Professional Documents
Culture Documents
AbstractA simple metric is presented for the accurate pre- cell library (i.e. the number of stacked transistors and cell
diction of path delay variability during the automated synthesis strength).
of digital VLSI circuits. This allows circuit variability to be Interestingly, the proposed metric permits statistical eval-
assessed at early stages within the design process with minimal
computational effort, as extensive Monte Carlo or SSTA runs are uation of the path delay by simply resorting to standard
not required. This paper introduces the metric and investigates (deterministic) CAD tools. In other words, no statistical timing
its effectiveness. The final predictions of path delay variability analyser is needed. This also permits a dramatic reduction of
are found to be within 10% of measured path delay variability, the computational effort involved in SSTA in the early digital
with an average error of 3%, for a series of test paths synthesised design phases, and in the preliminary design space exploration
from randomised models of a 130nm technology library. These
randomised models are generated from a 3D atomistic simula- and comparison of different design options. Using the pro-
tor and provide more accuracy than traditional Monte Carlo posed approach, SSTA can be performed only to validate the
simulation runs. final design version (sign-off) with a greater accuracy (and
computational effort).
I. I NTRODUCTION
The paper is structured as follows. Section II introduces the
The accuracy of timing validation tools and methods is metric, describing the assumptions made during the derivation.
being challenged by the scaling of fabrication process dimen- Section III contains an assessment of the effectiveness and
sions down to the nanometre level. Designers must be able to limitations of the metric. A summary and conclusions is
alter their designs to accommodate random physical intradie provided in section IV.
variations, but this relies on the ability to predict their effects
within design tools, something which is not yet fully available. II. D ESCRIPTION OF THE PROPOSED METRIC
There has been much work within the literature to im-
Delay variations due to process variability of a given logic
prove statistical methods for the modelling of random intradie
gate depend on
variations; including correlations between delays [1]; adding
signal transition time and power supply variations [2] [3]; 1) Gate topology (i.e., logic function - number of stacked
incorporating non-Gaussian distributions [4] and a metric of transistors)
sensitivity [5]. These incarnations of SSTA have increased 2) Transistor size
computational complexity over STA algorithms and have not 3) Input rise/fall time
yet been adopted by the industry. 4) Output parasitic capacitance of the gate
This paper presents the derivation and evaluation of a 5) Load capacitance (interconnect + input capacitance of
simple but effective metric for the prediction of path delay the subsequent gate)
variability in CMOS logic circuits. This metric allows for The following observations can be made about these vari-
a rapid comparison of path variability, without the need for ability sources in order to simplify the process of modelling the
extensive SSTA or Monte Carlo simulations. The variability intradie variability (i.e. variations that have a different impact
of a given path is defined as path
path
where path and path are on different transistors within the same die [6]):
the mean and standard deviation of the path delay respectively. 1) Topology affects the current Igate delivered by the gate
The proposed metric permits the evaluation of the path to the output capacitance. In [6], it was demonstrated
delay variability in any synthesis methodology (i.e., based on that the variations in the on current delivered by a
wire load models or physically based), and strongly simplifies generic logic gate Igate (and hence the resulting delay
the path timing analysis process for modelling the effects of variations) are proportional to the reciprocal of the
intradie variations on delays. The information required for square root of the number of stacked transistors nstack :
predicting path delay variability is minimal, as it depends on
very basic parameters of the cell involved in the synthesis, delay 1
(1)
which are known directly from the knowledge of the adopted delay nstack
2499
(a) INVERTERS
Fig. 3. A plot of measured path delay variability ( path ) against predicted
path
(d) NAND & NOR Gates path delay variability (Equation 5) for test path Figure 1(a). k has been initially
set to 1
Fig. 1. Test Circuits
Fig. 4. A plot of measured path delay variability ( path ) against predicted
path
path delay variability (Equation 5) for the test paths in Figure 1. k is set
Fig. 2. Testbench used for SPICE simulations to the calculated value of 0.0567
2500
logic.
IV. C ONCLUSIONS AND F UTURE W ORK
A simple metric has been developed which accurately
predicts the variability of delays through circuits comprising
of simple CMOS logic. Equation 5 allows for the variability of
different paths to be compared without the need for extensive
Monte Carlo or SSTA analysis, allowing circuit variability to
be assessed at early stages within the design process with
minimal computational effort. The predictions are based on
Gaussian distributions of gate delays, and have been used to
Fig. 5. A plot of measured path delay variability ( path ) against predicted
path
accurately predict the variability of paths consisting of non-
path delay variability (Equation 5) for the test paths in Figures 1. The values Gaussian gate delays for logic depths greater than two cells.
of nstack have been changed to 1.2 for cells with two transistors in series The randomspice based modelling of random physical vari-
ations within transistors used within this work provides a
more accurate reference than results based on typical Monte
Carlo simulations of commercial design kits (where a few
device parameters are varied randomly according to a Gaussian
probability distribution).
The proposed metric therefore provides an accurate and
rapid statistical evaluation (always within 10% and usually
within 3%) of designs at the first stages of synthesis for circuits
consisting of simple CMOS gates. This permits to avoid
performing Monte Carlo simulations during design iterations,
thereby dramatically reducing the computational effort. Hence,
Monte Carlo simulations can be performed only at the signoff
stage to provide greater accuracy when required.
Fig. 6. A plot of measured path delay variability ( path ) against predicted
path
path delay variability (Equation 5) for the test path in Figures 1(b), where the Future work will investigate the limitations of the metric
number of inputs to the NAND gates are varied when predicting the variability of circuits containing complex
cell structures, and will verify its effectiveness for other
technology nodes.
value of nstack are given in Figure 5. R EFERENCES
The accuracy of the metric increases with a greater depth [1] S. Tsukiyama, M. Tanaka, and M. Fukui, A statistical static timing
of logic, and the differences between the predicted values and analysis considering correlations between delays, in Design Automation
measured values have a maximum error of 10% and an average Conference, 2001. Proceedings of the ASP-DAC 2001. Asia and South
Pacific, 2001, pp. 353358.
error of 3% for N > 2. The increase in accuracy with logic [2] T. Kouno and H. Onodera, Consideration of Transition-Time variability
depth may be due to the assumptions within the metric of the in statistical timing analysis, 2006, pp. 207210.
Gaussian nature of the delays, as the randomspice models are [3] D. Kouroussis, R. Ahmadi, and F. Najm, Voltage-Aware static timing
analysis, Computer-Aided Design of Integrated Circuits and Systems,
not based on Gaussian distributions of transistor parameters. IEEE Transactions on, vol. 25, no. 10, pp. 21562169, 2006.
The accuracy improves because the delays through the path [4] V. Khandelwal and A. Srivastava, A general framework for accurate sta-
become increasingly Gaussian in nature as the logic depth tistical timing analysis considering correlations, in Design Automation
Conference, 2005. Proceedings. 42nd, 2005, pp. 8994.
increases, as predicted by the Central Limit Theorem. [5] S. V. Kumar, C. V. Kashyap, and S. S. Sapatnekar, A framework
The corrected value of nstack was verified by repeating for block-based timing sensitivity analysis, in Design Automation
the analysis of test path Figure 1(b) with three, four and five Conference, 2008. DAC 2008. 45th ACM/IEEE, 2008, pp. 688693.
[6] M. Alioto, G. Palumbo, and M. Pennisi, Understanding the effect of
input NAND gates replacing the previously analysed two input process variations on the delay of static and domino logic, Very Large
NAND gates. An increase of 0.2 was applied to nstack for Scale Integration (VLSI) Systems, IEEE Transactions on, vol. PP, no. 99,
each additional input, and the predicted variability was found p. 1, 2009.
[7] M. Orshansky, S. Nassif, and D. Boning, Design for Manufacturability
to remain within 8%, again with an average error of 3%. Figure and Statistical Design. Springer Verlag, 2007.
6 provides a plot of the path delay variability predictions for [8] A. Asenov, Meeting the design challenges of nano-CMOS
multiple input NAND gates, where nstack was set to 1.2, electronics, http://www.nanocmos.ac.uk/, 2008. [Online]. Available:
http://www.nanocmos.ac.uk/
1.4, 1.6 and 1.8 for the two, three, four and five input gates [9] X. Wang, B. Cheng, S. Roy, and A. Asenov, Simulation of strain
respectively. enhanced variability in nMOSFETs, in Ultimate Integration of Silicon,
At this level of accuracy the proposed metric provides a 2008. ULIS 2008. 9th International Conference on, 2008, pp. 8992.
[10] T. Drysdale, A. Brown, G. Roy, S. Roy, and A. Asenov,
simple and efficient method of predicting the variability of Capacitance variability of short range interconnects, Journal
circuits synthesised from simple CMOS logic, especially if of Computational Electronics, Dec. 2007. [Online]. Available:
the circuit is constructed from simple NAND or NOR based http://www.nanocmos.ac.uk/publications/capacitancevariability.pdf
2501