true CCE of 0.75 is calculated using individualphoton events. This level of CCE performance is quite goodby today's CCD standards.

3.6. Photon -transfer histogram

The accuracy of determining K and J can be improved byusing Eqs. (14) and (16) directly (as opposed to the graphicalapproach used in Sec. 3.5). The signal S (DN) and the noise as(DN) are found from the CCD in the same manner as for thephoton- transfer curve discussed above. The read noise aR(DN) is found from a dark image. After applying these formu-las to many different 20 X20 pixel subarrays across the sensor,the resulting values of K (or J) are compiled into a form we calla "photon- transfer histogram." An example histogram using4000 A illumination is shown in Fig. 4. It produces a veryaccurate value of K = 1.5 e -/ DN. Using many 20 X20 pixelsubarrays allows elimination of those regions on the devicethat contain blemish artifacts, which give erroneous values forK. Areas that are not well behaved can be easily recognized asdata points outside the main histogram, as Fig. 4 shows. Asimilar histogram also can be generated over the entire CCDarray for 77E (= K/ J) at a specific wavelength of interest. Thistype of histogram is quite valuable in characterizing the vari-ability of CCE performance across the array of the CCD. InSec. 4 we depict the use of histograms of rlE under differentoperating conditions of the CCD.

4. PHOTON -TRANSFER USEIn this section we apply the photon- transfer technique tomeasuring CCE performance for two different types of CCDs,namely, the thick TI frontside- illuminated virtual -phase CCD(TI VPCCD) and the thin TI backside -illuminated three -phase CCD (TI 3PCCD). These devices are discussed in con-siderable detail elsewhere. 1- 3 The CCDs used in tests discussedhere have approximately the same read noise floor, charge -transfer efficiency, and quantum efficiency performance for

JANESICK, KLAASEN, ELLIOTT

100 10* 105102 103

SIGNAL (S), DN

Fig. 2. Photon-transfer curve using 7000 A illumination (T?J = 1).

10000

1000

100

0.001 0.01 0.1 1 10 100 1000 10000 100000

SIGNAL, DN

Fig. 3. Photon-transfer curves taken at three photon energies, each yielding a different effective quantum yield.

dark current were subtracted from the data before the signal level was determined.) The ordinate, as (DN), is the standard deviation of the signal of those 400 pixels at each exposure. The standard deviation is found after the CCD pixel-to-pixel nonuniformity has been removed. This can be accomplished by differencing (pixel by pixel) two frames taken at the same light level, calculating the standard deviation of the resultant difference, and dividing by 2, which yields the desired as (DN).

The read noise aR (DN), indicated in Fig. 2, represents the intrinsic noise associated with the readout circuitry, i.e., the CCD on-chip amplifier and any other noise sources that are independent of the signal level. As the signal is increased, the noise eventually becomes dominated by the shot noise of the signal and is characterized by a line of slope 1/2. From Eq. (14) we note that the intersection of the slope 1/2 line of the signal axis [i.e., as (DN) = 1] represents the desired conver­ sion constant K.

The same graphical approach can be used in determining the constant J when rj{ > 1 [Eq. (16)]. For example, in Fig. 3 we show three photon-transfer curves (taken with the same CCD camera and CCD) generated from flat fields at wave­ lengths of 7000 A, 1216 A, and 2.1 A. The corresponding intersections on the signal axis at each of these wavelengths are 2.3, 0.77, and 1.62 X10~3 . Since r]E = r]- = 1 for 7000 A illumination, the signal at as (DN) = 1 for this photon- transfer curve represents the value of constant K (i.e., K = 2.3 e~/DN). The other two intersections, for the wave­ lengths 1216 A and 2.1. A, represent values for J that can be used in conjunction with K to find rjE (— K/ J), yielding an

I I I I I I I I I I I TI3PCCD X= 4000 A

1.5

0.50 0.75 1.0 1.25 1.50 1.75 2.00 2.25 2.50

INTERACTING PHOTONS/DN

Fig. 4. Photon-transfer histogram using 4000 A illumination(T* = 1).

average of 3 e~ and 1420 e~ per affected pixel per interacting photon, respectively.

In the case of 2.1 A (Ex = 5.9 keV; rj = 1610 e~), the photon-transfer technique yields a K/J of 1420 e~. An actual rjE of 1215 e~ is readily determined by measuring individual photon events [Eq. (5)], which gives a value for e~ l in Eq. (19) of 0.85. From Eq. (4), an upper limit of 0.88 for CCE perfor­ mance is calculated using K/ J found from the photon-transfer curve, while a true CCE of 0.75 is calculated using individual photon events. This level of CCE performance is quite good by today's CCD standards.

