Professional Documents
Culture Documents
.
.
.
.
.
.
.
7
7
7
8
8
9
9
10
.
.
.
.
.
.
.
.
13
13
13
14
14
14
14
15
16
3 Phase-Locked Loops
1
Duration: 2 weeks . . . . . . . . . . . . . . . . . . . . . . . . .
2
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3
Laboratory Work . . . . . . . . . . . . . . . . . . . . . . . . .
19
19
19
20
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.re
jin
pa
ul
.c
Contents
w
w
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
w
w
.re
jin
pa
ul
.c
.
.
.
.
.
.
.
.
.
.
20
22
22
23
24
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
27
27
27
27
28
29
29
29
29
30
31
31
31
32
32
33
33
33
34
om
3.1
Charge Pumps .
3.2
Loop Filters . . .
3.3
Open-Loop Test .
3.4
Closed-Loop Test
Laboratory Report . . .
om
.c
List of Figures
ul
.re
jin
pa
. 20
.
.
.
.
21
22
23
24
w
w
om
.c
List of Tables
w
w
.re
jin
pa
ul
om
Chapter 1
.re
jin
pa
ul
.c
Current-Controlled Ring
Oscillators
1
Duration : 2 weeks
Introduction
w
w
vn2
,
(dvo /dt)2
.c
2 =
om
edges of the waveform of the oscillator. (ii) A fast transition minimizes the
timing jitter of the oscillator. As pointed out in [2] the noise of the MOS devices and the switching noise of the VCOs output at the threshold-crossing
shifts the actual threshold-crossing time (timing jitter) by an amount that is
proportional to the power of the total noise injected at the threshold-crossing
and inversely proportional to the slew rate of the output voltage [2].
(1.1)
3
3.1
.re
jin
pa
ul
where 2 is the timing jitter, vn2 is the power of the total noise injected at
the threshold-crossing, and dvo /dt is the slew rate of the output voltage of
the oscillator at the threshold-crossing. (iii) The latch effectively rejects the
noise presented on the power and ground rails once the latch is established.
Laboratory Work
Oscillator
w
w
VDD2
Vo+
VoVin+
4
Vin-
om
M6
M5
I ctrl
CL
M3
M1
M2
M4
.c
3.2
ul
3.3
.re
jin
pa
A static CMOS inverter whose dimension is much larger than the dimension
of the transistors in the delay stage of the oscillator is powered by the same
power supply as that of the oscillator. Let the load of the large CMOS
inverter be a large capacitor. The input of the large CMOS inverter is an
ideal square wave generator whose period is much smaller than the oscillation
period of the oscillator. This clocked large inverter is used to simulate the
injection of switching noise to the power rail of the oscillator.
w
w
Construct the symbol of the delay cell of the oscillator that has (i)
two input terminals vin+ and vin , (ii) two output terminals vo+ and
vo , (iii) one VDD connection terminal, (iv) one VSS connection terminal, (v) one n-well connection terminal vnwell , and (vi) one substrate
connection terminal vsub , as per Fig.1.2.
Construct the current controlled oscillator by including (i) all delay
stags, (ii) the biasing circuit, (iii) the switching noise injection circuit,
and (iv) VDD and VSS circuitry including bonding wires.
Perform time-domain analysis of the designed oscillator. The start of
10
Vn-well
Vo-
Vin+
Vo+
Vsub
.c
VSS
om
Delay cell
Vin-
ul
.re
jin
pa
Plot the output voltage of each stage of the oscillator with two different
control currents. Measure the rise and fall times of the output voltage
in both cases. Measure the timing jitter. Comment on your findings.
w
w
Laboratory Report
11
om
.c
The waveform of the voltage of the output of each stage of the currentcontrolled oscillator with two different transistor widths of the latch.
Measure the peak-to-peak timing jitter at the threshold-crossing points
of the output waveform of the oscillator.
