Static Timing Analysis Basics by Selva Kumar
-
Upload
gaurav-gopalkrishna -
Category
Documents
-
view
109 -
download
1
description
Transcript of Static Timing Analysis Basics by Selva Kumar
Selvakumar @ [email protected] 2
Overview
� Even though a digital circuit may be logically correct, one needs to know how it will perform on physical implementation
� Timing analysis is required • To meet a performance specification
• To evaluate how the design operates
� Timing Analysis : Point-point analysis of design without stimulus vectors
Selvakumar @ [email protected] 3
Session Topics
� Timing Paths
� Clock Skew
� Timing Constraints
� Slack
Selvakumar @ [email protected] 8
Timing Analysis� A ‘N’ bit input design to be verified completely requires
a 2N test patterns
� The approach of generating 2N patterns and applying them at different time & sequence is impractical
� The complexity of analysis of netlist is reduced by breaking the process into two steps
• One of functional verification
• Other of timing Verification
� The process of verifying the netlist through test benches and test patterns after P&R is known as Dynamic Timing Analysis
• Requires input from user / verification engineer
Selvakumar @ [email protected] 9
Dynamic Timing Analysis
� Dynamic timing analysis / simulation guarantees 100% coverage
� Though this guarantee comes at a cost of time
• Time of generation of patterns
• Time of application of patterns etc
� It is design specific
� One method cannot be used to all designs
� Hence for each design a unique test bench needs to be created
� A generalized approach is required so that verification process is speeded up
Selvakumar @ [email protected] 18
•Static Timing Analysis is a method of calculating the
expected timing of a digital circuit without requiring
simulation.
•STA is method of analyzing and validating the timing
performance of a design.
STA – What is Static Timing Analysis?
�Advantages:
�Much faster than gate-level simulation.
�Exhaustive
�Proper circuit functionality is not checked.
�Vector generation NOT required.
Selvakumar @ [email protected] 19
What is Timing Analysis?
Setup Requirement
Hold Requirement
�What are our circuit timing requirements?
Clk
100 200 300 400 5000
Data
Data Cannot
Change Within
These Windows
Data
Clk
D Q
QBOutput
OutputBar
Selvakumar @ [email protected] 20
Timing Terminologies
� Setup Time : Time for which data should be stable at the input of the ff before the arrival of clock (active edge) at the ff’s clock pin
� Hold Time : Time for which data should be stable at the input of the ff after the arrival of clock (active edge) at the ff’s clock pin
TSETUP
THOLD
Selvakumar @ [email protected] 21
Metastability
� For any flop to work its setup time and hold time requirements should be met
Selvakumar @ [email protected] 27
STA Process
� STA happens in three steps
� Circuit is broken down to sets of timing paths
� Delay of each path is calculated
� Path delays are checked to see if timing constraints have been met
Selvakumar @ [email protected] 28
Timing Paths
� ‘Timing’ refers to the time taken by data to travel from one point to the other
� The point at which data is expected to originate is considered a ‘start point’• Input Port
• Clock Pin of sequential elements
� The location at which data terminates is considered as the stop point• Output Port
• Data Pin of sequential elements
Selvakumar @ [email protected] 29
Three Steps in Timing Analysis
• Circuit is broken down into sets of timing paths.
• Delay of each path is calculated.
• Path delays are checked to see if timing constraints
have been met.
Selvakumar @ [email protected] 30
What is a Timing Path?
•A Timing Path is a point-to-point path in a design which
can propagate data from one flip-flop to another.