3.6. Photon-transfer histogram

The accuracy of determining K and J can be improved by using Eqs. (14) and (16) directly (as opposed to the graphical approach used in Sec. 3.5). The signal S (DN) and the noise as (DN) are found from the CCD in the same manner as for the photon-transfer curve discussed above. The read noise aR (DN) is found from a dark image. After applying these formu­ las to many different 20 X20 pixel subarrays across the sensor, the resulting values of K (or J) are compiled into a form we call a "photon-transfer histogram." An example histogram using 4000 A illumination is shown in Fig. 4. It produces a very accurate value of K = 1.5 e~/DN. Using many 20X20 pixel subarrays allows elimination of those regions on the device that contain blemish artifacts, which give erroneous values for K. Areas that are not well behaved can be easily recognized as data points outside the main histogram, as Fig. 4 shows. A similar histogram also can be generated over the entire CCD array for rjE (= K/ J) at a specific wavelength of interest. This type of histogram is quite valuable in characterizing the vari­ ability of CCE performance across the array of the CCD. In Sec. 4 we depict the use of histograms of rjE under different operating conditions of the CCD.

4. PHOTON-TRANSFER USE

In this section we apply the photon-transfer technique to measuring CCE performance for two different types of CCDs, namely, the thick TI frontside-illuminated virtual-phase CCD (TI VPCCD) and the thin TI backside-illuminated three- phase CCD (TI 3PCCD). These devices are discussed in con­ siderable detail elsewhere. ! ~ 3 The CCDs used in tests discussed here have approximately the same read noise floor, charge- transfer efficiency, and quantum efficiency performance for

976 / OPTICAL ENGINEERING / October 1987 / Vol. 26 No. 10

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/21/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

SUBSTRATE

p+ 10.011 -cm)

EPI INTERFACESTATES

CHARGE -COUPLED -DEVICE CHARGE -COLLECTION EFFICIENCY AND THE PHOTON -TRANSFER TECHNIQUE

EPITAX IAL LAYER

v,

Ec

p 11012-cml

TII - --1 µmr-

p+ DIFFUSION

fl2µm - 5µm

-

-e i POTENTIAL

E -iV ¡'

Ef i

240 µm

FIELD

FREE -HREGIME

SPLIT -PARTIALEVENTS

e e

QUAS I -FERMI LEVEL

FOR ELECTRONS

10µm

DEPLETION

REGION

FIELD

FREE

REGIME

SPLIT -PARTIALEVENTS

h`-

SiO2

1200 A

POLYS I LICON

GATE

BACKS IDE

ELECTRIC

FIELD

5000 A

a

Fig. 5. Cross section of a thick frontside -illuminated CCD, showingenergy band structure and locations at which partial and split eventgeneration occurs.

the wavelengths used. However, we show that CCE perfor-mance for the two CCDs is significantly different owing to thenumber of partial and split events generated in each device.

4.1. Frontside illumination (TI VPCCD)In general, we find numerous partial and split events for thethick frontside -illuminated epitaxial CCD. Such a device isschematically represented in Fig. 5, which shows the primaryregions within the CCD that are responsible for generating splitand partial events. Photons that interact in the substrate regionproduce a charge cloud that has a high probability of recombi-nation owing to the high concentration of holes in this regionand the existence of bulk interface states at the epitaxial inter-face. The physical size of the charge cloud also is likely to spanthe boundary between adjacent pixels. Such interactions tend toresult in split and partial events. Photons that interact betweenthe epitaxial interface and frontside depletion region alsogenerate split events due to charge diffusion in the field -freeregion and partial events due to signal charge diffusing throughthe weak field produced by the p+ diffusion into the bulk trapsat the epitaxial interface (see Fig. 5).

Figure 6 shows the response of the TI VPCCD uniformlyilluminated by an Fe55 x -ray source with a mean flux ofapproximately i x ray per 500 pixels. The DN of each pixel onthe CCD containing signal charge is measured and appro-priately binned in histogram form. Ideally, the response of Fig.6 should show only two prominent peaks, located at 1150 DN(1610 e) and 1270 DN (1780 e) due to the Mn -Ka andMn -Kß x rays generated by the F55 source. The events withsignals below these two lines are the result of split and partial

500

400

? 300

oo

200

z100

THICK FRONTSIDEILLUMINATED

Fe55

1.4e -IDN

3.65 eVle-

PARTIAL AND SPLIT EVENTS

Mn Ka

fMn Kß

200 400 600 800 1000 1200 1400 1600 1800 2000

SIGNAL, DN

Fig. 6. Fe55 x -ray histogram for a thick frontside -illuminated CCD,showing numerous partial and split events.

200

150 -

100 -

50 -

THICK FRONTSIDEILLUMINATED

Fe55

IDEAL

I

i

I

IJ i1 i

0 200 400 600 800 1000 1200 1400 16100 1800 2000

QUANTUM YIELD í'7E1, e

Fig. 7. Histogram of effective quantum yield nE for a thick frontside-illuminated CCD.

events and constitute the majority of the events observed.Figure 7 shows a histogram of effective quantum yield '7E

calculated by the photon- transfer method for the same TIVPCCD used in generating the histogram shown in Fig. 6.The histogram is generated by uniformly stimulating the CCDwith a mean flux of about 5 x rays per pixel (Fe55) andcalculating 17E for several different 40X40 pixel subarraysacross the sensor. We find from Fig. 7 an average effectivequantum yield of 900 a -/ pixel / interacting photon, which issignificantly less than the ideal quantum yield rli of 1610 e .

