A Variable Dynamic Range Single-Photon Imager Designed for Multi-Radiation Tolerance

Full text

1
A Variable Dynamic Range Single-Photon Imager Designed
for Multi-Radiation Tolerance
Lucio Carrara**, Matt Fishburn+, Cristiano Niclass*
Noémy Scheidegger†, Herbert Shea‡, Edoardo Charbon*+
*Quantum Architecture Group (AQUA), EPFL, Lausanne, Switzerland
+CAS Dept. of Electrical Engineering, Computer Sciences, and Mathematics (EWI), TU
Delft, Delft, The Netherlands
**ESPROS Photonics, Baar, Switzerland
†Oerlikon Space, Zurich, Switzerland
‡Microsystems for Space Technologies Laboratory (LMTS), EPFL, Lausanne, Switzerland
Contact author: Edoardo Charbon (e.charbon@tudelft.nl)
Abstract
We present new measurements conducted on a single-photon imager implemented in CMOS
technology. The sensor was designed to sustain massive doses of Gamma irradiation, X-rays,
and proton bombardment. The chip was also designed to operate continuously in large
magnetic fields without measurable noise and distortion.
The imager consists of an array of 32x32 photon counting pixels with a pitch of 30µm. Each
pixel comprises a single-photon avalanche diode (SPAD) based on [1], a 1-bit counter, and
miniaturized readout electronics. The chip measures 2.00x2.35mm2. The block diagram and a
photomicrograph of the sensor are shown in Fig. 1. The pixel array is read out in rolling
shutter mode via the high-speed row decoder and may be reset after each read operation or
read out non-destructively.
Fig. 1. Block diagram of the sensor (left). Photomicrograph of the sensor system and pixel inset (right).

2
All 32 columns are read in parallel, thus enabling a complete 1024-pixel frame readout in
!
Tmin
=1.2µs with 1-bit depth. To achieve a higher number of gray levels we accumulate N
frames, thus reaching an intensity resolution of log2(N) bits at the expense of lower frame
rates. The saturation count rate is
!
1/ Tmin
, SNRmax for integration time tint is computed as
!
SNRmax =20log tint
Tmin
"
#
$
%
&
' (10log tint
Tmin
+Var[DC]
"
#
$
%
&
'
,
where the noise power is given by the sum of Poisson noise power and Var[DC], i.e. the
variance of the stochastic process underlying dark count generation. The latter is
approximated by the average of dark counts during integration, or DCR tint, where DCR is the
dark count rate of the detector. In this paper we use median DCR. We believe that this figure
is a better representation of the noise performance of the chip as it represents the DCR upper
bound for 50% of the pixels.
The intrinsic dynamic range of each pixel is limited from below by DCR and from above by
the inverse of dead time. In this design it amounts to 120dB at 1s integration. The dynamic
range of the system is programmable by acting upon the speed of the readout. Tab. 1 shows
the trade-off between speed and dynamic range in comparison to the maximum achievable
dynamic range assuming a dead time of 40ns and negligible DCR. In modern CMOS
technologies, high readout speeds may be easily achieved, thus it is often advantageous to
reduce the counter depth on-pixel in favor of reduced pitch. In this design, for example, a
frame rate comparable to [2]was achieved with 3 bits of counter depth (as opposed to 8 bits),
but with 11 times higher pixel density. The pixel density of this design is 8 times higher than
that of [3],[4], a 60x48 SPAD array whose pixel comprises two 8-bit counters and a pitch of
85µm.
Frame rate (fps)
Dead time limited
dynamic range (dB)
Counter depth limited
dynamic range (dB)
System counter
depth (bits)
833,333
29
6
1
104,166
47
18
3
3,255
77
48
8
813
89
60
10
25
118
90
15
Tab. 1. Trade-off between frame rate and dynamic range. Reported are dead time limited and counter
depth limited dynamic ranges.
The sensor was designed and laid out using techniques to maximize radiation resilience [5].
Details of the pixel and of the design can be found in [6], whereas the chip has now been
tested for much larger doses of gamma radiation reaching 300 kGy (30 MRad, Si). To the
best of our knowledge, this is the largest dose ever used on a SPAD array and it equals the
experiment performed by Fossum et al. [7]. However, in our work the observed noise
degradation is negligible if compared to that reported in [7], where dark current increased
several orders of magnitude.
The results of the full characterization of the chip are reported in Tab. 2. The radiation testing
was performed in four separate measurement campaigns. All measurements from these
campaigns were with an excess bias of 2.8 V, unless otherwise noted. The gamma radiation
campaigns were performed at ESA-ESTEC in Noordwijk (Netherlands) at the Reactor
Institute Delft (RID) in Delft (Netherlands). During the campaign at ESTEC, the sensor
received a total dose of 14 kGy (1.4 Mrad, Si) at a dose rate of roughly 4 mGy / sec. During
the second campaign at RID, the sensor received a total dose of 300 kGy (30 Mrad, Si) at a
dose rate of approximately 800 mGy / sec.

