- Email: [email protected]
Design and characterisation of a highly miniaturised radiation monitor HMRM
Nuclear Instruments and Methods in Physics Research A 731 (2013) 154–159
Contents lists available at ScienceDirect
Nuclear Instruments and Methods in Physics Research A journal homepage: www.elsevier.com/locate/nima
Design and characterisation of a highly miniaturised radiation monitor HMRM N. Guerrini a,n, R. Turchetta a, D. Griffin a, T. Morse a, A. Morse a, O. Poyntz-Wright a, S. Woodward a, E. Daly b, A. Menicucci b, H. Araujo c, E. Mitchell c a
STFC—Rutherford Appleton Laboratory, UK ESA—Noordwijk, The Netherlands c Imperial College London, UK b
art ic l e i nf o
a b s t r a c t
Available online 28 June 2013
Reliable data on the ionising radiation environment is regarded as very important to ensure an efficient design and operation of spacecraft. Here we present a novel Highly Miniaturised Radiation Monitor (HMRM) that aims to greatly reduce costs and complexity of radiation detectors. At the core of the current design is a CMOS Image Sensor. Size and mass are considerably reduced thanks to this approach and there is also scope for a reduction in power consumption. This makes the HMRM much easier to integrate on a spacecraft. The innovative architecture of the proposed radiation monitor will also make particle identification possible. The image sensor is based on a 50 by 51 pixel array. The selected pixel is a 4T, to reduce the noise. The array is read out in snapshot mode at a frame-rate of 10,000 fps. Biasing currents and voltages are generated on-chip to reduce the number of signals required to control the sensor. The sensor is designed to work on a large range of temperatures, from −40 1C to +80 1C; hence a temperature sensor has been integrated. The digital output data is obtained with a three-bit column parallel ADC with programmable thresholds. An analogue readout has been also designed to characterise and debug the ASIC. In this following paper we also want to present the results obtained from the measurements on the prototype. Preliminary PTC plots show a gain of 60 mV/e− with CDS and a noise of 17 e− rms, which includes the noise from the external board. & 2013 Elsevier B.V. All rights reserved.
Keywords: Radiation monitoring Radiation hard design CMOS sensors
1. Introduction The Highly Miniaturised Radiation Monitor (HMRM) is being developed by the UK Science and Technology and Facility Council (STFC), Rutherford Appleton Laboratory (RAL) and Imperial College London, within the framework of a European Space Agency (ESA) technology development contract. The purpose of the HMRM is to provide housekeeping data on the ionising radiation environment in and around spacecraft and spacecraft subsystems. The information provided by the HMRM can be used to assist in the planning of spacecraft and payload operations, provide evidence for the analysis of in-flight anomalies, provide data for correlation of performance degradation of systems during flight, and help Engineers gain a better understanding of the performance of space hardware to inform future design activities.
n
Corresponding author. Tel.: +44 1235 44 6212. E-mail address: [email protected] (N. Guerrini).
0168-9002/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.nima.2013.06.073
The aim of the current phase of the overall HMRM development programme is to design, manufacture and characterise a prototype version of the monitor. In the subsequent programme phases, the HMRM will undergo further development, testing, and obtain space flight qualification so that it can become generally available for integration into future space missions. Monitors currently in use weigh more than 2 Kg, require several watts of power, are quite large (410 cm in each dimension). Our aim is to develop a “chip sized” (that is smaller, lighter, and less power hungry) prototype radiation monitor suitable for use in a wide range of applications such as coarse radiation housekeeping save and alerting functions, and support to platform and payload systems. In order to obtain an instrument that represents a real progress compared to the current state-of-the-art several requirements have been set. First of all the radiation monitor has to minimise the number of external components; saving, space, power, and keeping the cost low. Total power dissipation lower than 200 mW and a weight less that 20 g were the initial targets for power and mass. Because of its nature the instrument has to be able to
N. Guerrini et al. / Nuclear Instruments and Methods in Physics Research A 731 (2013) 154–159
withstand a TID of at least 100 kRad whilst being latch-up free. Operating in a space environment sets also challenging requirements in terms of operating temperature range (from −40 to +80 1C). Finally, in-flight calibration has also to be possible. For what concerns the functionalities, HMRM has to be able to perform as a dosimeter, together with a particle rate metre function; up to 107 particles/cm2/s. In addition we also want to achieve particle-species identification. More details about this will be provided in Section 2. Because of the requirements presented, and the need for compactness yet flexibility, the choice was made to base the entire instrument on active pixel sensor technology (APS). The use of CMOS process helps in keeping the cost low while still allowing the integration of a large number of complex functionalities on a single chip. The sensor was manufactured on standard (low) and high resistivity epitaxial substrates: the former being the baseline for the technology, with the latter being the one which will be used in the instrument. The high-resistivity allows the charge to be collected by drift and not diffusion, thus reducing the overall cross-talk. In both substrates, the epitaxial layer is 12 mm thick.