The relative CCE performance for this device is readily calcu-lated to be 0.56 using Eq. (4).

4.2. Backside illumination (TI 3PCCD)For the backside -illuminated CCD, in which the substrateand epitaxial interface are removed (Fig. 8), the number ofpartial and split events is significantly reduced. It has beendemonstrated that an internal QE of 100% can be achieved(i.e., pe /77i = 1) for the CCD that is properly thinned andbackside treated.4 -6 The main factor that determines CCEperformance for the backside -illuminated CCD is the splitevent.

To minimize the number of split events for the backside -illuminated CCD, it is important that an electric field ofgreater than 105 V/ cm be provided throughout the entirephotosensitive depth of the CCD. Regions in which the field

OPTICAL ENGINEERING / October 1987 / Vol. 26 No. 10 / 977

CHARGE-COUPLED-DEVICE CHARGE-COLLECTION EFFICIENCY AND THE PHOTON-TRANSFER TECHNIQUE

SUBSTRATE EPITAXIAL LAYER 500

BACKSIDE FRONTS IDE

Fig. 5. Cross section of a thick frontside-illuminated CCD, showing energy band structure and locations at which partial and split event generation occurs.

the wavelengths used. However, we show that CCE perfor­ mance for the two CCDs is significantly different owing to the number of partial and split events generated in each device.

4.1. Frontside illumination (TI VPCCD)

In general, we find numerous partial and split events for the thick frontside-illuminated epitaxial CCD. Such a device is schematically represented in Fig. 5, which shows the primary regions within the CCD that are responsible for generating split and partial events. Photons that interact in the substrate region produce a charge cloud that has a high probability of recombi­ nation owing to the high concentration of holes in this region and the existence of bulk interface states at the epitaxial inter­ face. The physical size of the charge cloud also is likely to span the boundary between adjacent pixels. Such interactions tend to result in split and partial events. Photons that interact between the epitaxial interface and frontside depletion region also generate split events due to charge diffusion in the field-free region and partial events due to signal charge diffusing through the weak field produced by the p+ diffusion into the bulk traps at the epitaxial interface (see Fig. 5).

Figure 6 shows the response of the TI VPCCD uniformly illuminated by an Fe55 x-ray source with a mean flux of approximately 1 x ray per 500 pixels. The DN of each pixel on the CCD containing signal charge is measured and appro­ priately binned in histogram form. Ideally, the response of Fig. 6 should show only two prominent peaks, located at 1150 DN (1610 e~) and 1270 DN (1780 e") due to the Mn-Ktt and Mn-K0 x rays generated by the F55 source. The events with signals below these two lines are the result of split and partial

\ 300 r

. 200 -

100 -

THICK FRONTS IDE ILLUMINATED

1.4e"/DN 3.65eV/e~

7 PARTIAL AND SPLIT EVENTS

KQ,

200 400 600 800 1000 1200 1400 1600 1800 2000 SIGNAL, DN

Fig. 6. Fe55 x-ray histogram for a thick frontside-illuminated CCD, showing numerous partial and split events.

200

150

8 100

50

THICK FRONTSIDE ILLUMINATED

/IDEAL

0 200 400 600 800 1000 1200 1400 1600 1800 2000 QUANTUM YIELD t^), e~

Fig. 7. Histogram of effective quantum yield r?E for a thick frontside- illuminated CCD.

events and constitute the majority of the events observed.Figure 7 shows a histogram of effective quantum yield r]E

calculated by the photon-transfer method for the same TI VPCCD used in generating the histogram shown in Fig. 6. The histogram is generated by uniformly stimulating the CCD with a mean flux of about 5 x rays per pixel (Fe55 ) and calculating rjE for several different 40X40 pixel subarrays across the sensor. We find from Fig. 7 an average effective quantum yield of 900 e~/ pixel/ interacting photon, which is significantly less than the ideal quantum yield rj{ of 1610 e~. The relative CCE performance for this device is readily calcu­ lated to be 0.56 using Eq. (4).

4.2. Backside illumination (TI 3PCCD)

For the backside-illuminated CCD, in which the substrate and epitaxial interface are removed (Fig. 8), the number of partial and split events is significantly reduced. It has been demonstrated that an internal QE of 100% can be achieved (i.e., ^pe/Tjj = 1) for the CCD that is properly thinned and backside treated.4 " 6 The main factor that determines CCE performance for the backside-illuminated CCD is the split event.

To minimize the number of split events for the backside- illuminated CCD, it is important that an electric field of greater than 105 V/cm be provided throughout the entire photosensitive depth of the CCD. Regions in which the field

OPTICAL ENGINEERING / October 1987 / Vol. 26 No. 10 / 977

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/21/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

BACKSIDE

e

e

BACKSIDECHARGE e

FRONTS IDE

n

DEPLETION

REG

1µm

E e

Ef

Ev

ACCUMULATIONLAYER

20

Si02

6-10µm

JANESICK, KLAASEN, ELLIOTT

Fig. 8. Cross section of a thin backside- illuminated CCD that isbackside charged, showing energy band structure.

