Embedded Systems: Hardware: - PowerPoint PPT Presentation

About This Presentation
Title:

Embedded Systems: Hardware:

Description:

Embedded Systems: Hardware: Using Combinational Logic in Applications: Signal Levels, Time, Physical Properties; Testing, Structural and Functional Faults; – PowerPoint PPT presentation

Number of Views:77
Avg rating:3.0/5.0
Slides: 41
Provided by: Andre924
Learn more at: https://eecs.ceas.uc.edu
Category:

less

Transcript and Presenter's Notes

Title: Embedded Systems: Hardware:


1
  • Embedded Systems Hardware
  • Using Combinational Logic in Applications
  • Signal Levels, Time, Physical Properties
  • Testing, Structural and Functional Faults
  • Using Verilog to Model Timing Delays
  • Main reference Peckol, Chapter 2

2
  • Main theme of this chapter
  • The world is ANALOG, not digital
  • Even in designing combination logic, we need to
    take analog effects into account
  • Software doesnt change but hardware does
  • --different manufacturers of the same component
  • --different batches from the same manufacturer
  • --environmental effects
  • --aging
  • --noise
  • 3 main areas of concern
  • --signal levels
  • --timing
  • --how to deal with effects of unwanted
    resistance, capacitance, induction

3
table_02_00
SIGNAL LEVELS 0, 1, and actual
values ideal Logical 0 0 volts Logical 1
5 volts (or 3.3 volts or ) actual (vendor
specifications) output gt 2.5 V to be
interpreted as 1 output lt 0.4 V to be
interpreted as 0 High-level noise immunity
(margin) VOH VIH 0.5 Low-level noise
immunity (margin) VIL VOL 0.4
table_02_00
0.8
4
fig_02_00
0
0
1
1
Logic level variations summary (bubble logic 0)
fig_02_00
fig_02_00
fig_02_01
5
fig_02_02
fig_02_02
fig_02_03
fig_02_04
TTL
MOS (CMOS)
Typical transistor families Q where is the
resistor in the MOS circuit? Q why is CMOS
preferred for todays ICs?
6
fig_02_05
Fan-in devices input current requirements how
much current does the device source to other
devices when its input is in logical 0 state how
much does it sink from other devices when input
is in logical 1 state? and Fan-out how many
devices can this gate drive with out degraing its
specified minimum and maximum output levels how
much current device sources to other devices in
logic high state and how much it sinks from other
devices in logic low state
Terminology In top picture Inv1 is sourcing
current to Inv2 and Inv2 is sinking current from
Inv1 In bottom picture Inv2 is sourcing current
to Inv1 and Inv1 is sinking current from Inv2
fig_02_05
table_02_01
Example values (SNL4LS04 data sheet)
7
fig_02_06
Computing fanout for the SNL4LS04 device use
lower of 2 values for logic 1/0 states Output
0 can sink 8mA, source -400?A fanout
8mA/-400?A 20 Output 1 similar calculation
gives fan-out 20 (same in this case) Example
2.0
Driver driving LED and several gates current
through LED when driver is at logic 0 Specs
inverter IIL -400?A IIH 20uA
driver IOL 24mA at VOL 0.2V IOH -15mA at
VOH 3.5V Logic 1 (fig. 2_07) at N1, i1 i2
i3 0 LED is off
i3 i1 15mA
fanout i3/i4 15mA / 20?A 750 Logic 0 (fig
2_08) at N1, i1 i2 i3 0 LED is on
i3 i1 i2 i3 15mA 10ma
5 mA fanout i3/i4
5mA/20?A 250
fig_02_06
fig_02_08
fig_02_07
8
fig_02_09
Adding resistors to measure ON resistance from
transistors
fig_02_09
Output Sinking current drop across R2 and
output voltage will be above 0.0 VDC Sourcing
current drop across R1, decrease in ideal output
of 5 VDC Input Sourcing current (input 0)
drop across R3 worst case if input 0.0 VDC
output is VOL but making input negative can
damage the part Sinking current drop across R4.
but forcing device to exceed limit can damage it.
9
fig_02_10
fig_02_10
TIME rise and fall time not 0 in
real life must allow for
these Verilog code p. 61 //syntax
(riseTime, fallTime) deviceinstance //in part
model parameter riseTime 1 parameter fallTime
2 Not (riseTime, fallTime) myNot (sigOut,
sigIn)
10
fig_02_11
TIME propagation delay Embedded systems
usually trying to meet a deadline, so use
longest combinational delay Verilog code p.
64 //syntax // delay LHS RHS //RHS changes
and is assigned to LHS //after a
delay //inclusion in part model (2 time
units) parameter propagationDelay 2 Not
(propagationDelay) myNot (sigIn,
sigOut) //example Q NOT in Alterawhy?
fig_02_11
fig_02_12
11
fig_02_13
Transport and inertial delays Transport delay
model input changes, after a specified interval,
output changes Inertial delay accounts for
physical movement of electronic charge within the
device voltage level within device much reach
specified minimum level before recognized as 0 or
q (so signal duration must be greater than the
specified minimum level) usually set to less
than or equal to the propagation delay of the
device
fig_02_13
12
  • Race conditions and hazards (glitches)
  • Critical state or output depends on order of
    arrival at decision point
  • Noncritical output value does not depend on
    order of arrival of inputs
  • Hazard (also called a decoding spike or a
    glitch) present in a circuit if the circuit has
    the possibility of giving an incorrect output
  • 2 types of hazards
  • Static glitch may occur because of race between
    2 or more input signals when output expected to
    remain at steady level
  • Static-0 may produce erroneous 1 static-1 the
    opposite
  • Dynamic output may erroneously change more than
    once as result of one single input transition