•Each path has a startpoint and an endpoint
•Startpoints:
•Input ports, Clock pins of flip-flops
•Endpoints:
•Output ports, Data input pins of flip-flops
Selvakumar @ [email protected] 31
Timing Terminologies
� Critical Path : Theoretically path which has maximum delay
� Arrival Time : Time taken by data to reach a end point from a specific start point. Depends on complexity of logic through which data traverses
� Required Time : Time at which data is required at a particular end point. Depends on the requirements / specifications
� Slack : Difference in required time and arrival time. For a design to work the slack value should always be positive
Selvakumar @ [email protected] 32
Timing Paths
� Depending on the logic through which data propogates a design can be considered to comprise of 4 paths
• input –> reg : Data from Input port to the first flip-flop
• reg –> reg : Data from one flip-flop to another ff
• reg –> out : Data from last flip-flop to output port
• input –> output : Data from Input – Output with no
sequential components in between
Selvakumar @ [email protected] 34
Timing paths
� Input -> Reg
� Reg -> Reg
� Reg -> Output
� Input -> Output
• Input / Output also called as ‘Pad’ / ‘Pin’
• ‘Reg’ can be any of - flop, latch, RAM etc
• ‘Reg’ also called as clock / setup
Selvakumar @ [email protected] 35
Input –> Reg
� Path which starts at any input port and ends at the data pin of any sequential element that is first encountered
� Is controlled by the capture clock
� Data may start independently but must reach the sequential element before clock reaches the element
Selvakumar @ [email protected] 36
Reg –> Reg
� Path which starts at the clock pin of a ff or enable pin of latch and ends at the data pin of any sequential element that is next encountered
Selvakumar @ [email protected] 37
Reg –> Reg
� Path which starts at the clock pin of a ff or enable pin of latch and ends at the data pin of any sequential element that is next encountered
� Is controlled by both the launch as well as capture clock
� Data leaves the first sequential element on the clock and then races forward to reach the next sequential element before clock / enable reaches the element
Selvakumar @ [email protected] 38
Reg –> Out
� Path which starts at the clock pin of a ff or enable pin of latch and ends at the output port of the design
� Clock launches the data after which data runs uncontrolled
Selvakumar @ [email protected] 39
Input –> Output
� Path which starts at any input port and ends at any output port without traversing through any sequential element
� Clock has no role to play in this path
� Is seen mainly in purely combinational circuits practically rare in designs
Selvakumar @ [email protected] 40
Timing Paths Example
� How many start points are there in the circuit
� How many end points are there in the circuit
� How many paths are there in the circuit
Selvakumar @ [email protected] 41
Anatomy Of A Sequential Path
Combinational
Logic
Tclock-to-Q Tlogic Tsetup
Flip
Flop
φ
Flip
Flop
φ
One clock cycle
setupcomboQtoclockcycle TTTT ++= −−
Selvakumar @ [email protected] 43
Propagation Delay
� Delay of a cell depends on many factors such as
• Operating Conditions
• Functionality of the cell
• Type of Inputs
� Based on the factors above a cell can propagate data either quickly or slowly
� Depending on the kind of analysis being done at any point of time one needs to consider either minimum delays or maximum delays
Selvakumar @ [email protected] 44
Real world effects
� All parameters vary based on PVT which stands for Process, Voltage, Temperature.
• Higher voltage usually speeds things up.
• Higher temperature usually slows things down.
• Process variation is more difficult to quantify.
� Most datasheets specify maximum delays.
Selvakumar @ [email protected] 46
Delay Types
� Assume the ‘and’ gate shown below has a maximum delay of 3 ns at max operating conditions and for a 0-1 transition
� Let the ‘and’ gate have a delay of 0.6 ns at minimum operating conditions and for a 1-0 transition
� Which is the delay to be considered for analysis?
Selvakumar @ [email protected] 47
Clock Skew
� Clock feeds multimillion flip-flops
� Theoretically clock should arrive at same instance at all flip-flops – Practically impossible
Selvakumar @ [email protected] 48
Clock Skew
• Clock Skew: The maximum difference in arrival time of the
clock signal to each register in the design
clock
Clock arrivaltime at 1.1ns
Clock arrival time at 1.3ns
Skew = 1.3ns - 1.1ns = .2ns
•It is also defined as the difference in time that a single clock signal
takes to reach two different registers.