.re
jin
pa
ul
The waveforms of the voltage of the output of each stage of the currentcontrolled oscillator with two different control currents. Measured the
rise and fall times in these two cases. Measure the peak-to-peak timing
jitter at the threshold-crossing points of the output waveform of the
oscillator.
w
w
om
.c
Bibliography
ul
[1] X. Mailand, F. Devisch, and M. Kuijk, A 900 Mb/s CMOS data recovery DLL using half-frequency clock, IEEE Journal of Solid-State
Circuits, vol. 37, No. 6, pp 711-715, June 2002.
w
w
.re
jin
pa
12
om
Chapter 2
.re
jin
pa
ul
.c
D-flipflop Phase/Frequency
Detectors
1
Duration : 2 weeks
Introduction
w
w
13
14
3.1
Laboratory Work
D-Flipflops
.c
om
3.2
.re
jin
pa
ul
w
w
3.3
AND2 Gate
D R
QA
QB
Vm
Vm
0 2p
Df
QB
Vm
A
B
QA
QB
QB
(a) Df=0
Df
Vm /4
ul
QA
-2p
om
QA
.c
15
(b) Df=p/2
3.4
.re
jin
pa
w
w
16
CLK
Q
CLK
CLK
CLK
(a)
(b)
.c
om
CLK
ul
.re
jin
pa
Laboratory Report
w
w
17
om
.
2
.c
.re
jin
pa
ul
w
w
3
.
2
om
.c
Bibliography
ul
[1] J. Yuan and C. Svensson, New single-clock CMOS latches and flipflops
with improved speed and power savings, IEEE J. Solid-State Circuits,
vol. 32, No.1, pp. 62-69, Jan. 1997.
.re
jin
pa
w
w
[3] K. Martin, Digital Integrated Circuit Design, New York : Oxford University Press, 2000.
18
Duration: 2 weeks
Introduction
.re
jin
pa
ul
.c
Phase-Locked Loops
om
Chapter 3
w
w
19
20
om
Loop filter
Voltage-controlled oscillator
PFD
Vvco
DN
C1
C2
R2
ul
JDN
.c
Vin
.re
jin
pa
Laboratory Work
Charge Pumps
w
w
3.1
21
.c
om
JDN from the output node, depending upon the control signals UP and DN
from the preceding D-flipflop phase/frequency detector. For more information on charge pumps, please read reference [1] and the main reference text
of the course.
Construct the schematic of the charge pump and its symbol. The charge
pump symbol should have the following inputs : UP, UP, DN, DN, Vout , VDD
and VSS . All n-well and substrate contacts are connected to the same power
supply and ground as those of the charge pump.
M6
ul
M5
.re
jin
pa
vo
UP
UP
M3
M4
J UP
DN
M1
DN
M2
J DN
w
w
M7
M6
M5
vo
M8
UP
UP
M3
M10
M4
DN
DN
M1
M11
M2
.c
M9
om
Vcp
22
Loop Filters
.re
jin
pa
3.2
ul
w
w
In order to convert the output current of the preceding charge pump into a
voltage whose value is proportional to the pulse width of the control signals
UP and DN of the charge pump, a loop filter consisting of two capacitors C1
and C2 and a resistor R2 is employed at the output of the charge pump, as
shown in Fig.3.1. Note that C2 C1 . C2 and R2 form the main part of the
loop filter. Note that high-frequency sparks can pass through R2 C2 network
and create reference spurs at the output of the oscillator. To minimize this
unwanted effect, a smaller capacitor C1 is used in parallel with R2 C2 to
filter out the high-frequency sparks.
Construct the schematic of the loop filter and its symbol. The value of
C1 , C2 , and R2 should be assigned as variables so that they can be tuned
when the phase-locked loop is constructed.