500

Nz 375

_U° 250

z 125

TOTAL EVENTS = 2250

2.3 e /DN

3.65 eV /e

CI

KaESCAPE PEAK

Mn Ka

11620e-1 -

300 400 500 600 700

SIGNALIS), DN

Fig. 9. Fe55 x -ray histogram for a thin backside -illuminated CCD,showing very few partial and split events.

IrMn Kß

i(1780é I

_

800 900

strength is lower permit the charge cloud to diffuse to a largesize, making it likely to overlap more than one pixel and split.For the TI 3PCCD, it has been demonstrated that a high fieldcondition throughout a 7 µm region can be achieved usingbackside charging or a flash gate and proper bias conditionsacross the n- channel and substrate (Fig. 8).

Figure 9 shows a histogram response to Fe55 x rays of a TI3PCCD that is properly thinned (i.e., thinned to the frontsidedepletion edge) and backside treated. Comparing Figs. 6 and9, we note that the number of partial and split events issignificantly less for the backside -illuminated CCD than forthe frontside -illuminated CCD.

Figure 10 shows a TI 3PCCD histogram of the effectivequantum yield for Fe55 x rays, generated using the photon -transfer method. We note from this figure that two regions onthe CCD (labeled thin and overly thin) produce differentquantum yields. The overly thin region is slightly thinner thanthe thin region and represents where the frontside depletionedge has reached the backside. The thicker region exhibitsdegraded performance because of a larger field -free region,which causes more charge splitting due to diffusion andresults in a lower effective quantum yield. The variability ofquantum yield across the CCD membrane can be seen moreclearly in Fig. 11, in which TIE is displayed in image format as aresult of calculating K/J for 160,000 subarrays across theCCD. Figure 12 is an expanded view of the upper right -handcorner of Fig. 11, and Fig. 13 is a corresponding printout offE. It is seen from Figs. 12 and 13 that the quantum yield forthis device varies considerably from 1400 e- in the corners(white regions) to 1200 a in the middle (black regions). The

978 / OPTICAL ENGINEERING / October 1987 / Vol. 26 No. 10

900

800 -

700 -

600 -

500 -

400 -

300 -

200 -

100 -

THIN BACKSIDEILLUMINATED

Fe55

VnP = 17 volts

0 i It I I '.'0 200 400 600 800 1000 1200 1400 1600 1800 2000

QUANTUM YIELD 1'7E1' e-

Fig. 10. Histogram of effective quantum yield TE for a thin backside -illuminated CCD.

THIN

OVERLY -THIN -/IDEAL

differences in 77E are the result of thinning nonuniformities,where the four corners are physically thinner than the middleof the array because of the thinning method employed.

Figure 14 shows 77E histograms for a different TI 3PCCDbiased with different frontside depletion voltages (VnP) acrossthe n- channel and p- substrate.5 The distance that the edge ofthe depletion region extends into the substrate is proportionalto V42 and therefore influences the field -free region of theCCD. From these figures we see that the effective quantumyield decreases significantly as Vn is lowered due to anincrease in field -free material, which causes more diffusionand charge splitting. Figure 15 shows an 77E image over theCCD array for VnP = 7 V; note two distinct areas that cor-respond to the thin and overly thin regions indicated in Fig.14(b).

The last r1E histogram, shown in Fig. 16, is similar to that ofFig. 10 but has an elevated frontside depletion voltage ofVn = 27 V and represents the best CCE performanceachieved for the TI 3PCCD. For the overly thin region, anaverage quantum yield of 1420 a was measured, which yieldsa relative CCD = 0.88. We believe that this limit in CCEperformance for the TI 3PCCD is due primarily to two fac-tors: First, the initial cloud diameter size for a 5.9 keV photonevent is significant (cloud diameter ~ 0.6 µm); therefore,charge splitting between the TI 3PCCD's 15 Am pixels occurseven if there is no field -free material. Second, it is known thata field -free region exists beneath the channel stop regions,which allows charge diffusion and charge splitting to occur.Both of these effects are discussed in detail in Ref. 7, where it isshown that increasing the size of the pixel improves the CCEperformance further because of these two factors.

5. FUTURE IMPROVEMENTS IN CCEAt this time in the development of the CCD, the backside -illuminated device is superior to the frontside -illuminateddevice in achieving high CCE performance. This is becausethe backside -illuminated CCD allows field control over mostof the photosensitve volume and elimination of neutral bulkand trapping centers at which photogenerated charge candiffuse and recombine to produce partial and split events.

Improvements in CCE for the backside -illuminated CCDwill be made principally in two areas. The first area for

JANESICK, KLAASEN, ELLIOTT

BACKSIDE FRONTS IDE

e"

e"

BACKSIDE CHARGE e^

e"1 11

p n ^

, DEPLETION II P —— REGION — ̂ 11

P*-—— e^ ——I""'!

^NW

1 ~^Ev\ h*- ACCUMULATION X.LAYER ^

^ 2QA Si0 2

_ ̂ 111 1200 A

Fig. 8. Cross section of a thin backside-illuminated CCD that is backside charged, showing energy band structure.

