In this paper, we describe the design of Neurogrid, a neuromorphic system for simulating large-scale neural models in real time. Neuromorphic systems realize the function of biological neural systems by emulating their structure. Designers of such systems face three major design choices: 1) whether to emulate the four neural elements-axonal arbor, synapse, dendritic tree, and soma-with dedicated or shared electronic circuits; 2) whether to implement these electronic circuits in an analog or digital manner; and 3) whether to interconnect arrays of these silicon neurons with a mesh or a tree network. The choices we made were: 1) we emulated all neural elements except the soma with shared electronic circuits; this choice maximized the number of synaptic connections; 2) we realized all electronic circuits except those for axonal arbors in an analog manner; this choice maximized energy efficiency; and 3) we interconnected neural arrays in a tree network; this choice maximized throughput. These three choices made it possible to simulate a million neurons with billions of synaptic connections in real time-for the first time-using 16 Neurocores integrated on a board that consumes three watts.

/pdf/neurogrid-a-mixed-analog-digital-multichip-system-for-large-494ipl0j1g.pdf

Neurogrid: A Mixed-Analog-Digital Multichip System for Large-Scale Neural Simulations

Random device mismatch plays an important role in the design of accurate analog circuits. Models for the matching of MOS and bipolar devices from open literature show that matching improves with increasing device area. As a result, accuracy requirements impose a minimal device area and this paper explores the impact of this constraint on the performance of general analog circuits. It results in a fixed bandwidth-accuracy-power tradeoff which is set by technology constants. This tradeoff is independent of bias point for bipolar circuits whereas for MOS circuits some bias point optimizations are possible. The performance limitations imposed by matching are compared to the limits imposed by thermal noise. For MOS circuits the power constraints due to matching are several orders of magnitude higher than for thermal noise. For the bipolar case the constraints due to noise and matching are of comparable order of magnitude. The impact of technology scaling on the conclusions of this work are briefly explored.

Device mismatch and tradeoffs in the design of analog circuits

Modern and future ultra-deep-submicron (UDSM) technologies introduce several new problems in analog design. Nonlinear output conductance in combination with reduced voltage gain pose limits in linearity of (feedback) circuits. Gate-leakage mismatch exceeds conventional matching tolerances. Increasing area does not improve matching any more, except if higher power consumption is accepted or if active cancellation techniques are used. Another issue is the drop in supply voltages. Operating critical parts at higher supply voltages by exploiting combinations of thin- and thick-oxide transistors can solve this problem. Composite transistors are presented to solve this problem in a practical way. Practical rules of thumb based on measurements are derived for the above phenomena.

/pdf/analog-circuits-in-ultra-deep-submicron-cmos-1g96qlvoip.pdf

Analog circuits in ultra-deep-submicron CMOS

This paper summarizes recent trends in the area of low-power A/D conversion. Survey data collected over the past eleven years indicates that the power efficiency of ADCs has improved on average by a factor of two every two years. A closer inspection on the impact of technology scaling is presented to explain the observed trend in the context of shrinking supply voltages and increasing device speed. Finally, a discussion on minimalistic and digitally assisted design approaches is used to sketch a route toward further improvements in ADC power efficiency and performance.

A/D converter trends: Power dissipation, scaling and digitally assisted architectures

As complementary metal-oxide-semiconductor (CMOS) technologies are scaled down into the nanometer range, a number of major nonidealities must be addressed and overcome to achieve a successful analog and physical circuit design. The nature of these nonidealities has been well reported in the technical literature. They include hot carrier injection and time-dependent dielectric breakdown effects limiting supply voltage, stress and lithographic effects limiting matching accuracy, electromigration effects limiting conductor lifetime, leakage and mobility effects limiting device performance, and chip power dissipation limits driving individual circuits to be more energy-efficient. The lack of analog design and simulation tools available to address these problems has become the focus of a significant effort with the electronic design automation industry. Postlayout simulation tools are not useful during the design phase, while technology computer-aided design physical simulation tools are slow and not in common use by analog circuit designers. In the nanoscale era of analog CMOS design, an understanding of the physical factors affecting circuit reliability and performance, as well as methods of mitigating or overcoming them, is becoming increasingly important. The first part of the paper presents factors affecting device matching, including those relating to single devices as well as local and long-distance matching effects. Several reliability effects are discussed, including physical design limitations projected for future downscaling. In some cases, it may be helpful to exceed foundry-specified drain-source voltage limits by a few hundred millivolts. Models are presented for achieving this, which include the dependence on the shape of the output waveform. The condition Vsb > 0 is required for cascode circuit configurations. The role of other terminal voltages is discussed, as Vsb > 0 increases both hot and cold carrier damage effects in highly scaled devices. The second part of the paper focuses on trends in device characteristics and how they influence the design of nanoscale analog CMOS circuits. A number of circuit design techniques employed to address the major nonidealities of nanoscale CMOS technologies are discussed. Examples include techniques for establishing on-chip accurate and temperature-insensitive bias currents, digital calibration of analog circuits, and the design of regulator and high-voltage circuits. Achieving high energy efficiency in ICs capable of accommodating 109 devices is becoming critically important. This paper also presents a survey of the evolution of figure of merit for analog-to-digital converters.

Analog Circuit Design in Nanoscale CMOS Technologies

Due to device and voltage scaling scenarios for present and future deep-submicron CMOS technologies, it is inevitable that the off-state current (I/sub off/) of MOSFET transistors increases as the technology minimum dimensions scale down. Experimental evidence shows that the leakage current distribution of modern deep-submicron designs not only has a higher mean value but it also presents a larger variability as well. In this paper, we investigate the impact of threshold voltage mismatch as one plausible source for this increased variability. In digital circuit design, it is commonly assumed that the threshold voltage difference (mismatch) of static CMOS cells is negligible. However, threshold voltage mismatch (/spl Delta/V/sub to/) has a two-sided effect on the off-state current. Namely, the total cell's current can increase or decrease depending upon the direction of the V/sub t/ mismatch shift. This effect can be so severe that I/sub off/ can increase by more than one order of magnitude with respect to its nominal value due only to V/sub to/ mismatch. We further show through experimental results that the V/sub to/ mismatch of paired transistors working in the subthreshold regime can be worse by a factor of two as compared to transistors working in the saturation or linear regions. A factor of two larger spread is obviously quite devastating in terms of area, speed, and power consumption, should it be desired to attain the same I/sub off/ level as for a V/sub to/ mismatch characterized out of the subthreshold regime.

/pdf/threshold-voltage-mismatch-and-intra-die-leakage-current-in-2hi0m9u07h.pdf

Threshold voltage mismatch and intra-die leakage current in digital CMOS circuits

CMOS evolution introduces several problems in analog design. Gate-leakage mismatch exceeds matching tolerances requiring active cancellation techniques. One strategy to deal with the use of lower supply voltages is to operate critical parts at higher supply voltages, by exploiting combinations of thin and thick-oxide transistors.

/pdf/designing-outside-rail-constraints-2y02bac9sp.pdf

H. Tuinhout

Papers

Analog circuits in ultra-deep-submicron CMOS

Threshold voltage mismatch and intra-die leakage current in digital CMOS circuits

Designing outside rail constraints

Intra-die leakage current behavior of digital CMOS circuits: a threshold voltage mismatch perspective,