3.3
Open-Loop Test
23
Charge pump
JUP
.c
Loop filter
Vc
PD
DN
C2
R2
C1
.re
jin
pa
JDN
ul
A
B
UP
om
3.4
Closed-Loop Test
w
w
Construct a 4-stage fully differential cross-coupled voltage-controlled oscillator as per Fig.3.5. The VCO is very similar to the CCO that you designed in
previous laboratories except that the oscillation frequency is now controlled
by a voltage rather than a current. The oscillation frequency of the oscillator
should be in the neighborhood of 900 MHz. Simulate the VCO for a given
control voltage Vc and record the oscillation period (frequency). Make sure
that the oscillation frequency of the oscillator covers 900 MHz.
Construct a test schematic containing (i) the D-flipflop phase/frequency
detector, (ii) the current-steering charge pump, (iii) the loop filter design,
and the 4-stage 900 MHz fully differential cross-coupled voltage-controlled
oscillator as per Fig.3.5. Apply a 900 MHz square wave to the input A of the
phase/frequency detector. The other input of the phase/frequency detector
B comes from the output of your VCO. Plot the waveform of A and B, as
well as the control voltage Vc of the VCO. When a lock state of the PLL
24
Vo-
Vin-
.re
jin
pa
Vin+
Vo+
ul
Vc
.c
om
Vc
Laboratory Report
w
w
25
om
w
w
.re
jin
pa
ul
.c
om
.c
Bibliography
w
w
.re
jin
pa
ul
26
om
Chapter 4
Duration : 4 weeks
Introduction
.re
jin
pa
ul
.c
w
w
CMC provides many packaging options, depending upon the operation frequencies of the design chips. In this lab, you will use 24 CFP (Ceramic
27
28
om
Flat Pack), which has 24 pins, a small cavity and is suitable for highfrequency (up to 4 GHz) designs. CMC has developed a library for both the
package and testfixture for 24 CFP. These files are located in
Library :
.c
package
ul
Cellview :
24cfp_140_half_cell, 24cfp_140_test_fixture_half
.re
jin
pa
Note that only a half of the package is provided. You need to make a
copy and rotate it to complete the package. These models are valid up to 4
GHz. CMC requires that all unused pins be grounded. This will increase the
level of signal isolation and improve the performance.
w
w
Bond pads are not part of the above package. You need to add bond pads
explicitly in your design to account for their capacitive effect. CMC does
not have bond pad cells developed in package library. CMC requires that
the bond pads must not be smaller than 6464. We use 6464 in this
laboratory. You need to use this value to estimate the parasitic capacitance
of the bond pads. Bond pads should be modeled as a shunt capacitor between
signal path and the substrate. M6 M5 are to be connected together and
used for the bonding pads. You need to look at the electrical parameter file
of TSMC-0.18m CMOS technology to find out the capacitance from M5 to
substrate per unit area to estimate the capacitance of the pads.
29
.re
jin
pa
ul
.c
om
Bond wires are also not part of the above package. You need to add bond
wires explicitly in your design in order to account for their inductive effect.
CMC has developed a bond wire cell located at
1) Library : package
2) Cellview : bondwire
The bond wires have a diameter of 31.75 m. The length of the bond
wires depends upon the pin/pad assignment and the location of pads. They
can only be determined once the complete layout is available. For your
simulation, we use the default parameters given in the library (hwl (horizontal
wire length)=2000 m, d = 31.75m, r = 0.05,...).
Schematics
In order to take into account the effect of bond wires, bond pads, and packaging, you need to add
Bond pads
24cfp_140_half_cell
24cfp_140_test_fixture_half
w
w
External test devices, such as supply voltage and loads can be connected
directly to the testfixture. The arrangement is shown in Fig.4.1.