155
The design has been then validated with full simulations of the geometry in five different reference orbits. The four detectors are effectively sampling the energy deposition curve, thus allowing particle identification (electron or proton), not just particle detection. Moreover the energy of the incident particle can be also extrapolated. The results of this extensive simulation campaign are presented in Fig. 3a and b. Fig. 3a shows the results for electrons. On the horizontal axis the energy of the incoming electron is reported, while on the vertical axis is the frequency of the hits. Data relating to different detectors in the stack are shown with different shades. The results show an operative range from 0.063 to 6 MeV for electrons, and from 1.3 to 500 MeV for protons. The maximum omni-directional flux considered is 1 108 cm−2 s−1. In particular electrons from 0.063 MeV to 0.1 MeV (see Fig. 3a) will be detected only by the first sensor in the stack (see Fig. 1), electrons from 0.1 MeV up to 0.5 MeV will be detected by the first and the second sensor and so on. A similar situation will occur for the protons (Fig. 3b).
3. Architecture
23.4
mm
17mm
m
The energy deposited in a single sensor by a particle with low kinetic energy may equal the deposit from a higher energy particle, for example a minimum ionising particle (MIP). This is a result of the differing energy loss rates and penetration depths in the detection medium, as illustrated in Fig. 1. Attempts to identify particle energy through energy deposit measurements in a single sensor will therefore be subject to this, and other, ambiguities Thin sensors, often known as ΔE detectors [1], may be used to measure the energy loss rate, dE/dx, of transmitted particles. In combination with a second, thicker sensor (which stops the particle) the total particle energy may also be measured, giving two values (E, ΔE) from which to make an identification. This method has, for example, been used successfully to identify ion particle species in a mixed environment [2]. By using a stack of thin sensors in a “telescope” configuration, multiple ΔE measurements may be obtained and used in similar identification schemes. Although it is not possible to stop the most energetic particles, transmission through the full stack gives a lower limit for these energies. Full simulations of the prototype design (shown in Fig. 2) have been performed using the Geant4 Monte Carlo toolkit [3], involving more than 109 particles. Particle identification algorithms, developed using simulated monitor response models, are now being optimised in preparation for operational testing of the HMRM.
The HMRM instrument block description is presented in Fig. 4. In Figs. 2 and 3, four APS detectors have been considered. Due to mechanical constrains during the final assembly the number of
30m
2. Application
HMRM ASIC
PCBs
Outer casing Flexible connections between PCBs
Fig. 2. GEANT4 simulation model (casing and structural layers are transparent for clarity).
Fig. 1. Detectors with different number of layers.
156
N. Guerrini et al. / Nuclear Instruments and Methods in Physics Research A 731 (2013) 154–159
Fig. 3. Hits vs. energy for the detector stack.
Shielding is required between the detectors to reduce the energy of highly-penetrating radiation and permit the monitor to distinguish between electron and proton fluxes. This is achieved with a combination of the shielding provided by the silicon substrate of the ASICs, and a single Titanium shield between the last two detectors in the stack. A shield in the front aperture of the monitor is required to define the low energy threshold of the monitor, and to ensure that stray-light photons do not produce a response in the detectors. This shielding is produced by a thin metalized polymer film. Full simulations of the interaction of electrons and protons with the monitor have been carried out with GEANT4 [3] and indicate that monitor is capable of characterising the radiation environment. 3.1. ASIC
Fig. 4. HMRM block diagram.
Fig. 5. HMRM case.
APS detectors has been reduced to three. GEANT4 simulations showed that the particle identification algorithm can give a reliable output even with three detectors. The electronic and sensing elements of the Integrated Monitor are encased in a Titanium shield structure labelled as the Casing in Fig. 5. The Casing ensures that the ionising radiation which interacts with the detector arrives through a controlled solid angle defined by the geometry of the Aperture and the Detector Stack. It also has the effect of limiting the geometric acceptance of the detectors and reduces the onset of regular particle pile-up events in high particle flux regimes.