3UU

<yo

y 375

12500

Ii 125

0

': • TOTAL EVENTS = 2250

; «2.3e~/DN

r • 3. 65 eV/e-

:

~

; ESCAPE PEAK r

LrrdL***^*^^

^ /- Mn Ka ;

(1620e") .:

:

Mn K :

/ (1780e"):f -

U, 1300 400 500 600 700

SIGNAL (S), DN800 900

Fig. 9. Fe x-ray histogram for a thin backside-illuminated CCD, showing very few partial and split events.

strength is lower permit the charge cloud to diffuse to a large size, making it likely to overlap more than one pixel and split. For the TI 3PCCD, it has been demonstrated that a high field condition throughout a 7 /zm region can be achieved using backside charging or a flash gate and proper bias conditions across the n-channel and substrate (Fig. 8).

Figure 9 shows a histogram response to Fe55 x rays of a TI 3PCCD that is properly thinned (i.e., thinned to the frontside depletion edge) and backside treated. Comparing Figs. 6 and 9, we note that the number of partial and split events is significantly less for the backside-illuminated CCD than for the frontside-illuminated CCD.

Figure 10 shows a TI 3PCCD histogram of the effective quantum yield for Fe55 x rays, generated using the photon- transfer method. We note from this figure that two regions on the CCD (labeled thin and overly thin) produce different quantum yields. The overly thin region is slightly thinner than the thin region and represents where the frontside depletion edge has reached the backside. The thicker region exhibits degraded performance because of a larger field-free region, which causes more charge splitting due to diffusion and results in a lower effective quantum yield. The variability of quantum yield across the CCD membrane can be seen more clearly in Fig. 11, in which rjE is displayed in image format as a result of calculating K/J for 160,000 subarrays across the CCD. Figure 12 is an expanded view of the upper right-hand corner of Fig. 11, and Fig. 13 is a corresponding printout of rjE . It is seen from Figs. 12 and 13 that the quantum yield for this device varies considerably from 1400 e~ in the corners (white regions) to 1200 e~ in the middle (black regions). The

900

800

700

600

500

400

300

200

100

0

THIN BACKSIDE ILLUMINATEDF055

Vnp = 17 volts

OVERLY-THIN - /IDEAL

200 400 600 800 1000 1200 1400 1600 1800 2000 QUANTUM YIELD (%), e"

Fig. 10. H istogram of effective quantum yield IJE for a thin backside- illuminated CCD.

differences in rjE are the result of thinning nonuniformities, where the four corners are physically thinner than the middle of the array because of the thinning method employed.

Figure 14 shows rjE histograms for a different TI 3PCCD biased with different frontside depletion voltages (V ) across the n-channel and p-substrate. 5 The distance that the edge of the depletion region extends into the substrate is proportional to V^2 and therefore influences the field-free region of the CCD. From these figures we see that the effective quantum yield decreases significantly as V is lowered due to an increase in field-free material, whicn causes more diffusion and charge splitting. Figure 15 shows an r]E image over the CCD array for Vnp — 1 V; note two distinct areas that cor­ respond to the thin and overly thin regions indicated in Fig.

The last rjE histogram, shown in Fig. 1 6, is similar to that of Fig. 10 but has an elevated frontside depletion voltage of V = 27 V and represents the best CCE performance acnieved for the TI 3PCCD. For the overly thin region, an average quantum yield of 1420 e~ was measured, which yields a relative CCD = 0.88. We believe that this limit in CCE performance for the TI 3PCCD is due primarily to two fac­ tors: First, the initial cloud diameter size for a 5.9 keV photon event is significant (cloud diameter « 0.6 )um); therefore, charge splitting between the TI 3PCCD's 15 pm pixels occurs even if there is no field-free material. Second, it is known that a field-free region exists beneath the channel stop regions, which allows charge diffusion and charge splitting to occur. Both of these effects are discussed in detail in Ref. 7, where it is shown that increasing the size of the pixel improves the CCE performance further because of these two factors.

5. FUTURE IMPROVEMENTS IN CCE

At this time in the development of the CCD, the backside- illuminated device is superior to the frontside-illuminated device in achieving high CCE performance. This is because the backside-illuminated CCD allows field control over most of the photosensitve volume and elimination of neutral bulk and trapping centers at which photogenerated charge can diffuse and recombine to produce partial and split events.

Improvements in CCE for the backside-illuminated CCD will be made principally in two areas. The first area for

978 / OPTICAL ENGINEERING / October 1987 / Vol. 26 No. 10

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/21/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

CHARGE -COUPLED -DEVICE CHARGE -COLLECTION EFFICIENCY AND THE PHOTON -TRANSFER TECHNIQUE

Fig. 11. Pictorial display of the effective quantum yield for a thinbackside -illuminated CCD, showing regions of different thickness.

