Trends in Integrated Circuit Design For Particle Physics Experiments
Trends in Integrated Circuit Design For Particle Physics Experiments
Trends in Integrated Circuit Design For Particle Physics Experiments
Abstract. Integrated circuits are one of the key complex units available to designers of
multichannel detector setups. A whole number of factors makes Application Specific Integrated
Circuits (ASICs) valuable for Particle Physics and Astrophysics experiments. Among them the
most important ones are: integration scale, low power dissipation, radiation tolerance. In order
to make possible future experiments in the intensity, cosmic, and energy frontiers today ASICs
should provide new level of functionality at a new set of constraints and trade-offs, like low-noise
high-dynamic range amplification and pulse shaping, high-speed waveform sampling, low power
digitization, fast digital data processing, serialization and data transmission. All integrated
circuits, necessary for physical instrumentation, should be radiation tolerant at an earlier not
reached level (hundreds of Mrad) of total ionizing dose and allow minute almost 3D assemblies.
The paper is based on literary source analysis and presents an overview of the state of the art
and trends in nowadays chip design, using partially own ASIC lab experience. That shows a
next stage of ising micro- and nanoelectronics in physical instrumentation.
1. Introduction
Integrated circuits are one of the key complex units available to designers of multichannel
detector setups. Complexity of custom or Application Specific Integrated Circuits (ASICs) for
particle physics experiments is strongly correlated with integration scale. According to Moore
Law, the number of transistors in a chip (integrated circuit) roughly doubles every two years.
As a result the scale gets smaller and transistor number increases at a regular pace to provide
improvements in integrated circuit functionality and performance while decrease costs [1]. Now
the interpretation of the law is not classical quantity of elements over space (area), but effective
usage of the fixed space for increasing functionality. Thus if we refer to scaling as the reduction
in cost (somehow) of the money/transistor on a chip, then surely several more generations of
Moores Law are ahead of us [2].
Acccording to forecast of the International technology roadmap for semiconductors (ITRS)
[3], the main acting factors for further chip development or system integration chalenges are:
design productivity, power management, manufacturability, bandwidth, cooling. All integrated
circuits, necessary for physical instrumentation, should be radiation tolerant at new earlier not
reached level (hundreds of Mrad) of total ionizing dose and allow minute almost 3D assemblies
[2, 4].
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
International Conference on Particle Physics and Astrophysics IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 798 (2017) 012204 doi:10.1088/1742-6596/798/1/012204
• Digital signal processing part more and more replaces the analog one. The 250 nm CMOS
process provides 10 Kgates/mm2 alowing to place in the area of 100*100 µm2 a few registers
only. Having shifted to the 28 nm process with density of 3900 Kgates/mm2 the designer
can place a whole microprocessor in the same area. Thus the trend pushes the designer in
the era of necessary minimum analog and analog-to-digital conversion design followed by
programming all other almost unlimited functionality in digital domain;
• Design cost increases comparing to the one of manufacture;
2
International Conference on Particle Physics and Astrophysics IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 798 (2017) 012204 doi:10.1088/1742-6596/798/1/012204
• Design timeline and number of prototype cycles (respins) are necessary to be reduced to
get more faster the final result;
• Technologies available today are well ahead of most requirements foreseen for the next
generation experiments [2];
• For the majority of projects the minimum set of ASICs is to be developed in a one well
verified technology (LHC – IBM 130 nm, SLHC – TSMC 130 and 65 nm, FAIR – UMC 180
nm). That is in fact a standardization trend. Technology choice is based on the necessity to
use processes more and more advanced (with life time beyond 2020), but well characterized
for design tools and having a reasonable cost;
• Extremely compressed timeline for each cycle (respin) of custom design (4-6 months);
• Start-to-finish (system level) design is necessary: Structural and behavioral modeling –
Design of IP blocks at transistor level and that of systems at high-level language – Layout
– Verification (ERC, DRC, LVS, PE, PS) – GDSII;
• Possibility of ASIC adaptation for other projects as criterion of economic efficiency, thus
providing a universalization trend;
• Functionality increase at a limited power consumption budget;
• Increase in demand for a high technology (expensive) product. The accent at design is
done at system (architectural) questions. ASICs as Systems On a Chip (SOCs) dictate a
mixed-signal character of design route. Today almost all projects are mixed-signal one. For
example, digital ASICs need an analog interface blocks, while analog chips are supplied by
digital slow control or power saving funtionality;
• The demanded number of ASIC wafers is small and can’t interest a manufacturer. For
example, if an experiment requires 106 channels, ASIC contains 100 channels per chip and
one can get 1000 chips from each wafer, that results in 10 wafers only;
• Design is based on the usage of constantly and fastly advanced tools at all levels of design:
Computer Aided Design (CAD) systems (Cadence, Mentor Graphics, Synopsys, Agilent
and other), technological libraries or Design Kits (elements, standard digital and IP blocks),
including design rules of manufacturer, the quantity of which quickly grows for advanced
processes (for example, more than 2000 for 22 nm node). As shown in figure 2 there is a
strong correlation between the number of design rules and layers for CMOS process. At
the same time figure 3 shows that due to technology trend for shift from 2D to 3D space
the complexity of element models grows quickly [8], demanding more and more powerful
computing;
• Necessity of a fast and actually continuous retraining of experts as well as a skilled staff of
young engineers (to support team competence level for experiment at least);
• Close integration of the design centers for both chip and system (hardware) levels.