13
fig_02_14
Examples static-0 hazard Extra delay through
inverter Static-1 hazard Adding buffers to
match delays Will not work because of Parameter
variations occurring In real physical parts
fig_02_14
14
fig_02_15
Additional examples for analysis
fig_02_15
15
fig_02_16_01
fig_02_16_01
16
fig_02_16_02
fig_02_16_02
17
fig_02_17
Example dynamic hazard One slow path and one
fast path other devices are assumed to have
typical delays, all of the same value If B 0 ? 1
there will be 3 state changes in the output
before it settles
fig_02_17
18
fig_02_18
fig_02_18
19
fig_02_19
LEGACY OF THE EARLY PHYSICISTS RESISTANCE,
CAPACITANCE, COUPLING (micro view, passive
components) Ampere current flowing in a wire
produces magnetic field Faraday, Lenz wire
moving in magnetic field has induced
current Gauss et al. capacitance Situations
to examine Coupling between two adjacent
wires Mutual capacitance between adjacent
circuits etc.
20
fig_02_19
PHYSICAL PROPERTIES RESISTANCE R
fig_02_19
R r (L / A)
Q what does this say about --length
of wires? --feature sizes? --noise
margins for low voltage?
Modeling resistance (first-order model, includes
inherent parasitic devices) for DC, L and C can
be ignored but in our circuits we will have
time-varying signals We are assuming a lumped
system (all resistance considered to be lumped
at one node) For a distributed system we would
look at R(x)dx, L(x)dx, C(x)dx
21
DC can ignore L and C Time-changing
signals Impedance ZL(w) Lw Capacitance
ZC(w) 1 / Cw Laplace transform Z(s) Ls
R 1/Cs Ls R/(RCs 1) Gives for w 0,
z(w) R for w ? infinity, z(w) L Z(w) for
R 10K, 1K, 0.1K at 10GHz, inductor becomes
dominant
fig_02_21
22
fig_02_22
Capacitance C e A/d Many instances of
capacitors on chip --Power/ground
planes --parallel wires --adjacent
pins --etc. Example part of signal in top wire
shows up as noise in adjacent wire
fig_02_22
fig_02_23
23
fig_02_24
First-order (lumped) model Using Laplace
transform gives Z(s) 1/Cs Ls R inductor
dominates at higher frequencies For C 1 muf,
0,1 muf, 0.01 muf
fig_02_24
fig_02_25
24
fig_02_26
How do these effects change logic
circuit? Example 2 inverters in
series Resistor connecting path Capacitor
device, wire, IC package, coupling to other
devices VOUT (s) 1/Cs / R 1/Cs V(s)IN
1/(RCs1) V(s)IN
1/(RCs1) VIN/s for VIN a step
function VOUT(t) VIN(1-exp(-t/RC)) Rise (and
fall) times are slowed Components can be damages
or Data rate can be reduced
fig_02_26
fig_02_27 interconnect
fig_02_28interconnect, driver
fig_02_29 rise time
25
fig_02_30
Example tristate driver Enable different data
sources to use system bus If driver disables,
pullup resistor controls bus VOUT(t)
VIN(1-exp(-t/RC)) Rise time is increased
and Receiving device can enter metastable region
where there is oscillation in its output
fig_02_30
fig_02_31
fig_02_32
26
fig_02_33
Example why you should never leave Gate inputs
floating (using a 3-input AND gate for a 2-input
application)
  • 1
  • VOUT(s) C1/(C1C2)VIN(s)
  • If voltage too low, output is always 0
  • Cap C1 C2
  • This doubles time constant, reduces rise/fall
    time can give metastable behavior on switching
  • 3. State of ununsed pin defined by pullup
    resistor, this will work