The CMOS ASIC could be seen as the core of the entire instrument. Radiation is converted into charge and then into binary information in the ASIC. A commercially-available 0.18 mm CMOS Image Sensor technology have been selected. This helped in keeping the cost down and also allowed the re-use of some already designed IPs. The radiation sensitive part of the ASIC is a 50 51 pixels array. Each pixel is 20 mm, and the sensitive area is 1 mm2. The pixel used is commonly known as 4T, which allows correlated double sampling (CDS). The pixel array is read out in a snapshot mode thanks to 25 parallel output lines. The data from the pixel is then converted into 3-bit information using a column parallel ADC. The maximum frame rate is about 10,000 fps. A secondary analogue output channel has been implemented to ease the debugging of the ASIC. To reduce the number of external components and the complexity of the PCB, biasing currents and ADC thresholds are generated by on-chip DACs controlled by externally programmable registers. Voltage references are generated on-chip too via bandgap Fig. 6. A 9-bit temperature sensor has been also designed. The temperature data is embedded in the data stream. More details about the readout chain can be found in Fig. 10. 3.2. Pixel The selection of the pixel architecture came down to a comparison between two different solutions. The first, known as 3T (three transistors in the pixel) is the simplest solution, with well proven radiation hardness [4] but higher noise (see Fig. 7.a). The second, 4T, is the most popular active pixel [5], with low noise and good radiation hardness, and is presented in Fig. 7b.
N. Guerrini et al. / Nuclear Instruments and Methods in Physics Research A 731 (2013) 154–159
The noise performance of a 4T pixel already fabricated by RALSTFC [5] are summarised in Fig. 8. The noise histogram, obtained from PTC (Photon Transfer Curve) measurements [5,6] show a most probable noise of 3.6 e− rms and an average noise of 4.5 e− rms. The same measurements gave a gain of 65 μV/e− and a linear full well capacity of 17,900 e−. The radiation requirements of HMRM were set at 100– 150 kRad. The noise cumulative probability for a 4T pixel at 0 and 110 kRad has been reported in Fig. 8b. The increase in noise due to irradiation is quite clear and, as shown also in Fig. 8a, some pixels might be particularly noisy. For this reason HMRM has the possibility to mask the pixel array and exclude pixel with noise outside the specification. Measurements taken on the RAL-STFC design [5] are reported in Fig. 9. As expected, the lines in Fig. 9a and b show an increase in noise and dark current for increasing levels of irradiation, but the pixel performance is not compromised to the point that it is unusable for this specific application. The noise requirement is, in fact, a total noise between 10e-rms and 15e-rms, while the Fig. 9a shows a maximum noise of 8.4 e− rms. The particularly low noise, the
157
minor effect of irradiation on the rms noise, and the possibility mentioned above of masking the pixels lead to the choice of the 4T pixel over the 3T one. 3.3. Biasing currents The HMRM instrument is based on a number of ASICs arranged in a stack. They require a certain number of signal lines to be operated and, because of the space/weight requirements, reducing the number of line will have a beneficial impact on the entire assembly. For this reason biasing currents and thresholds for the column ADC (see Section 3.4) are generated on-chip using register controlled DACs. 3.4. Readout The pixel array (50 51) is readout in the snapshot mode and the sampled data is stored at the bottom (and top) of the array itself. Reset value and integrated value are both stored to perform Correlated Double Sampling (CDS). The pixel array is readout from two sides. Reading from one side only would have required running 50 parallel lines, something that is extremely challenging in a 20 μm pitch. There are then two ADCs per column; once again, one at the bottom and one at the top of the array. Data from the storing capacitors is multiplexed to the input of the ADC to be converted. Having 25 pixels per side and targeting a frame rate of 10,000 fps, the analogue-to-digital conversion has to take place in 100 ms, leaving 4 μs per sample. 3.5. ADC
Fig. 6. HMRM block diagram.
Fig. 7. 3T pixel (a) and 4T pixel (b) schematics.
