Voltage Budget Example EE290C Spring 2011

Lecture 6: Link Performance Analysis

Elad Alon Dept. of EECS

EE290C Lecture 5 4

Motivation
Vin ,ampl Voff BER = 1 erfc 2 2 noise

Voltage Budget Example

Does eqn. above predict everything?

EE290C

Lecture 5

EE290C

Lecture 5

Traditional Approach
Borrowed from computer systems
Built to be error free Worst-case analysis

Timing Budget: Jitter Definitions

Like voltage need to separate deterministic (bounded) from random (unbounded) Total jitter (TJ) = Deterministic (DJ) + random (RJ) DJ is measured as peak-to-peak, added linearly RJ is measured as rms
For BER = 10-12, peak-to-peak RJ is 14*RMS value

Voltage/Time (VT) Budget

Also called link budget (especially in wireless) Key: separate random vs. deterministic error sources

Dont forget that jitter gets filtered too

E.g., source synchronous, CDR
EE290C Lecture 5 3 EE290C Lecture 5 6

Jitter Tolerance Mask

CDR Tolerance & XAUI Sinusoidal Jitter Tolerance Mask
9.00 8.50 8.00 7.50 7.00

Issues with Traditional Approach

Is worst-case realistic?
What if you had 100 taps of residual ISI?

Jitter Amplitude (UI)

6.50 6.00 5.50 5.00 4.50 4.00 3.50 3.00 2.50 2.00 1.50 1.00 0.50 0.00
0.010 22KHz 0.100

XAUI Sinusoidal Jitter Mask

XAUI Mask LV at Vtt-Rx=1V & 3.125Gbps

Can you really treat timing and voltage noise completely separately?

1.875MHz 0.1UI 1.000

20 MHz

0.1UI

10.000

100.000

Jitter Frequency (MHz)

Specifies amplitude of sinusoidal jitter (SJ) vs. frequency for which link must maintain BER spec
Mask drives CDR loop filter characteristics
EE290C Lecture 5 7 EE290C Lecture 5 10

Jitter in Source Synchronous Systems

Communications Approach
Model small deterministic errors as Gaussian
Find , multiply it to get peak-to-peak value at given BER

Works well at low BER (1e-3), but

EE290C

Lecture 5

EE290C

Lecture 5

11

Jitter in Source Synchronous Systems

Issues with Communications Approach

Cumulative ISI distribution
probability [cdf] Steady-State Phase Probability 0 -2 -4 -6 -8 -10 0 -2 -4 -6 -8 -10
4% T s ym bol erro r @ 10
-1 0

CDR phase distribution

9% T s ym bol

log

10

40m V erro r @ 10 25% o f eye h eig h t

25

50 75 100 re sidual IS I [m V ]

log

-1 0

10

80

100 120 140 160 180 p hase co unt

Gaussian model breaks at lower probabilities

EE290C Lecture 5 9 EE290C Lecture 5 12

A More Modern Approach

Use direct noise and ISI statistics
probability distribution of residual ISI

Real ISI Dist.: 5-tap TX FIR

0.04

Dont treat timing noise separately

Integrate with voltage noise sources I.e., need to map from time to voltage

0.03

0.02

Ref: V. Stojanovic, Ph.D. thesis

0.01

probability distribution of residual ISI

0.016

0.012

0.008

0.004

0 -60

-40

-20

0 20 voltage [mV]

40

60

0 -150

-100

-50

50 100 voltage [mV]

150

Data sample distribution

EE290C Lecture 5 13 EE290C Lecture 5

Edge sample distribution

16

Residual ISI
Generally cant correct for all ISI
Equalizers are finite length Even with infinite equalizers, the coefficients are quantized And you may not be able to estimate coefficient values perfectly

Effect of Timing Noise

Need to map from time to voltage
Voltage noise when receiver clock is off
Jittered sampling Ideal sampling

Voltage noise

Need to find distribution of this residual ISI Effect depends on jitter magnitude, input sequence, and channel
EE290C Lecture 5 14 EE290C Lecture 5 17

Generating ISI Distributions

Convolution
probability [pmf]
1 0.75 0.5

Effect of TX Jitter: Decomposition

bk
bk
1
TX1 k+
(k + 1)T
kT

ideal
(k +1)T

Equalized pulse response

200

ISI distribution
probability [pmf]
1 0.75 0.5 0.25 0

kTX
kT

0.25 0

TX k

bk
TX1 k+

voltage [mV]

-15

[mV]

15

probability [pmf]

bk
-25 -5 5 25

noise

1 0.75 0.5 0.25 0

15 -10 -3 -2 -1 0 1 2

sample #

voltage [mV]

Decompose output into ideal + noise Jitter is pulse at front and end of symbol
Width of pulse set by jitter magnitude

-10

[mV]

10

EE290C

Lecture 5

15

EE290C

Lecture 5

18

Converting to Voltage
b
TX1 k+
kTX
TX k k +1

Implications
Power spectral density [dBV]

bk

For same source jitter (white)

Noise from TX jitter larger than RX jitter TX jitter enhanced by channel

-30

Source white jitter Noise from Tx jitter

-40

bk

TX bk k

-50

Variable width pulses annoying to deal with Approximate noise pulses with deltas of same area
Channel is low-pass and will filter them anyways
EE290C Lecture 5 19

Noise from Rx jitter

-60

Bandwidth of jitter sets final magnitude

Like any other white noise source
EE290C Lecture 5

0.5

1.5 2 2.5 3 frequency [GHz ]

22

Jitter Propagation Model

TX
precoder

Including the CDR

+

ak
nvdd

w
Hjit(s)

bk

ideal

pulse response

p( jT )

x kISI

xk

ak
kRX
RX

RX Slicer

dn

dn

dataOut deserializer Phase control Phase mixer ref Clk PLL

nin

PLL
TX k ,k +1

noise h jT + T
2

x kjitTX

PD

en
data Clk

impulse response

en (late) dn-1

x ISI ( kT + i + kRX ) =
x jitter (kT + i + kRX ) =

j = sbS

sbE

k j

p ( jT + i )

edge Clk

j = sbS

sbE

k j

T T RX TX RX TX h( jT + 2 + i ) k k j h( jT 2 + i ) k k +1 j

Need jitter on actual sampling clock

Not just jitter from source, PLL

Channel bandwidth matters

If h(T/2) is small, jitter voltage noise is small
EE290C Lecture 5 20

EE290C

Lecture 5

23

Implications
TX Jitter
High frequency (period) jitter is bad
Changes the energy (area) of the symbol) Uncorrelated noise sources add up

CDR Model
Probability

Model as a state machine

Current phase position is the state Transitions caused by early/late signals

1 p-up 0.8
Accumulate-reset filter, length 4

p-dn

0.6 0.4 0.2 0 0 p-early

Low frequency jitter isnt as bad

Just shifts waveform Correlated noise sources partially cancel

p-no-valid transitions p-late p-hold

RX jitter
EE290C

kRx

Lecture 5

kRx

On average move to right position

50

100

150

200 250 Phase count

But probability of incorrect transition not small

Shifts TX sequence same as low f TX jitter

21 EE290C Lecture 5 24

What You Really Care About

Final steady-state probability of being in each phase position (state)
Probability 1 p-up 0.8
Accumulate-reset filter, length 4

Putting It All Together: BER Contours

5 tap Tx Eq
150
-5

5 tap Tx Eq + 1 tap DFE

150
-5

100
p-dn

100
-10

pdn,i

margin [mV]

margin [mV]

phold ,i

50 0 -50

50 0 -50 -100 -150 0

-10

0.6 0.4 p-early

p-no-valid transitions p-late p-hold

-15

-15

i1 i

pup,i

-20

-20

i +1

0.2 0 0

-100
50 100 150 200 250 Phase count

-25

-25

-150 0

20

40

Can find using Markov model

Fancy way of saying that you iteratively apply the transition probabilities
EE290C Lecture 5 25

60 80 100 120 140 160 time [ps]

-30

20

40

60 80 100 120 140 160 time [ps]

-30

Voltage margin @ given BER:

Distance from probability contour to threshold (0)
EE290C Lecture 5 28

Result: CDR Phase PDF

0 log 10 Steady-State Probability -5

Model and Measurements

0 -2 log10(BER) -4 -6 -8 -10 -12

-10

-15

50

100

150

200 250 Phase Count

-14 80 60 40 20 0 -20 -40 -60 -80 Voltage Margin [mV]

This is the jitter distribution

(With nominal phase subtracted out)
EE290C Lecture 5 26 EE290C

Lecture 5

29

ISI and CDR Distributions

300 2PAM - lin. eq 200 100 0 -100 -200 -300 50
-10 -5

voltage [mV]

-15

-20

-25

60

70

80 90 time [ps]

100

110

-30

Vertical slice: ISI distribution @ given time Horizontal weight: CDR phase distribution

EE290C

Lecture 5

27