Fig. 12. Magnified view of upper right -hand corner of Fig. 11.

improvement will come from depleting the entire photosensi-tive volume of the CCD. Although complete depletionalready has been demonstrated for the TI 3PCCD in certainregions on the array, depletion extends only a maximum of 7µm. Therefore, high energy efficiency and cosmic ray back-ground rejection characteristics are still poor for this sensor.Increasing the thickness of the CCD and depleting it fullythrough deep depletion technology (high resistivity silicon) inconjunction with backside treatment should yield both highCCE and high energy sensitivity for the CCD. Work at severalCCD manufacturers has been initiated to accomplish thisgoal.

Second, larger pixels (-30 µm) would increase CCE per-formance further by making a larger "target " for the incomingphoton and reducing the effects of the initial event clouddiameter. In addition, a smaller fraction of the pixel is devotedto "overhead" functions such as the channel stop regions,which tend to increase the number of split events by diffusion.

1248 1251 1323 1346 1382 1369 1348 1440 1420 1478 1526 1434 1415

1273 1246 1261 1344 1354 1314 1339 1420 1420 1441 1471 1475 1437

1192 1157 1230 1248 1255 1321 1344 1303 1413 1375 1391 1489 1431

1058 1117 1149 1139 1232 1416 1374 1293 1337 1342 1462 1500 1428

1131 1118 1141 1129 1280 1270 1328 1339 1402 1413 1456 1391 1358

1099 1127 1183 1154 1241 1185 1270 1337 1399 1326 1332 1373 1411

1157 1154 1207 1142 1171 1116 1116 1186 1231 1280 1324 1461 1441

1199 1221 1217 1193 1151 1168 1171 1142 1180 1282 1379 1461 1369

1130 1130 1207 1172 1185 1191 1116 1152 1202 1146 1236 1318 1319

1166 1167 1198 1204 1220 1188 1121 1171 1117 1131 1157 1202 1325

1231 1203 1170 1135 1207 1207 1134 1121 1046 1216 1171 1159 1316

1202 1221 1279 1241 1264 1129 1094 1157 1204 1239 1175 1136 1187

1135 1063 1147 1189 1203 1082 1153 1175 1186 1146 1116 1157 1142

CORNER QUANTUM YIELDS, TIE

Fig. 13. Printout representing effective quantum yield of the regionshown in Fig. 12.

(a)

(b)

1 I

VnP=17volts

Fe55

OVERLY-TH INr

THIN

0 200 400 600 800 1600 1200 1400 1600 1800 2000

QUANTUM YIELD I77E1, e-

VIP = 7 volts

Fe 55

THIN

/OVERLY-THIN

\n h1

200 400 600 800 1000 1200 1400 1600 1800 2010

QUANTUM YIELD ITIEI, e-

600 800 1000 1200 1400 1600 1800 2000

QUANTUM Y IELD I' El, e-

Fig. 14. Effective quantum yield histograms for a thin backside -illuminated CCD using frontside depletion voltages of (a) 17 V,(b) 7 V, and (c) 2 V.

OPTICAL ENGINEERING / October 1987 / Vol. 26 No. 10 / 979

CHARGE-COUPLED-DEVICE CHARGE-COLLECTION EFFICIENCY AND THE PHOTON-TRANSFER TECHNIQUE

Fig. 11. Pictorial display of the effective quantum yield for a thin backside-illuminated CCD, showing regions of different thickness.

Fig. 12. Magnified view of upper right-hand corner ot Fig. 11.

improvement will come from depleting the entire photosensi­ tive volume of the CCD. Although complete depletion already has been demonstrated for the TI 3 PC CD in certain regions on the array, depletion extends only a maximum of 7 /urn. Therefore, high energy efficiency and cosmic ray back­ ground rejection characteristics are still poor for this sensor. Increasing the thickness of the CCD and depleting it fully through deep depletion technology (high resistivity silicon) in conjunction with backside treatment should yield both high CCE and high energy sensitivity for the CCD. Work at several CCD manufacturers has been initiated to accomplish this goal.

Second, larger pixels (—30 jum) would increase CCE per­ formance further by making a larger "target"for the incoming photon and reducing the effects of the initial event cloud diameter. In addition, a smaller fraction of the pixel is devoted to "overhead" functions such as the channel stop regions, which tend to increase the number of split events by diffusion.

1248 125!

1273 1246

1192 1157

1058 1117

1131 1118

1099 1127

1157 1154

1199 1221

1130 1130

1166 1167

1231 1203

1202 1221

1135 1063

1323 1346

1261 1344

1230 1248

1149 1139

1141 1129

1183 1154

1207 1142

1217 1193

1207 1172

1198 1204

1170 1135

1279 1241

1147 1189

1382 1369

1354 1314

1255 1321

1232 1416

1280 1270

1241 1185

1171 1116

1151 1168

1185 1191

1220 1188

1207 1207

1264 1129

1203 1082

1348 1440

1339 1420

1344 1303

1374 1293

1328 1339

1270 1337

1116 1186

1171 1142

1116 1152

1121 1171

1134 1121

1094 1157

1153 1175

1420 1478 1526

1420 1441 1471

1413 1375 1391

1337 1342 1462

1402 1413 1456

1399 1326 1332

1231 1280 1324

1180 1282 1379

1202 1146 1236

1117 1131 1157

1046 1216 1171

1204 1239 1175

1186 1146 1116

1434 1415

1475 1437

1489 1431

1500 1428

1391 1358

1373 1411

1461 1441

1461 1369

1318 1319

1202 1325

1159 1316

1136 1187

1157 1142

CORNER QUANTUM YIELDS, T] EFig. 13. Printout representing effective quantum yield of the region shown in Fig. 12.