3
International Conference on Particle Physics and Astrophysics IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 798 (2017) 012204 doi:10.1088/1742-6596/798/1/012204
Figure 2. The number of design rules and layers for CMOS processes.
4
International Conference on Particle Physics and Astrophysics IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 798 (2017) 012204 doi:10.1088/1742-6596/798/1/012204
For example, further presented is the practical dilemma. What is better: either a rad-hard
350 nm process (for example, silicon-on-isolator) or a bulk CMOS 130 (90) nm process?
The latter is better also in terms of economics.
• Traditional methods for radiation hardening are: 1) Circuital (reservation, redundancy,
coding, adaptive biasing and so on); 2) Layout ones (guard rings, ring transistors, and so
on); 3) Technological (low volume CMOS, SOI, trench isolation, and so on). The required
rad-hardness level is provided by a semiconductor structure choice and use of the special
technological processes.
• Rad-hard by design concept consists in the following:
(i) Refinement of traditional non- rad hard libraries for computer aided design (CAD);
(ii) Elaboration of additional design rules, improving rad-hardness at the following 3 levels
of design:
(a) Transistor level,
(b) Level of cells, gates and IP blocks,
(c) System level;
(iii) Creation of calibrated test structures and expansion of the scaled element libraries;
(iv) Monitoring and statistical analysis of the ASIC rad-hard stability, provided at
manufacture;
(v) Rad-hard tests of chips at both passive and active work modes.
5
International Conference on Particle Physics and Astrophysics IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 798 (2017) 012204 doi:10.1088/1742-6596/798/1/012204
6
International Conference on Particle Physics and Astrophysics IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 798 (2017) 012204 doi:10.1088/1742-6596/798/1/012204
7
International Conference on Particle Physics and Astrophysics IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 798 (2017) 012204 doi:10.1088/1742-6596/798/1/012204
7. Conclusions
The future of instrumentation is tightly bound with a wide usage of ASICs due to the new
challenge in integration scale. Commercial technologies available today are well ahead of
most requirements foreseen for particle physics‘ ASICs. The solution of system (architectural,
structural) problems will demand substantially higher efforts to create not only separate
components, but whole systems on chip (SoC). Almost all new chips have a mixed-signal
character.
Nowadays the design is based on microelectronic CAD tools of different suppliers and the
adapted Design Kits of the chosen manufacturer. Reduction of timelines for design, extension of
reusable IP block libraries, new level of system tests for readout electronics all of that requires
the usage of advanced CAD in distributed fast networks of powerful computing clusters.
New projects more and more require a tight collaborative work of expanding team and
accurate sharing all design infractructure (computing clusters, CAD tools etc). The necessity of
much stronger collaboration and coordination between all participants, sharing: the process of
preparing human resources; the cost of expensive CAD and chip prototyping as well as updating
existing design competences.
The ”rad-hard by design” approach for commercial CMOS technology is considered to be
the best way of providing an acceptable rad-hardness by circuital and layout techniques. For
radiation hardness, temperature stability and reliability of the chips it is important to refine the
existing libraries and design routes.
Long term planning becomes an extremely important factor, especially taking into account
the ”life time” of both the chosen technology and the design staff.
Acknowledgments
This work was supported by the Ministry of Education and Science of the Russian Federation in
the frames of the Competitiveness Programme of National Research Nuclear University MEPhI
and grant no. 14.A12.31.0002 in accordance with the RF government resolution no. 220.
References
[1] http://isscc.org/videos/2016 plenary.html Moores Law: A Path Forward
[2] Marchioro A 2015 Past and Future of Microelectronics in HEP Proc. of TWEPP-2015
[3] http://www.semiconductors.org/main/2015 international technology roadmap for semiconductors itrs/ In-
ternational technology roadmap for semiconductors 2.0: 2015 Executive report
[4] De Geronimo G et al 2013 Integrated Circuit Design in US High-Energy Physics arXiv:1307.3241 [physics.ins-
det]
[5] Weber M 2016 Highlights and Trends of Detector Instrumentation and Technology Development in Germany
Proc. of TWEPP-2016
[6] Roy K, Jung B and Raghunathan A 2010 Integrated systems in the more-than-Moore Era: Designing lowcost
energy efficient systems using heterogeneous components Proc. of 23rd International Conference on VLSI
Design
[7] Lacoe R 2008 Improving Integrated Circuit Performance Through the Application of Hardness-by-Design
Methodology IEEE Trans. on NS 55 (4) 1903
[8] http://www.cadence.com/mail/can/apr 2016.html Cadence EMEA 2016 Science-Fiction becoming Reality