In designing the ADC, conflicting requirements have to be balanced. On one side low power, simple and reliable design, low bit number is desired, with a single ramp ADC being the obvious choice. On the other hand, short conversion time and fine resolution will help the particle identification algorithm. In this case a flash ADC might be the best option, with a sensible impact on the power consumption. But fine resolution is required mainly for low signals, while large signals do not need it and, because of the frame rate constrains, data has to be converted in less than 4 μs. In addition to the above, CDS, subtraction of the sampled voltage from the reset voltage has to be also implemented. These considerations lead to the ADC topology presented in Fig. 11, which can be regarded as a hybrid between ramp ADC and flash ADC. Reset and sample values are transferred onto two separate capacitances connected to the input of a comparator. The other plate of the capacitor where the reset value is stored is connected
Fig. 8. Noise histogram and noise cumulative probability measured for 4T pixel.
158
N. Guerrini et al. / Nuclear Instruments and Methods in Physics Research A 731 (2013) 154–159
Fig. 9. 4T pixel noise and dark current vs. irradiation.
Fig. 12. Output registers readout architecture.
Fig. 10. Column readout block diagram.
Assuming that the comparator offset has been completely cancelled by the trimming DAC, the comparator compares RESET− (SAMPLE+THRESHOLD)40 or,in other words, if RESET−SAMPLE4THRESHOLD. Comparing the RESET and the SAMPLE values with the seven thresholds will result in a 7-bit information which is then converted into 3-bit information and shifted out as shown in Fig. 12. The HMRM has been designed with all transistors with enclosed geometry layout to mitigate the effects of radiation. For the same reason all digital blocks are based on Triple Majority Voting (TMV).
3.6. Results
Fig. 11. Column ADC block diagram.
to an 8-bit trimming DAC that compensates for the comparator offset. Each ADC has its own trimming register, hence 51 2 8 (816) trimming bits have to be set before acquiring data. In a conventional flash ADC, the data to be converted is compared with 2N−1 thresholds in order to get N-bit outputs. Our proposal is to use just one comparator and add the thresholds to the data. For each threshold added the output of the comparator is recorder and stored into a 7-bit register.
The ASIC also has an analogue readout that can be used for debugging, which can be used to more accurately evaluate the pixel performance in terms of noise and full well. Preliminary measurement results are summarised in Fig. 13, where Photon Transfer Curve (PTC) [6] and signal response are presented. The PTC plot shows the standard deviation of signal vs. the mean of the signal in DN (Digital Number). When the slope of the log–log plot is 0.5 the noise is shot noise dominated (Poisson noise). In the log–log PTC plot (left plot in Fig. 13a) the points used to identify the 0.5 slope have been highlighted with a darker colour. The results obtained using the Photon Transfer Curve method [6] are in line with expected performance, as shown in Table 1.
N. Guerrini et al. / Nuclear Instruments and Methods in Physics Research A 731 (2013) 154–159
159
Fig. 13. Measured PTC and signal response. Table 1 Comparison between expected and measured performance.
Expected Measured
Gain (mV/e−)
Noise (e-rms)
Input capacitance (fF)
Full well (e−)
Linear full well (e−)
51.2 59.9-
13.5 16.8
2.3 2.1
– 14,600
– 6600
On the internal signal path the overall gain is different; we expect it to be 40.3 mV/e−, with a noise of 13.9 e−rms. Under these conditions the S/N ratio for a MIP should be equal to 69. 4. Programme status and conlusions More extensive tests are ongoing while the prototype assembly is finalised. The HMRM will be ready imminently to fly on a mission and collect helpful data. Moreover, tests on the effects of higher substrate resistivity will also take place in the near future.
Finally, we can conclude by summarising what has been presented in this paper. We have proved that using CMOS technology it is possible to fabricate a radiation monitor and detector suitable for space applications. The solution here-in presented shows decisive advantages in terms of power consumption and mass compared to currently available solutions.
References [1] F.S. Goulding, B.G. Harvey, Identification of nuclear particles, Ann. Rev. Nucl. Sci. 25 (1975) 167. [2] D.A. Bromley, Nuclear experimentation with semiconductor detectors, IRE Trans. Nucl. Sci. NS-9 (3) (1962) 135. [3] S. Agostinelli, et al., GEANT4—a simulation toolkit, Nucl. Instr. Meth. A 506 (3) (2003) 250–303. [4] N. Guerrini et al.A high frame rate, 16 million pixels, radiation hard CMOS sensor”, Proceedings of 12th IWORID, Cambridge, UK, 2010. [5] R. Coath, et al., A low noise pixel architecture for scientific CMOS monolithic active pixel sensors, IEEE Trans. Nucl. Sci. NS 57 (5) (2010). [6] J.R. Janesick, Photon Transfer DN-λ, SPIE press, Bellingham, USA, 2007.