7.1
Laboratory Work
Complete Test Circuit Construction
Construct the complete test circuit of the designed phase-locked loop. The
test circuit should contain the follows
30
Cavity
Physical boundary
of CFP chip
Bondwire
om
Pad
Your
design
Load
.c
External
voltage
source
Testfixture
ul
.re
jin
pa
w
w
7.2
To minimize the effect of switching noise, for each VDD and VSS connection
to the outside, four pads/bond wires should be used. The number of pads
are as the followings :
31
ul
7.3
.c
om
7.4
.re
jin
pa
To bring the output voltage of the VCO to the outside of the chip, buffers
are needed in order to drive the parasitics of the package and the load. Add
eight voltage buffers to the eight outputs of the VCO (Two inverters are
connected in series with the first inverter of small dimensions. Note that the
first inverter should not be made large. Otherwise, it will load the VCO.).
PLL Test
w
w
Apply a 900 MHz 50% duty-cycle square wave to the input A of the phase
detector. The other input of the phase detector B comes from the output of
your VCO. Plot the waveform of A and B, as well as the control voltage Vc of
the VCO. When a lock state of the PLL is established, the phase difference
between A and B should be zero and the control voltage Vc of the VCO
should settle down to a constant value.
Layout of PLL
Once the simulation of the designed PLL is completed and design specifications are met, it is the time to complete the physical design of the PLL
chipset. Since all of you already have the experience in layout in either
ELE704 or ELE734, you will find this step a rather enjoyable design expe-
32
rience. Note that you should not make any change to the dimension of all
transistors used in your schematic-level design.
Floor-Planning of PLL
om
8.1
8.2
.re
jin
pa
ul
.c
Before you proceed with the actual layout of the PLL chipset, the first step
that you should do is the floor planning. You MUST make a decision on
critical issues such as (i) where to route your power and ground rails. (ii)
Where to route signals, both low-frequency and high-frequency signals. (iii)
How to isolate high-frequency signals from others. (iv) How to assign I/O
pins. (v) How to ensure that n-wells are properly connected to the most
positive voltage. (vi) Where to put guard rails and how to put the guard
rails. (vii) How to protect dc biasing circuits from noisy digital circuits, (viii)
How to minimize switching noise, and (xi) How to minimize the interaction
between analog and digital circuits of the PLL system.
Layout of PLL
w
w
Once the floor of your PLL is planned, you can proceed with the layout
of each functional block of your PLL. The following guidelines should be
followed in the layout : (i) Try to put all n-wells together. This will reduce
the chip area required for n-wells. Use as many as n+ pull-up connections
as possible for n-wells to ensure that n-well is uniformly connected to the
most positive voltage of the circuit. (ii) Multiple p+ pull-down connections
of the substrate should be used throughout your layout to ensure that the
substrate of your chip is uniformly connected to the ground. (iii) Use as
many as pins/pads as possible for VDD and VSS connections to minimize
the inductance of the bonding wires of VDD and VSS pads subsequently
the switching noise associated with these pads. (iv) High-frequency signals
should be guarded with either ground or VDD rails to minimize their effect
on other parts of the circuit. The bottom line here is to minimize the loop
area of high-frequency signal lines so that their inductance is minimized.
33
Note that the shielding rails increase the capacitance of the high-frequency
signal lines.
om
8.3
8.4
.re
jin
pa
ul
.c
Layout of Pads
Following the design rules of TSMC for pads. Use only M6 and M5 for all
pads to minimize the capacitance from the pads to the substrate.
8.5
w
w
Output buffers are usually large in size and generate large switching noise and
substrate noise. To minimize their effect, the followings guidelines should be
followed : (i) Use a separate set of VDD and VSS pads and pins for the output
buffers to eliminate the injection of their switching noise to the system. (ii)
Use guard rings to isolate the output buffers from the rest of the system.
Guard ring VDD and VSS should be completely separated from those of the
rest of the system as they carry a high level of switching noise. (iii) Minimize
the length of the interconnects connecting the output buffers and their pads.
Techniques should be used to minimize the inductance of these interconnects.
34
Laboratory Report
om
.c
ul
.re
jin
pa
w
w