600

50000

g 400 -€£.y 300- |»z 100 -

(a)

~~i——i——i—• V np -17wlts

• Fe,55

OVERLY-THIN

0 ———————————I________-^_________________iX^.L . I

0 200 400 600 800 1000 1200 1400 1600 1800 2000QUANTUM YIELD(*?E ), e~

600

500 -

I

I 200

i

100

<*>) o

<V np =7voIts

• Fe55

THINOVERLY-THIN,

0 200 400 600 800 1000 1200 1400 1600 1800 2000 QUANTUM YIELDP?E ), e~

600

500

§ I 400

5o 300 fea:

I 200•^•z.

100

(c) o

.V np =2volts

200 400 600 1000 120Q 1400 1600 1800 2000

Fig. 14. Effective quantum yield histograms for a thin backside- illuminated CCD using frontside depletion voltages of (a) 17 V, (b) 7 V, and (c) 2 V.

OPTICAL ENGINEERING / October 1 987 / Vol. 26 No. 10 / 979

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/21/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

JANESICK, KLAASEN, ELLIOTT

Fig. 15. An effective quantum yFig. 14 (b).

900

800 -

eld image for the histogram of

,A 700 -

z

E 500 -óó 400 -w

Z

600 -

300 -

200 -

100 -

o

THIN BACKSIDEILLUMINATED

Vnp= 27 volts

Fe 55

0 600 800 1000 1200 1400

QUANTUM YIELD I'k1, e-

Fig. 16. Histogram of best effective quantum yield for the TI 3PCCD,using a frontside voltage of 27 V.

THIN

OVERLY -THIN

IDEAL

1600 1800 2000

6. ACKNOWLEDGMENTSThe authors acknowledge many rewarding conversations onthe subject of CCE with Morley Blouke (father of the TI3PCCD), Taher Daud, Andy Collins, Dave Campbell, JamesDeWitt, Arsham Dingizian, and James McCarthy. We alsothank Deborah Durham for reviewing this paper. Theresearch described was carried out by the Jet PropulsionLaboratory, California Institute of Technology, under con-tract with the National Aeronautics and Space Administration.

7. REFERENCESI. J. R. Janesick, T. Elliott, S. Collins, M. M. Blouke, and J. Freeman,

"Scientific charge -coupled devices," Opt. Eng. 26(8), 692 -714 (1987).2. M. M. Blouke, J. R. Janesick, T. Elliott, J. E. Hall, M. W. Cowens, and

P. J. May, "Current status of 800X800 charge -coupled -device imagersensor," Opt. Eng. 26(9), in press (1987).

3. J. R. Janesick, J. Hynecek, and M. M. Blouke, "Virtual phase imager forGalileo," in Solid State Imagers for Astronomy, J. C. Geary and D. W.Latham, ed., Proc. SPIE 290, 165 -173 (1981).

4. J. Janesick, T. Elliott, T. Daud, J. McCarthy, and M. Blouke, "Backsidecharging of the CCD," in Solid State Imaging Arrays, K. N. Prettyjohnsand E. L. Dereniak, eds., Proc. SPIE 570, 46 -79 (1985).

5. J. Janesick, T. Elliott, T. Daud, and D. Campbell, "The CCD flash gate,"in Instrumentation in Astronomy VI, D. L. Crawford, ed., Proc. SPIE627, 543 -582 (1986).

6. J. Janesick, D. Campbell, T. Elliott, T. Daud, and P. Ottley, "Flashtechnology for CCD imaging in the UV," in Ultraviolet Technology,R. Huffman, ed., Proc. SPIE 687, 36 -55 (1986).

7. J. Janesick, T. Elliott, J. McCarthy, H. Marsh, and S. Collins, "Presentand future CCDs for UV and x -ray scientific measurements," IEEETrans. Nuc. Sci. NS -32 (1), 409 -416 (1985).

James R. Janesick: Biography and photograph appear with the GuestEditorial in this issue.

Kenneth P. Klaasen is the supervisor of thePhotoscience Group at JPL. He received his BSdegree in physics from Calvin College, GrandRapids, Mich., and his MS in aerospace engi-neering from the University of Michigan in1969. He has been involved in many solar sys-tem exploration missions, including Mariner10, the Viking Orbiter, and currently the Gali-leo mission to Jupiter. He has served as theexperiment representative for the imagingexperiments on these missions and is cur-

rently a member of the Galileo Imaging Science Team. His primaryresponsibilities have included imaging system calibration, experimentplanning, and mission operations. He has published papers on therotation period of Mercury, Venus atmospheric dynamics, and the pho-tometry of Mercury, Mars, and the Martian moons, Phobos and Deimos,as well as several on various spacecraft imaging instruments andexperiments.

