High Definition 10µm pitch InGaAs detector with Asynchronous Laser Pulse Detection mode

R. Fraenkel, E. Berkowicz, L. Bykov, R. Dobromislin, R. Elishkov, A. Giladi, I. Grimberg,

I. Hirsh, E. Ilan, C. Jacobson, I. Kogan, P. Kondrashov, I. Nevo, I. Pivnik and S. Vasserman

SemiConductor Devices, P.O. Box 2250, Haifa 31021, Israel

ABSTRACT

In recent years SCD has developed InGaAs/InP technology for Short-Wave Infrared (SWIR) imaging. The first product, Cardinal 640, has a 640512 (VGA) format at 15µm pitch, and more than a thousand units have already been delivered. We now present Cardinal 1280, having the smallest pitch available today (10µm), with a 12801024 (SXGA) format. Cardinal 1280 addresses both long-range daylight imaging, and passive or active imaging in Low Light Level (LLL) conditions. The Readout Integrated Circuit supports snapshot imaging at 13 bit resolution with a frame rate of 160Hz at full format, or a frame rate of 640Hz with 22 binning. It also has a Low Noise Imaging (LNIM) mode with 35e- readout noise with internal Correlated Double Sampling (CDS). An asynchronous Laser Pulse Detection (ALPD) mode is implemented with 2x2 binning in parallel to SWIR imaging (with 10 µm resolution). The new 10 µm pixel is sensitive down to the visible (VIS) spectrum, with a typical dark current of ~ 0.5fA at 280K, and a quantum efficiency >80% at 1550nm. The Focal Plane Array is integrated into a ruggedized, high vacuum integrity, metallic package, with a Thermo-Electric Cooler (TEC) for optimized performance, and a high grade Sapphire window. In this paper we will present the architecture and preliminary measurement results. Keywords: Infrared Detector, Focal Plane Array, SWIR, High Definition, 10µm pitch, ALPD

1. INTRODUCTION In recent years SCD has developed InGaAs/InP technology for Short-Wave Infrared (SWIR) imaging. The first product, Cardinal 640, was launched in 2013 and since then more than a thousand units have already been delivered to numerous customers [1, 2]. The SWIR technology offers many benefits for a variety of electro-optical systems and applications: the image is reflective and thus more natural and intuitive compared with thermal. It penetrates fog and haze much better than CCD or CMOS detectors, especially for long range distances. For low light level conditions it can utilize the night glow phenomenon and unlike standard image intensifiers it can address the full intensity spectrum (from daylight to overcast). Another important advantage is the capability to perform active or gated imaging with "eye-safe" laser source. In this paper we present Cardinal 1280 designed for the new generation of high-end platforms such as airborne payloads and long range imaging systems. The FPA has the smallest pitch available today (10µm), with a

12801024 (SXGA) format. The pixel is sensitive down to the visible (VIS) spectrum, with a typical dark current of ~ 0.5fA at 280K, and a quantum efficiency >80% at 1550nm. The digital Readout Integrated Circuit (ROIC) supports snapshot imaging at 13 bit resolution with a maximum frame rate of 160Hz at full format and up to 640Hz with 22 binning. For imaging the ROIC includes 3 gain levels to suit various scenarios. The high gain mode is optimized for low light level conditions with 35e- readout noise. Our legacy Asynchronous Laser Pulse Detection (ALPD) mode [3] was also implemented with 2x2 binning superimposed on the standard imaging function (with 10 µm resolution). The Focal Plane Array is integrated into a ruggedized, high vacuum package, with a Thermo-Electric Cooler (TEC) and a high grade Sapphire window. Special means were implemented to reduce stray light and support high F/numbers as required by high end, long range systems. The combination of all these features makes Cardinal 1280 an ideal choice for the new generation of SWIR long-range daylight imaging platforms, as well as passive and active imaging under low light level (LLL) conditions

The paper is organized as follows: In section 2 we describe the ROIC's architecture and specification. In section 3 we present preliminary measurement results for the 10µm pixel. In the last section we will discuss the product status and our future plans.

2. ROIC ARCHITECTURE AND SPECIFICATION The Cardinal 1280 ROIC is a successor to the 2 previous designs: the multifunction SNIR [3] and the 10µm Blackbird detector [4]. While the pixel design and the special operation modes are derived from SNIR, the periphery and the interface are based on the Blackbird legacy. The ROIC’s main features are:

Snapshot integration with on-chip A2D Conversion at 13 bit resolution Sub LVDS video readout 160Hz Maximum frame rate (640Hz with 2X2 binning) Day light imaging: 2 gain levels (medium and low) Low light level imaging: high gain level with floor noise lower than 35 electrons (with CDS) Asynchronous Laser Pulse Detection (ALPD) at 2X2 binning (one stage per 4 adjacent pixels) Improved active imaging with a 2µsec time constant

The ROIC functional block diagram is shown in Figure 1. The ROIC pixel matrix outputs the image data through column wires to a column-parallel ADC array. Additionally, an ALPD bit is output indicating the detection of a laser pulse in each 4 pixel group. The image data is converted by the ADCs at 13 bit resolution and packed with the ALPD bit to form a 14 bit word. The pixel data at the ADC output is multiplexed to the video output. The digital controller provides all necessary controls for the internal blocks and handles the communication with the system. The controller enables the user to set various sensor parameters via communication: e.g. gain, exposure, windowing, readout direction. The internal structure of the controller enables high flexibility is programming various operation modes to support a wide variety of applications. Figure 2 exhibits the ROIC pixel structure: each 10µm pixel contains 3 basic modes: daylight imaging, low light level imaging and active imaging. Due to area constraints the ALPD function is common to a group of 4 pixels as depicted in the figure. Each one of the 4 diodes can detect the laser pulse and activate the ALPD bit (and of course it can be done by more than one diode simultaneously). The ALPD with an effective VGA format is superimposed on the standard 13bit HD image.

Figure 1: Cardinal 1280 ROIC Functional Block Diagram

Figure 2: Cardinal 1280 ROIC Pixel

We now describe in detail the various operation modes of the ROIC:

Daylight imaging – for daylight imaging the ROIC provides 2 gain levels. The medium gain consists of an Integrate-While-Read (IWR) 0.5Me capacitor that should cover most of the day time scenarios. For specific

applications that require higher dynamic range one can use the 1Me capacitor, which is Integrate-Then-Read (ITR). The calculated readout noise is 150 and 250 electrons respectively. The expected power dissipation is roughly 125mW for 30 Frames Per second (FPS).

Low light level imaging – the Capacitive Trans Impedance Amplifier (CTIA) "high gain" mode supports low light level scenarios. The capacitor is 15Ke per pixel with a readout noise of 35 electrons (with Correlated Double sampling (CDS)). For extremely low light conditions, one can utilize the binning mode that will provide an effective 20µm VGA array with double the signal-to-noise (SNR) ratio.

Active imaging – active and gated imaging is growing in importance in recent years for both military and civilian applications [5, 6]. Hence a considerable effort was invested in optimizing this feature in Cardinal 1280. A time constant < 2µsec can be achieved for a 128x128 window of interest. In Figure 3 we exhibit the simulated response of Cardinal 1280 compared with measured data of Cardinal 640. The time constant is less than 2 µsec compared with 5 µsec for the earlier design. As a result, the clutter rejection is much better.

ALPD – this is an extremely useful feature [7] especially for daylight SWIR applications. In the Cardinal 640 it was implemented for each pixel (15µm pitch) but due to area constraints of the 10µm pixel it was decided to implement one cell per four pixels (see Figure 2). The threshold laser sensitivity for 90% operability is estimated at 500 electrons/pixel.

Figure 3: Comparison between Cardinal 1280 and Cardinal 640 Active Imaging

-6 -4 -2 0 2 4 6 80

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Delay between exposure start and sensor illumination [usec]

Res

pons

e [A

.U]

Cardinal 640

Cardinal 1280

2.9 3 3.1 3.2 3.3 3.4 3.5 3.6 3.710-1

100

101

102

1000/Temperature (1/K)

Dar

k C

urre

nt (f

A)

Cardinal 640 - 15Cardinal 1280 - 10

3. 10 MICRON PIXEL TECHNOLOGY The Cardinal 1280 diode array is based on the mature 15µm P-i-N pixel technology that was developed a few years ago [1]. The scale down from 15µm to 10µm is not a trivial task: The increased ratio of surface to volume imposes new design rules in order to maintain high Quantum Efficiency (QE) and low dark current which is essential for low light level applications. These parameters are also affected by the various growth parameters. Specifically, the active layer thickness needs to be controlled in order to reduce the crosstalk between neighboring pixels. The InGaAs array is hybridized to the ROIC where the Indium bumps become considerably smaller which is an additional challenge.

The pixel technology was developed on 2 platforms: test-chip arrays based on the SNIR ROIC followed by hybridization to an existing 10µm ROIC [4]. Figure 4 shows a comparison between the measured dark current of the standard 15µm pixel (red circles) and the new 10µm pixel (blue squares). As expected the extracted activation energy is similar, but the current is roughly correlated to the area with a typical value of ~ 1fA at 20degC for the 10µm pixel.

Cross Talk (XT) measurements of the 10µm pixel are presented in Figure 5. The measurements were performed using Point Spread Function (PSF) measurement setup with a cavity Black Body (BB) set typically to 1000C. The radiation is passed through a filter wheel with a selected 1.5µm narrow-band filter and diffraction limited SWIR lens optics. The FPA is set on a translation stage where its movement is controlled by a precision DC motor controller. The pixel's net spatial photo-response is calculated by de-convoluting the measured 2D spatial response with the theoretical diffraction limited beam. This resulting spatial response or PSF, is then presented in a two dimensional color map. We provide the photo-signal integrated percentage values within 10µm pitch size blocks, which is the measured Cross Talk (Figure 5a). The slight asymmetry in the image is related to some measurement artifacts. The calculated Modulation Transfer Function (MTF) is exhibited in Figure 5b where the ideal (step function) 10µm curve is shown for reference.

Figure 4: Dark Current Temperature Dependence for 15 and 10 Micron Pixels

x ( m)

y (

m)

0.1%

0.2%

0.4%

0.2%

0.0%

0.1%

2.4%

8.6%

2.3%

0.1%

0.5%

8.7%

53.2%

9.0%

0.5%

0.2%

2.3%

8.1%

2.4%

0.2%

0.0%

0.1%

0.2%

0.1%

0.0%

-25 -15 -5 5 15 25

-25

-15

-5

5

15

25

-3

-2.5

-2

-1.5

-1

-0.5

0

QE

50 100 150 200 250 300

50

100

150

200

250

0.78

0.8

0.82

0.84

0.86

0.88

0 20 40 60 80 100 120 140 160 1800

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Frequency (1/mm)

MTF

Cardinal 10 pitchIdeal 10 pitch

0 500 1000 1500 2000 2500 3000 35000

5

10

15

20

25

30

Signal

Noi

se2 (D

L2 )

Figure 5: Measured PSF on a log scale (a) and Calculated MTF (b) of the 10µm Pixel

Quantum Efficiency (QE) measurements were performed with a halogen/tungsten lamp, passed through a 1550nm bandpass filter and integration sphere. The illumination intensity is measured independently and controlled by a sphere-optics variable aperture setup. Figure 6 depicts the calculated QE per pixel on a test chip 10µm FPA. The measured electronic gain of 124e/DL, for the nominally 0.5Me 13bit capacitor, is used for the QE calculation. The measured spatial low frequency pattern of the photo-response is due only to an un-normalized illumination effect. After extracting the spatial illumination dependence, the net QE is found to be 0.86.

Figure 6: Measured QE in a test chip (a) and Gain (b) for several 10µm pixels

4. PRODUCT STATUS AND FUTURE PLANS The Cardinal 1280 is in the final stages of development. The InGaAs diode array is already fully functional, and in order to evaluate the operability and uniformity we have hybridized it to an existing 10µm pitch ROIC. The daylight image is exhibited in Figure 7: The operability is well above 99.5% and the raw non-uniformity is similar to our mature 15µm technology.

We have also finalized the design and production of a ruggedized, high vacuum integrity, metallic package, with a Thermo-Electric Cooler (TEC) which is shown in Figure 8. The window is made of high grade Sapphire aimed for high end applications. Special attention was devoted to the elimination of "stray light" and to the support of very high F/number optics. All this makes the Cardinal 1280 an ideal candidate for high end payloads and other airborne systems.

We expect to deliver the first prototypes in Q3/2016, with low rate production towards the end of Q4. The basic specification of the product is summarized in Table 1.

Figure 7: Imaging with an existing 10µm ROIC

Figure 8: Cardinal 1280 Package

Table 1: Cardinal 1280 Basic Specification

Parameter Typical Value Format & Pitch 1280x1024, 10μm Spectral Range 0.4 - 1.7μm (VIS-SWIR) Quantum Efficiency >80% at 1550nm Dark current < 1fA @ 280K Pixel Operability > 99.5% NEP < 0.5fW @ High Gain, 27msec

integration time Well capacity and ROIC noise (typical)

High gain - 10Ke, 35e ITRMedium gain – 0.5Me, 170e, IWR Low gain – 1Me, 350e, ITR

Maximum FR at full window (Medium and low gain modes)

160 F/s @ 13 bit resolution (70MHz clock rate) 640 F/s with Binning

Windowing Flexible, 2 rows step

FPA Power Dissipation < 150mW @ 60 F/s , Standard Imaging

Active Imaging time constant 2µsec @ CTIA stage ALPD sensitivity < 500e per pixel @ 90%

operability Ambient operating temperature

-40oC to 71oC

Ambient non-operating temperature

-54oC to 80oC

Package Metallic (vacuum tight) , 34 x 34 x 10 mm3, weight < 50 gr

Vacuum lifetime > 14 years @ 25oC ambient Cooling capability ΔT ≥ 40oC @ 30oC ambient

5. SUMMARY AND CONCLUSIONS In this paper we have presented SCD's new InGaAs detector Cardinal 1280 designed for the new generation of high-end SWIR platforms such as airborne payloads and long range imaging systems. The FPA has the smallest pitch available today in industry (10µm), with a 12801024 (SXGA) format. The pixel is sensitive down to the visible (VIS) spectrum, with a typical dark current of ~ 0.5fA at 280K, and a quantum efficiency >80% at 1550nm. The digital Readout Integrated Circuit (ROIC) supports snapshot imaging at 13 bit resolution with a maximum frame rate of 160Hz at full format and up to 640Hz with 22 binning. For imaging the ROIC supports 3 gain levels for various scenarios. The high gain mode is optimized for low light level conditions with 35e- readout noise. Our legacy Asynchronous Laser Pulse Detection (ALPD) mode was also implemented with 2x2 binning superimposed on the standard imaging function (with 10 µm resolution).

ACKNOWLEDGEMENTS The work presented here was supported over the years by the Israeli Ministry of Economics. We are in debt to a large group of engineers and technicians who conducted this work. Their dedicated work and contribution to the development and production of the detectors is highly appreciated.

REFERENCES

1. L. Shkedy et al. "Multifunction InGaAs detector with on-chip signal processing", Proc. SPIE (2013). 2. R. Fraenkel et al. "SCD's cooled and uncooled photo detectors for NIR-SWIR", Proc. SPIE (2012). 3. L. Langof et al. "Advanced multifunction-function infra-red detector with on-chip processing", Proc. SPIE (2011). 4. G. Gershon et al. "3 Mega-pixel InSb Detector with 10µm Pitch", Proc. SPIE (2013). 5. Baker et al. "A Low Noise, Laser-Gated Imaging System for Long Range Target Identification", Proc. of SPIE

(2004) 6. J. Bentell et al. " Flip Chipped InGaAs Arrays for Gated Imaging with Eye-Safe Lasers" 7. US patent 9,215,386: "Detector pixel signal readout circuit using an AC signal component in implementing an event

detection mode per pixel"