Selvakumar @ [email protected] 49
Clock Skew
� Variation in arrival of clock at clock pin of subsequent / consecutive flip-flops is known as skew
Selvakumar @ [email protected] 50
Clock Jitter
� The difference in arrival of clock at different flip-flops could be due to
• Jitter – cycle to cycle variation in clock period due to aging of oscillator, anomalies in the plletc
Selvakumar @ [email protected] 52
Clock Skew
� Skew can be considered to be of two types
• When clock arrives earlier than expected –generally known as negative skew
• When clock arrives later than expected –generally known as positive skew
� +ve skew can occur when data & clock travel in same direction
� -ve skew can occur when data & clock travel in opposite directions
Selvakumar @ [email protected] 53
Positive & Negative Skew
R1In
(a) Positive skew
CombinationalLogic
D Q
tCLK1CLK
delay
tCLK2
R2
D QCombinational
Logic
tCLK3
R3
• • •D Q
delay
R1In
(b) Negative skew
CombinationalLogic
D Q
tCLK1
delay
tCLK2
R2
D QCombinational
Logic
tCLK3
R3
• • •D Q
delay CLK
Selvakumar @ [email protected] 54
Positive Skew
Launching edge arrives before the receiving edge
+ve skewTclk + δ >= tcq1 + tcombo + t su2
(th2 + δ) < tcq1 + tcombo
CLK1
CLK2
T CLK
δ
T CLK + δ
+ t h δ
2
1
4
3
Selvakumar @ [email protected] 55
Negative Skew
Receiving edge arrives before the launching edge
-ve skewTclk - δ >= tcq1 + tcombo + t su2
(th2 - δ) < tcq1 + tcombo
CLK1
CLK2
T CLK
δ
T CLK - δ
2
1
4
3
Selvakumar @ [email protected] 56
Timing Terminologies
� Critical Path : Theoretically path which has maximum delay
� Arrival Time : Time taken by data to reach a particular end point from a specific start point. Depends on complexity of logic through which data traverses
� Required Time : Time at which data is required at a particular end point. Depends on the requirements / specifications
� Slack : Difference in required time and arrival time. For a design to work the slack value should always be positive
Selvakumar @ [email protected] 57
Maximum Frequency
� For a circuit to meet its specifications the minimum time period or maximum frequency at which it works should be identified
� The frequency of operation of a circuit depends on the logic of the paths
� Each of the four kinds of paths have different required times
Selvakumar @ [email protected] 58
Frequency Calculations
� Arrival Time = TCOMBO, Max
� Required Time = TCP + TClock_Delay - TSETUP
Selvakumar @ [email protected] 59
Frequency Calculations
� Arrival Time : TClk-Q + TCombo, MAX
� Required Time = TCP + TClock_Skew - TSETUP
Selvakumar @ [email protected] 60
Frequency Calculations
� Arrival Time : TCombo, MAX
� Required Time : Explicit Timing Constraint
(If Specified)
Selvakumar @ [email protected] 61
Frequency Calculations
� Arrival Time = TClk-Q + TCombo, MAX
� Required Time = Explicit Timing Constraint
(If Specified)
Selvakumar @ [email protected] 62
Problem 1
� Find The maximum Frequency for the circuit shown in figure below
Selvakumar @ [email protected] 63
Solution Methodology
� Step One : Find the total number of paths in the design
� Step Two : Identify the paths that are constrained by clock
� Step Three : Identify the arrival times of all the paths identified in step 2 above
� Step Four : Identify the required time equations of all paths identified in step 2
� Step Five : Equate the arrival times and required times of all relevant paths to obtain the time periods
� Step Six : Select the maximum applicable clock frequency from the clocks identified in step 5
Selvakumar @ [email protected] 64
Solution
� Three paths in the design• Input – Reg
• Reg – Reg
• Reg – Out
� Reg – Out path not constrained hence can be discarded for timing calculation
� Relevant Paths are Input – Reg and Reg – Reg
� Input – Reg Path• Arrival Time = 3
• Required Time = CP + 0.5 + 0.6 -Tsetup= CP + 1.1 -0.65
• Equating the 2, CP = 3 – 1.1 + 0.65 = 2.55
Selvakumar @ [email protected] 65
Solution
� Reg – Reg Path• Arrival Time = 0.5 + 0.6 + 1.4 + 4 = 6.5
• Required Time = 0.5 + 0.7 + 0.8 + CP – 0.75 = CP + 1.25
• Equating the two, CP = 6.5 – 1.25 = 5.25
� Two time periods available from the calculations 5.25, 2.55. Though first path can work if clock time period is 5.25 the second path can not work if clock time period is 2.55
� Hence the maximum clock frequency at which the circuit can work is 1/5.25
Selvakumar @ [email protected] 66
Problem 2
Paths from Q1 to Q1:
Paths from Q1 to Q2:
Paths from Q2 to Q1:
Paths from Q2 to Q2:
None
TW ≥ max tPDFF +tJKsu = 20 +10 = 30 ns
TW ≥ max tPDFF + max tAND + tJKsu = 20 + 12 + 10 = 42 ns
TW ≥ max tPJKFF + tOR + TDsu = 25 + 10 + 5 = 40 ns
TW ≥ max tPJKFF + max tAND + tJKsu = 25 + 12 + 10 = 47 ns
TW ≥ 47 ns
Selvakumar @ [email protected] 67
Problem 3
� Find the maximum applicable clock frequency for the circuit shown in figure below
Selvakumar @ [email protected] 68
Problem 4
� For the circuit shown in figure below find the maximum applicable clock frequency
Selvakumar @ [email protected] 69
Problem 5
� For the circuit shown in figure below find the maximum applicable clock frequency
Selvakumar @ [email protected] 70
Failure / Data loss Due To Large Skew
Combinational
LogicFlip
Flop
Flip
Flop
δ delayclk
“B”
AoutAin
“A”
Bin
� If new data (Ain) gets to point “B” before clock does, system will fail by simply skipping over old data…
� For this illustration - ignore tsetup
Selvakumar @ [email protected] 71
Combinational
LogicFlip
Flop
Flip
Flop
δ delayclk
“B”
AoutAin
“A”
Bin
Clock arrives at point “A”
T = 0ns
Selvakumar @ [email protected] 72
Data arrives at comb logic input
Combinational
LogicFlip
Flop
Flip
Flop
δ delayclk
“B”
Ain
“A”
BinAout
T = tclk-to-Q
Selvakumar @ [email protected] 73
Data Exits Comb Logic
Combinational
LogicFlip
Flop
Flip
Flop
δ delayclk
“B”
Ain
“A”
Aoutnew
Bin
Bin
T = tclk-to-Q + tlogic
Selvakumar @ [email protected] 74
Clock Reaches “B”
Combinational
LogicFlip
Flop
Flip
Flop
clk
“B”
Ain
“A”
Aoutnew
Bin
Bin
δ delay
T = tclk-to-Q + tlogic T = tδ
Selvakumar @ [email protected] 75
Failure!!!
Combinational
LogicFlip
Flop
Flip
Flop
clk
“B”
Ain
“A”
Aoutnew
Bin
Bin
δ delay
New
Bin
What happened to old Bin???
If tclk-to-Q+tlogic < tδ it fails…
Selvakumar @ [email protected] 77
Example 1D Q
QCK
Q
TW ≥ max tPFF + tsu
� For the 7474, max tPLH = 25ns, max tPHL = 40ns, tsu = 20ns
TW ≥ max (max tPLH + tsu, max tPHL + tsu)
TW ≥ max (25+20, 40+20) = 60
Selvakumar @ [email protected] 79
Example 3
D Q
Q
D Q
Q
MUX
0
1
Q0 Q1
CK
TW ≥ max tPFF + max tPMUX + tsu
Selvakumar @ [email protected] 86
False paths• Paths that physically exist in a design but are not
logic/functional paths
• These paths never get sensitized under any input conditions
Mux 1
C C1 C2A
B
Mux 2
S
B1 B2
OUT
Total 4 timing paths
PATH 1 – A-C-C1-C2-OUT
PATH 2 – A-C-OUT
PATH 3 – B-B1-B2-C-C1-C2-OUT
PATH 4 – B-B1-B2-C-OUT
Only path1 and 4 above are valid logic paths as select line for the 2 muxes are the same
Selvakumar @ [email protected] 87
Multicycle Path
�Multicycle paths are paths which intentionally require more than one clock cycle to propagate.
�This information cannot possibly be inferred by the timing tool, so it must be specified by the designer so the analyzer can mark the path and correctly compute the timing.
�A start point, end point and/or "through" point is specified, along with the number of allowed clock
cycles.