Sources of Noise:
(3-6 are esp. important for fast systems)
Additionally, manufacturing between devices can introduced fixed noise (e.g. offsets)
'1'
'0'
Not all signals arise from a clock domain at all, such as physical, real-world inputs
A problem arises at a fanout point, where mutiple interpretations are performed in parallel
Consider that an inverter represents contrain on allowed input, output (V1,V2) relationships:
Consider another inverter with the input and output connections reversed:
Consider a coupled set of inverters, setting constraints between V1 and V2
Two constraints, A and B
Three Points are possible solutions to both constraints
1 is a quasi stable solution
Time analysis:
Consider ideal voltage sources and switches set two intial voltages
Switches are opened at time t=0
V1(0)=0
V2(0)=VDD
V1→0⇒A⇒V2→VDD⇒B⇒V1→0
V1(0)=VDD/2
V2(0)=VDD/2
V1→VDD/2⇒A⇒V2→VDD/2⇒B⇒V1→VDD/2
V1(0)=43VDD
V2(0)=41VDD
V1→43VDD⇒A⇒V2→0⇒B⇒V1→VDD
V1 is pushed to VDD, system stabalizes
@ (V1 = VDD, V2=0)
In theory, the middle point is also stable
We have assumed a symmetric transfer function and constraints
However, any noise will disrupt the balance.
If V1 is bumped more positive, V2 is pushed more negative and thus V1 is pushed more positive, etc...
Positive Feedback ensure until V1=VDD and V2=0
Think of of the quasi-stable point in the analogy of balancing a ball on a hill:
Middle point is quasistable and can only be preserved in a noiseless system. Starting points to the left of middle are pushed towards the left and starting points on the right are pushed to the right
Maintaining Highest Potential Energy:
Quasi-stable point can only be preserved in a noiseless system
First consider for what types of memories this is relevant:
Two Inverter (SRAM):
alternative view as extension of clocked SR latch:
The question of interest is how long does is take to reach a stable state as a function of starting position?
Given VA(0) how long to reach a stable point?
Given T, a time limit to reach stable point, what is the constraint on VA(0)?
Clocked latches can be placed near a metastable state if the clock transistions while data is
The problem is more than complex than "will it settle to the new or old value?"
If placed in a metastable state, the output can be stuck near the quasistable point for a very long time and the following stages will non-deterministically interpret the signal as 1 or 0 depending on thesholds and noise
Assume in the Logic, D1 and D2 represent two conditions that are normally never true but are often opposite in value. This logic is checking if the condition that D1 and D2 are ever true together
Here the output may glitch or become high in error under normal conditions, such as simutanuous tranistions D1:0→1 and D2:1→0
At some point in the system, parallel resolution must happen that can cause a problem
The same input signal (staring point) can be resolved differently in two components
A generic representation of positive feedback:
Assume the Ideal Amp, at all times VB=VA⋅A
General behaviour of positive feedback is to accellerate away from a quasi stable point (VA=VB=0)
Negative feedback for comparison:
fast than slow settle at a equilibrium
Analyze the temporal voltage change at the input storage capacitor. The voltage value here represents the state of the system
dtdVA=CI=R⋅CVB−VA=R⋅CAVA−VA
dtdVA=R⋅C(A−1)VA
VA(t)=VA(0)etKRCA−1
Time to reach VDD/2 ?
Approach:
Solve for t in VA(t)=VDD/2
t=K1ln(VA(0)VDD/2)
This is the time to reach VDD/2 given some starting point VA(0)
Bound t to be less than a desired clk-to-Q, i.e. time to resolve, Tr , and find ∣VA(0)∣min
Tr=K1ln(VA(0)VDD/2)
VA(0)=±(VDD/2)−1eTrK
Tr and VA(0) here represent the time and voltage constraints on the input
Starting values farther from the metastable equilibrium than ∣VA(0)∣min will resolve quickly
Starting values closer to the metastable equilibrium than ∣VA(0)∣min will resolve in a much longer amount of time, and meanwhile are more susceptible to noise corruption
Nature of Exponentials (Self-similartity)
For a given rise time, an input transition must stay outside a time window around the clock event
Implications of data transition slope and timing
The rise time and allowed arrival time of a 50% transition point are related
For the next analysis: instead of viewing the clk as the reference and data timing as the variable:
We will switch and consider the data edge to be the transition
2∣TW∣≈T10−90VDD2VA(0)
2∣TW∣≈T10−90e−KTr
∣TW∣=Ce−KTr
Tr<Fclk1−TPD→TW=Ce−KTr
TW is the window of error [eq 3.36] that helps define setup and hold time requirements and thus allowed combinatorial propagation delays
N⋅P failures in timespan Tspan
TspanN⋅P=ExpectedRateofFailure
The Rate of Failure is INVERSELY RELATED TO the mean time between failures
MTBF=R⋅P1;R=TspanN
R is the data transition rate
Actel 1989 ACT-1 Logic:
Page 127
R (rate of transitions) is 1/10th of the clock rate
Effective Time provided is doubled (2T):
V3(2T)=V3(T)V2(T)updateeKT=V1(0)eKTeKT=V1(0)eK2T
TW=Ce−2KT=C(e−KT)2
Pfail∝TW1
Pfail=C1K21 for 2 flip-flops
Pfail=C1KN1 for N flip-flops
Another Possibility to realize this effect for a multi stage (e.g. master-slave) flip-flop:
Metastability-Hardened FF (rather than just a use higher-power fast FF, it would be a low-power flipflop specifically designed with large K)
Sample Less Often (Sacrifice Responsiveness)
Use Edge-Sharpening Amplifiers at the Input (Fast Transitions)