Stythe T. Elliott was born in Van Nuys, Cali-fornia. He received his BA degree in geographyfrom the California State University at North-ridge. He joined the Jet Propulsion Laboratoryin 1979 and is presently working on develop-ing charge -coupled devices for NASA spaceimaging systems. He received a NASAAchievement Award in 1986. He is currentlyinvolved with the development of CCDs used inthe wide field /planetary Hubble Space Tele-scope camera.

980 / OPTICAL ENGINEERING / October 1987 / Vol. 26 No. 10

JANESICK, KLAASEN, ELLIOTT

7. REFERENCES

Fig. 15. An effective quantum yield image for the histogram of Fig. 14(b).

wu800'

a 7000

1 600

§500oo 400CSL UJ

§ 300

1 200

100

0

1 I I • 1 I 1 I 1 1

• THIN BACKS I IDEILLUMINATED'

• V » 27 volts |Lx/ THINi,, r ;I .OVERLY-THIN

1 an/

A i\ ̂ ID"L -, , , , , J. Ai .,

0 2DO 400 600 800 1000 1200 1400 1600 1800 2000 QUANTUM YIELD(r?E ), e~

Fig. 16. H istog ra m of best effective quantum yield for the Till 3 PC C D, using a fronts id e voltage of 27 V.

2.

I. J. R. Janesick, T. Elliott, S. Collins, M. M. Blouke, and J. Freeman, "Scientific charge-coupled devices;' Opt. Eng, 26(8), 692-714 (1987). M. M. Blouke, J. R. Janesick, T. Elliott, J. E. Hall, M. W. Cowens, and P. J. May, "'Current status of 800X800 charge-coupled-device imager sensor," Opt. Eng. 26(9), in press (1987),J, R. Janesick, J. Hynecek, and M. M. Blouke,, "Virtual phase imager for Galileo, "in Solid Stale Imagersfor Astronomy, J. C. Geary and D. W. Latham, ed., Proc. SPIE 290, 165-173 (1981)."J. Janesick, T. Elliott, T. Daud, J. McCarthy, and M. Blouke, "Backside charging of the CCD," in Solid State Imaging Arrays, K. N. Prettyjohns and E. L. Dereniak, eds., Proc. SPIE 570, 46-79 (1985). J. Janesick, T. Elliott, T. Daud, and D. Campbell, "The CCD flash gate," in Instrumentation in Astronomy VI, D. L. Crawford, ed., Proc. SPIE 627,543-582(1986).J. Janesick, D. Campbell, T. Elliott, T. Daud, and P. Ottley, "Flash technology for CCD imaging in the UV," in Ultraviolet Technology, R. Huffman, ed,, Proc. SPIE 687, 36-55 (1986).J... Janesick,, T. Elliott,, J. McCarthy, H. Marsh, and, S. Collins, "'Present and, future CCDs for UV and x-ray scientific measurements," IEEE Trans. Nuc. Sci. NS-32 (I),, 409-416 (.1985). s

James R. Janesick: Biography and photograph appear with the Guest Editorial in this issue.

Kenneth IP. Klaasen is the supervisor of the 1 P h otosci e nee G rou p at J P L. H e recei ved h i s BS :/: deg r e e i n p hys i cs fro m C a I v i n Co 11 eg e, G r a nd

Rapids, Mien., and his MS in aerospace engi­ neering from the University of Michigan in 19 6 9, H e h a s bee n i n vo I ved i n m a n y so I a r sys - tern exploration missions, including Mariner

1 I 10, the Viking Orbiter, and currently the Gali- | ^ leo mission to Jupiter. He has served as the I experiment representative for the imaging

experiments on these missions and is cur­ rently a member of the Galileo Imaging Science Team. His primary responsibilities have included imaging system calibration, experiment planning, and mission operations. He has published papers on the rotation period of Mercury, Venus atmospheric dynamics, and the pho­ tometry of Mercury, Mars, and the Martian moons, P hobos and Deirnos, as well as several on various spacecraft imaging instruments and experiments.

6. ACKNOWLEDGMENTSThe authors acknowledge many rewarding conversations on the subject of CCE with Morley Blouke (father of the TI 3PCCD), Taher Daud, Andy Collins, Dave Campbell, James DeWitt, Arsham Dingizian, and James McCarthy,., We alsothank Deborah Durham, for reviewing this paper. The research, described was carried out by the Jet Propulsion, Laboratory, California Institute of Technology, under con­ tract with the National Aeronautics and Space Administration.

Stythe T. Elliott was born in Van Nuys, Cali­ fornia. He received his BA degree in geography from the California State University at North- ridge. He joined the Jet Propulsion Laboratory in 1979 and is presently working on develop­ ing charge-coupled devices for NASA space imaging systems. He received a NASA Achievement Award in 1986. He is currently involved with the development of CCDs used in the wide field/planetary Hubble Space Tele­ scope camera.

980 / OPTICAL ENGINEERING / October 1 987 / Vol. 26 No. 10

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/21/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx