Accuracy Considerations in RC Extraction for STA Aya Keller, Igor - - PowerPoint PPT Presentation

accuracy considerations in rc extraction for sta
SMART_READER_LITE
LIVE PREVIEW

Accuracy Considerations in RC Extraction for STA Aya Keller, Igor - - PowerPoint PPT Presentation

Accuracy Considerations in RC Extraction for STA Aya Keller, Igor Keller TAU 2019 Monterey March, 19 1 Motivation Design trends: Finfet era: faster cells, higher logic density, clocks scale down fast Size grew significantly: 10


slide-1
SLIDE 1

1

Accuracy Considerations in RC Extraction for STA

Aya Keller, Igor Keller TAU 2019 Monterey March, 19

slide-2
SLIDE 2

2

Motivation

  • Design trends:
  • Finfet era: faster cells, higher logic density, clocks scale down

fast

  • Size grew significantly: 109-instance designs are here!
  • Higher process variations: many process corners, often >100
  • Requirements for modern STA:
  • High capacity and performance
  • Required accuracy: path delay within 2% of Spice
  • Waveform Aware Simulation-based delay calculation is a

MUST

2

Vi

V0

current source models, waveforms

Cx v v t BI Ax x E    ) , , (  

Fast and accurate ODE solvers efficient RC Reduction

slide-3
SLIDE 3

3

Motivation, cont.

  • Shrinking geometries lead to higher resistance and

capacitance of wires

  • Good old times of Ceff+TLU delay calculation are over!
  • Wire delays scale down slower than cell delays

3 IMAC’2017

20nm 7nm

higher R, C

slide-4
SLIDE 4

4

Motivation, cont.

  • Many sources of inaccuracy in STA
  • Is Spice an ultimate reference for delay?
  • Not really, uses many assumptions
  • Wire has a continuous though non-uniform

resistance and capacitance

  • RC extraction replaces it with lumped R’s and C’s
  • Lumping: spatial discretization
  • Effects of spatial discretization on STA accuracy is

not well studied

  • Goal of this work

1-2% 1-2% 5-10% ~5-10% ~5-10% Delay Calculation 3-8% ~3-4% IR drop Process Variations Crosstalk RC Extraction Library Characterization Model extraction from Silicon

+

1-5% Spice Reference Error >25% Total Error vs Silicon

Focus RC Extraction => Spatial discretization

slide-5
SLIDE 5

5

Problem Formulation

  • Uniform wire, far-end cap
  • Driver: step voltage source
  • Target: far-end voltage response
  • Split into N equal segments
  • Approximate each with Π-model
  • Build N-link RC chain
  • Goal: accuracy of far-end

response vs N

5

C/N, R/N

N segments

R/N

C/2N C/2N

Lump C, R

kC

~

u(t)

kC

~

u(t)

C/N C/N C/N C/2N

R/N R/N R/N R/N

slide-6
SLIDE 6

6

Discretized Problem, Numerical Solution

6

6

kC

~

u(t)

N N-1 2 1

C/N C/N C/N C/2N

R/N R/N R/N R/N

[time]=R*C

slide-7
SLIDE 7

7

Continuous Problem, Analytical Solution

7

C, R

~

u(t) x 1 0<x<1: t=0:

boundary conditions

Solution:

[time]=R*C

slide-8
SLIDE 8

8

Errors of Temporal Discretization

  • Numerical solution of ODEs has errors due to temporal discretization
  • Needs to be much smaller than spatial discretization
  • Errors depends on the method
  • Backward/Forward Euler: O(h)
  • Trapezoidal: O(h2)

8 Forward Euler Backward Euler Trapezoidal

analytical solution

slide-9
SLIDE 9

9

Accuracy of Wire Delay and Slew (k=0)

9

slide-10
SLIDE 10

10

Voltage Responses at Far End

10

slide-11
SLIDE 11

11

Results

11

Delay, slew and relative errors for different far-end load and N Time unit = R*C

slide-12
SLIDE 12

12

Conclusions, Future Work

  • Spatial discretization in RC extraction may cause large errors
  • We quantified errors of wire delay and slew for several far-end loads
  • Used analytical solution to ensure time discretization has negligible error
  • We demonstrated that number of segments needs to be larger than 2 for wire

delay error < 2%

  • Is important for wire-delay dominated paths
  • Future work:
  • include crosstalk
  • more complex wire topologies
  • non-uniform wire widths and spacing

12