Contents lists available at ScienceDirect
Nuclear Instruments and Methods in Physics Research A journal homepage: www.elsevier.com/locate/nima
Design and characterisation of a highly miniaturised radiation monitor HMRM N. Guerrini a,n, R. Turchetta a, D. Griffin a, T. Morse a, A. Morse a, O. Poyntz-Wright a, S. Woodward a, E. Daly b, A. Menicucci b, H. Araujo c, E. Mitchell c a
STFC—Rutherford Appleton Laboratory, UK ESA—Noordwijk, The Netherlands c Imperial College London, UK b
art ic l e i nf o
a b s t r a c t
Available online 28 June 2013
Reliable data on the ionising radiation environment is regarded as very important to ensure an efficient design and operation of spacecraft. Here we present a novel Highly Miniaturised Radiation Monitor (HMRM) that aims to greatly reduce costs and complexity of radiation detectors. At the core of the current design is a CMOS Image Sensor. Size and mass are considerably reduced thanks to this approach and there is also scope for a reduction in power consumption. This makes the HMRM much easier to integrate on a spacecraft. The innovative architecture of the proposed radiation monitor will also make particle identification possible. The image sensor is based on a 50 by 51 pixel array. The selected pixel is a 4T, to reduce the noise. The array is read out in snapshot mode at a frame-rate of 10,000 fps. Biasing currents and voltages are generated on-chip to reduce the number of signals required to control the sensor. The sensor is designed to work on a large range of temperatures, from −40 1C to +80 1C; hence a temperature sensor has been integrated. The digital output data is obtained with a three-bit column parallel ADC with programmable thresholds. An analogue readout has been also designed to characterise and debug the ASIC. In this following paper we also want to present the results obtained from the measurements on the prototype. Preliminary PTC plots show a gain of 60 mV/e− with CDS and a noise of 17 e− rms, which includes the noise from the external board. & 2013 Elsevier B.V. All rights reserved.
Keywords: Radiation monitoring Radiation hard design CMOS sensors
1. Introduction The Highly Miniaturised Radiation Monitor (HMRM) is being developed by the UK Science and Technology and Facility Council (STFC), Rutherford Appleton Laboratory (RAL) and Imperial College London, within the framework of a European Space Agency (ESA) technology development contract. The purpose of the HMRM is to provide housekeeping data on the ionising radiation environment in and around spacecraft and spacecraft subsystems. The information provided by the HMRM can be used to assist in the planning of spacecraft and payload operations, provide evidence for the analysis of in-flight anomalies, provide data for correlation of performance degradation of systems during flight, and help Engineers gain a better understanding of the performance of space hardware to inform future design activities.
n
Corresponding author. Tel.: +44 1235 44 6212. E-mail address: [email protected] (N. Guerrini).
0168-9002/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.nima.2013.06.073
The aim of the current phase of the overall HMRM development programme is to design, manufacture and characterise a prototype version of the monitor. In the subsequent programme phases, the HMRM will undergo further development, testing, and obtain space flight qualification so that it can become generally available for integration into future space missions. Monitors currently in use weigh more than 2 Kg, require several watts of power, are quite large (410 cm in each dimension). Our aim is to develop a “chip sized” (that is smaller, lighter, and less power hungry) prototype radiation monitor suitable for use in a wide range of applications such as coarse radiation housekeeping save and alerting functions, and support to platform and payload systems. In order to obtain an instrument that represents a real progress compared to the current state-of-the-art several requirements have been set. First of all the radiation monitor has to minimise the number of external components; saving, space, power, and keeping the cost low. Total power dissipation lower than 200 mW and a weight less that 20 g were the initial targets for power and mass. Because of its nature the instrument has to be able to
N. Guerrini et al. / Nuclear Instruments and Methods in Physics Research A 731 (2013) 154–159
withstand a TID of at least 100 kRad whilst being latch-up free. Operating in a space environment sets also challenging requirements in terms of operating temperature range (from −40 to +80 1C). Finally, in-flight calibration has also to be possible. For what concerns the functionalities, HMRM has to be able to perform as a dosimeter, together with a particle rate metre function; up to 107 particles/cm2/s. In addition we also want to achieve particle-species identification. More details about this will be provided in Section 2. Because of the requirements presented, and the need for compactness yet flexibility, the choice was made to base the entire instrument on active pixel sensor technology (APS). The use of CMOS process helps in keeping the cost low while still allowing the integration of a large number of complex functionalities on a single chip. The sensor was manufactured on standard (low) and high resistivity epitaxial substrates: the former being the baseline for the technology, with the latter being the one which will be used in the instrument. The high-resistivity allows the charge to be collected by drift and not diffusion, thus reducing the overall cross-talk. In both substrates, the epitaxial layer is 12 mm thick.
155
The design has been then validated with full simulations of the geometry in five different reference orbits. The four detectors are effectively sampling the energy deposition curve, thus allowing particle identification (electron or proton), not just particle detection. Moreover the energy of the incident particle can be also extrapolated. The results of this extensive simulation campaign are presented in Fig. 3a and b. Fig. 3a shows the results for electrons. On the horizontal axis the energy of the incoming electron is reported, while on the vertical axis is the frequency of the hits. Data relating to different detectors in the stack are shown with different shades. The results show an operative range from 0.063 to 6 MeV for electrons, and from 1.3 to 500 MeV for protons. The maximum omni-directional flux considered is 1 108 cm−2 s−1. In particular electrons from 0.063 MeV to 0.1 MeV (see Fig. 3a) will be detected only by the first sensor in the stack (see Fig. 1), electrons from 0.1 MeV up to 0.5 MeV will be detected by the first and the second sensor and so on. A similar situation will occur for the protons (Fig. 3b).
3. Architecture
23.4
mm
17mm
m
The energy deposited in a single sensor by a particle with low kinetic energy may equal the deposit from a higher energy particle, for example a minimum ionising particle (MIP). This is a result of the differing energy loss rates and penetration depths in the detection medium, as illustrated in Fig. 1. Attempts to identify particle energy through energy deposit measurements in a single sensor will therefore be subject to this, and other, ambiguities Thin sensors, often known as ΔE detectors [1], may be used to measure the energy loss rate, dE/dx, of transmitted particles. In combination with a second, thicker sensor (which stops the particle) the total particle energy may also be measured, giving two values (E, ΔE) from which to make an identification. This method has, for example, been used successfully to identify ion particle species in a mixed environment [2]. By using a stack of thin sensors in a “telescope” configuration, multiple ΔE measurements may be obtained and used in similar identification schemes. Although it is not possible to stop the most energetic particles, transmission through the full stack gives a lower limit for these energies. Full simulations of the prototype design (shown in Fig. 2) have been performed using the Geant4 Monte Carlo toolkit [3], involving more than 109 particles. Particle identification algorithms, developed using simulated monitor response models, are now being optimised in preparation for operational testing of the HMRM.
The HMRM instrument block description is presented in Fig. 4. In Figs. 2 and 3, four APS detectors have been considered. Due to mechanical constrains during the final assembly the number of
30m
2. Application
HMRM ASIC
PCBs
Outer casing Flexible connections between PCBs
Fig. 2. GEANT4 simulation model (casing and structural layers are transparent for clarity).
Fig. 1. Detectors with different number of layers.
156
N. Guerrini et al. / Nuclear Instruments and Methods in Physics Research A 731 (2013) 154–159
Fig. 3. Hits vs. energy for the detector stack.
Shielding is required between the detectors to reduce the energy of highly-penetrating radiation and permit the monitor to distinguish between electron and proton fluxes. This is achieved with a combination of the shielding provided by the silicon substrate of the ASICs, and a single Titanium shield between the last two detectors in the stack. A shield in the front aperture of the monitor is required to define the low energy threshold of the monitor, and to ensure that stray-light photons do not produce a response in the detectors. This shielding is produced by a thin metalized polymer film. Full simulations of the interaction of electrons and protons with the monitor have been carried out with GEANT4 [3] and indicate that monitor is capable of characterising the radiation environment. 3.1. ASIC
Fig. 4. HMRM block diagram.
Fig. 5. HMRM case.
APS detectors has been reduced to three. GEANT4 simulations showed that the particle identification algorithm can give a reliable output even with three detectors. The electronic and sensing elements of the Integrated Monitor are encased in a Titanium shield structure labelled as the Casing in Fig. 5. The Casing ensures that the ionising radiation which interacts with the detector arrives through a controlled solid angle defined by the geometry of the Aperture and the Detector Stack. It also has the effect of limiting the geometric acceptance of the detectors and reduces the onset of regular particle pile-up events in high particle flux regimes.