3 methods 1 2 3
fig_02_35
fig_02_33
fig_02_34
27
fig_02_36
Second-order add parallel inductor This
adds a damping factor Natural frequency wn 1/
(LC)1/2 damping d (R/2) (L/C)1/2 d lt 1
underdampedcan have oscillation, noise d 1
damped okay d gt 1 overdampedcan have
metastability
fig_02_37
fig_02_36
28
fig_02_39
Testing combinational circuits Fault-unsatisfactor
y condition or state can be constant,
intermittent, or transient can be static or
dynamic Error static inherent in system most
can be caught Failure undesired dynamic event
occurring at a specific timetypically
randomoften occurs from breakage or age cannot
all be designed away Physical faults one-fault
model Logical faults Structuralfrom
interconnections Functional within a component
29
fig_02_39
Structural faults Stuck-at model (a single
fault model) s-a-1 s-a-0 may be because circuit
is open or there is a short
30
fig_02_39
Testing combinational circuits s-a-0 fault
fig_02_39
31
fig_02_40
Modeling s-a-0 fault
fig_02_40
32
fig_02_41
S-a-1 fault
fig_02_41
33
fig_02_42
Modeling s-a-1 fault
fig_02_42
34
fig_02_43
Open circuit fault appears as a s-a-0 fault
fig_02_43
35
fig_02_44
Bridging fault bad connections, broken flakes,
errant wire pieces
fig_02_44
36
fig_02_45
Examples of bridging faults
fig_02_45
37
fig_02_46
Bridging faults can be feedback or non-feedback
faults Non-feedback faults Between input or
output and power rail use stuck-at model Between
signal traces or logic pins inputs model as
common signal to both inputs internal who wins?
fig_02_46
38
fig_02_47
Modeling a competitive fault result of fault
depends on logic family being used
fig_02_47
39
fig_02_48
Feedback bridging faults Number of inversions is
important Circuit A Circuit B In A
there are an odd number of inversions on the
path this can cause oscillation can sometimes
be modeled as competing signals In B there are an
even number of inversions this can oftn be
modeled as a stuck-at fault
fig_02_48
40
fig_02_49
Functional faults Example hazards, race
conditions Two possible methods A consider
devices to be delay-free, add spike generator B
add delay elements on paths Method
A Method B As frequencies increase,
eliminating hazards through good design iseven
more important
fig_02_49
fig_02_50
Write a Comment
User Comments (0)
About PowerShow.com