3
In the third experiment, the sensor was exposed to two separate proton beams at a constant
energy of 11MeV and 60Me V, respectively. The experiment was performed at the Paul
Scherrer Institute in Villigen (Switzerland).
In the fourth experiment, the chip was exposed to a massive X-ray dose at the University
Institute for Radiation Physics in Lausanne (Switzerland). The X-ray beam, generated by a
Bipolar Metal-Ceramic tube Comet-Yxlon TU 320-D03, achieved fluence and total dose
levels reported in the table, that also lists the results of the DCR change. Preliminary
irradiations were performed without any filtering and using a large collimation (27 mm). A
series of irradiations at 15kV, 120kV and 200kV showed negligible impact on DCR, PDP,
and afterpulsing.
Irradiation
type
Source
Fluence/
Flux
Dose (Si)
Initial
DCR
Final
DCR
DCR after Annealing
(anneal time, temp)
41.6 mGy/s
14 kGy
153
13487
276 (172 h, 80° C)
Gamma
Co60
797.5 mGy/s
300 kGy
128
N/A
25877 (1500 h, 20° C)
4.3 AsV2
0.25 mGy
204
N/A
204 (1 min, 20° C)
324 AsV2
0.25 mGy
204
N/A
204 (1 min, 20° C)
X *
Comet-Yxlon
TU320-D03
900 AsV2
0.5 mGy
204
N/A
204 (1 min, 20° C)
1.8x107p/cm2/s
(11MeV)
400 Gy
140
6298
3884 (10 d, 20° C)
Proton
Accelerator
8.3x107p/cm2/s
(60 MeV)
400 Gy
142
6290
1299 (21 d, 20° C)
Tab. 2. Irradiation experiment summary. The median DCR is reported in Hz at room temperature.
Photon detection and afterpulsing probabilities, and maximum frame rate remained unchanged after all
three types of irradiation. The final DCR, when reported, was measured with the sensor still being
irradiated. *Note: X-ray measurements were performed with an excess bias of 3.3 V rather than 2.8 V.

4
References
[1] C. Niclass A. Rochas, P.A. Besse, and E. Charbon, “Design and Characterization of a CMOS 3-D Image
Sensor Based on Single Photon Avalanche Diodes”, IEEE J. of Solid-State Circuits, Vol. 40, N. 9, Sep.
2005.
[2] S. Tisa, F. Guerrieri, A. Tosi, F. Zappa, “100kframe/s 8 Bit Monolithic Single-photon Imagers”, IEEE
European Solid-State Device Research Conference (ESSDERC), Sep. 2008.
[3] C. Niclass, C. Favi, T. Kluter, F. Monnier, E. Charbon, “Single-Photon Synchronous Detection”, IEEE
European Solid-State Circuits Conference (ESSCIRC), Sep. 2008.
[4] C. Niclass, C. Favi, T. Kluter, F. Monnier, E. Charbon, “Single-Photon Synchronous Detection”, to appear,
IEEE Journal of Solid-State Circuits, Vol. 44, N. 9, Sep. 2009.
[5] J. Bogaerts, B. Dierickx, and C. Van Hoof, “Radiation-induced Dark Current Increase in CMOS Active
Pixel Sensors,” in SPIE Photonics for Space Environments VII, Vol. 4134, 2000, pp. 105–114.
[6] L. Carrara, C. Niclass, N. Scheidegger, H. Shea, E. Charbon, “A Gamma, X-ray and High Energy P roton
Radiation-Tolerant CMOS Image Sensor for Space Applications”, IEEE Intl. Solid-State Circuits
Conference (ISSCC), Feb. 2009.
[7] El-Sayed Eid, T.Y. Chan, E. R. Fossum, R. H. Tsai, R. Spagnuolo, J. Deily, W. B. Byers, J. C. Peden,
“Design and Characterization of Ionizing Radiation-Tolerant APS Image Sensors up to 30 Mrd (Si) Total
Dose”, IEEE Trans. Nuc. Sci, Vol. 48, N. 6, Dec. 2001.