The CMOS ASIC could be seen as the core of the entire instrument. Radiation is converted into charge and then into binary information in the ASIC. A commercially-available 0.18 mm CMOS Image Sensor technology have been selected. This helped in keeping the cost down and also allowed the re-use of some already designed IPs. The radiation sensitive part of the ASIC is a 50 51 pixels array. Each pixel is 20 mm, and the sensitive area is 1 mm2. The pixel used is commonly known as 4T, which allows correlated double sampling (CDS). The pixel array is read out in a snapshot mode thanks to 25 parallel output lines. The data from the pixel is then converted into 3-bit information using a column parallel ADC. The maximum frame rate is about 10,000 fps. A secondary analogue output channel has been implemented to ease the debugging of the ASIC. To reduce the number of external components and the complexity of the PCB, biasing currents and ADC thresholds are generated by on-chip DACs controlled by externally programmable registers. Voltage references are generated on-chip too via bandgap Fig. 6. A 9-bit temperature sensor has been also designed. The temperature data is embedded in the data stream. More details about the readout chain can be found in Fig. 10. 3.2. Pixel The selection of the pixel architecture came down to a comparison between two different solutions. The first, known as 3T (three transistors in the pixel) is the simplest solution, with well proven radiation hardness [4] but higher noise (see Fig. 7.a). The second, 4T, is the most popular active pixel [5], with low noise and good radiation hardness, and is presented in Fig. 7b.
N. Guerrini et al. / Nuclear Instruments and Methods in Physics Research A 731 (2013) 154–159
The noise performance of a 4T pixel already fabricated by RALSTFC [5] are summarised in Fig. 8. The noise histogram, obtained from PTC (Photon Transfer Curve) measurements [5,6] show a most probable noise of 3.6 e− rms and an average noise of 4.5 e− rms. The same measurements gave a gain of 65 μV/e− and a linear full well capacity of 17,900 e−. The radiation requirements of HMRM were set at 100– 150 kRad. The noise cumulative probability for a 4T pixel at 0 and 110 kRad has been reported in Fig. 8b. The increase in noise due to irradiation is quite clear and, as shown also in Fig. 8a, some pixels might be particularly noisy. For this reason HMRM has the possibility to mask the pixel array and exclude pixel with noise outside the specification. Measurements taken on the RAL-STFC design [5] are reported in Fig. 9. As expected, the lines in Fig. 9a and b show an increase in noise and dark current for increasing levels of irradiation, but the pixel performance is not compromised to the point that it is unusable for this specific application. The noise requirement is, in fact, a total noise between 10e-rms and 15e-rms, while the Fig. 9a shows a maximum noise of 8.4 e− rms. The particularly low noise, the
157
minor effect of irradiation on the rms noise, and the possibility mentioned above of masking the pixels lead to the choice of the 4T pixel over the 3T one. 3.3. Biasing currents The HMRM instrument is based on a number of ASICs arranged in a stack. They require a certain number of signal lines to be operated and, because of the space/weight requirements, reducing the number of line will have a beneficial impact on the entire assembly. For this reason biasing currents and thresholds for the column ADC (see Section 3.4) are generated on-chip using register controlled DACs. 3.4. Readout The pixel array (50 51) is readout in the snapshot mode and the sampled data is stored at the bottom (and top) of the array itself. Reset value and integrated value are both stored to perform Correlated Double Sampling (CDS). The pixel array is readout from two sides. Reading from one side only would have required running 50 parallel lines, something that is extremely challenging in a 20 μm pitch. There are then two ADCs per column; once again, one at the bottom and one at the top of the array. Data from the storing capacitors is multiplexed to the input of the ADC to be converted. Having 25 pixels per side and targeting a frame rate of 10,000 fps, the analogue-to-digital conversion has to take place in 100 ms, leaving 4 μs per sample. 3.5. ADC
Fig. 6. HMRM block diagram.
Fig. 7. 3T pixel (a) and 4T pixel (b) schematics.
In designing the ADC, conflicting requirements have to be balanced. On one side low power, simple and reliable design, low bit number is desired, with a single ramp ADC being the obvious choice. On the other hand, short conversion time and fine resolution will help the particle identification algorithm. In this case a flash ADC might be the best option, with a sensible impact on the power consumption. But fine resolution is required mainly for low signals, while large signals do not need it and, because of the frame rate constrains, data has to be converted in less than 4 μs. In addition to the above, CDS, subtraction of the sampled voltage from the reset voltage has to be also implemented. These considerations lead to the ADC topology presented in Fig. 11, which can be regarded as a hybrid between ramp ADC and flash ADC. Reset and sample values are transferred onto two separate capacitances connected to the input of a comparator. The other plate of the capacitor where the reset value is stored is connected
Fig. 8. Noise histogram and noise cumulative probability measured for 4T pixel.
158
N. Guerrini et al. / Nuclear Instruments and Methods in Physics Research A 731 (2013) 154–159
Fig. 9. 4T pixel noise and dark current vs. irradiation.
Fig. 12. Output registers readout architecture.
Fig. 10. Column readout block diagram.
Assuming that the comparator offset has been completely cancelled by the trimming DAC, the comparator compares RESET− (SAMPLE+THRESHOLD)40 or,in other words, if RESET−SAMPLE4THRESHOLD. Comparing the RESET and the SAMPLE values with the seven thresholds will result in a 7-bit information which is then converted into 3-bit information and shifted out as shown in Fig. 12. The HMRM has been designed with all transistors with enclosed geometry layout to mitigate the effects of radiation. For the same reason all digital blocks are based on Triple Majority Voting (TMV).
3.6. Results
Fig. 11. Column ADC block diagram.
to an 8-bit trimming DAC that compensates for the comparator offset. Each ADC has its own trimming register, hence 51 2 8 (816) trimming bits have to be set before acquiring data. In a conventional flash ADC, the data to be converted is compared with 2N−1 thresholds in order to get N-bit outputs. Our proposal is to use just one comparator and add the thresholds to the data. For each threshold added the output of the comparator is recorder and stored into a 7-bit register.
The ASIC also has an analogue readout that can be used for debugging, which can be used to more accurately evaluate the pixel performance in terms of noise and full well. Preliminary measurement results are summarised in Fig. 13, where Photon Transfer Curve (PTC) [6] and signal response are presented. The PTC plot shows the standard deviation of signal vs. the mean of the signal in DN (Digital Number). When the slope of the log–log plot is 0.5 the noise is shot noise dominated (Poisson noise). In the log–log PTC plot (left plot in Fig. 13a) the points used to identify the 0.5 slope have been highlighted with a darker colour. The results obtained using the Photon Transfer Curve method [6] are in line with expected performance, as shown in Table 1.
N. Guerrini et al. / Nuclear Instruments and Methods in Physics Research A 731 (2013) 154–159
159
Fig. 13. Measured PTC and signal response. Table 1 Comparison between expected and measured performance.
Expected Measured
Gain (mV/e−)
Noise (e-rms)
Input capacitance (fF)
Full well (e−)
Linear full well (e−)
51.2 59.9-
13.5 16.8
2.3 2.1
– 14,600
– 6600
On the internal signal path the overall gain is different; we expect it to be 40.3 mV/e−, with a noise of 13.9 e−rms. Under these conditions the S/N ratio for a MIP should be equal to 69. 4. Programme status and conlusions More extensive tests are ongoing while the prototype assembly is finalised. The HMRM will be ready imminently to fly on a mission and collect helpful data. Moreover, tests on the effects of higher substrate resistivity will also take place in the near future.
Finally, we can conclude by summarising what has been presented in this paper. We have proved that using CMOS technology it is possible to fabricate a radiation monitor and detector suitable for space applications. The solution here-in presented shows decisive advantages in terms of power consumption and mass compared to currently available solutions.
References [1] F.S. Goulding, B.G. Harvey, Identification of nuclear particles, Ann. Rev. Nucl. Sci. 25 (1975) 167. [2] D.A. Bromley, Nuclear experimentation with semiconductor detectors, IRE Trans. Nucl. Sci. NS-9 (3) (1962) 135. [3] S. Agostinelli, et al., GEANT4—a simulation toolkit, Nucl. Instr. Meth. A 506 (3) (2003) 250–303. [4] N. Guerrini et al.A high frame rate, 16 million pixels, radiation hard CMOS sensor”, Proceedings of 12th IWORID, Cambridge, UK, 2010. [5] R. Coath, et al., A low noise pixel architecture for scientific CMOS monolithic active pixel sensors, IEEE Trans. Nucl. Sci. NS 57 (5) (2010). [6] J.R. Janesick, Photon Transfer DN-λ, SPIE press, Bellingham, USA, 2007.