Voltage Controlled Oscillator Design - Short Course at NKFUST, 2013

壓控振盪器設計
李健榮助理教授
國立臺北科技大學電子工程系
Department of Electronic Engineering
National Taipei University of Technology

Outline
• Resonator
• Feedback, Two-port Reflection, and Negative Resistance
• Feedback Loop Analysis
• Amplifier Configurations
• Capacitor Ration with Colpitts Oscillators
• ADS Simulation Tips
• Phase Noise and Lesson’s Model
• Summary
Department of Electronic Engineering, NTUT2/60

Introduction
• An oscillator is a circuit that generates a periodic waveform.
• Oscillators are used with applications in which a reference
tone is required. In most RF applications, sinusoidal references
with a high degree of spectral purity (low phase noise) are
required.
90o
 I t
cos ct
 Q t
Low Noise Amplifier
(LNA)
Baseband
Processor
LPF
LPF
90o
 I t
cos ct
 Q t
 ms t
Power Amplifier
(PA)
Antenna
Baseband
Processor

LC Resonator
• An LC resonator determines the oscillation frequency and
often forms part of the feedback mechanism.
If i (t) = Ipulsed (is applied to the parallel
resonator, the system time response:
 
2
2 2
2 1 1
cos
4
t
RC
pulse
out
I e
v t t
C LC R C

  
    
   
2 2
1 1
4
osc
LC R C
  
1
osc
LC
 
L C R
 outv t
 i t
Time
Amplitude
R  
   pulsedi t I t

Adding Negative Resistance Through Feedback
• In any practical circuit, oscillations will die away unless
feedback is added to generate a negative resistance in order to
sustain the oscillation.
L C pR nR
L C sr
nr
feedback
active device
Parallel RLC Resonator Series RLC Resonator
feedback
active device

Design Method: Feedback System
• The oscillator can be seen as a linear feedback system.
• The gain of the system:
• Barkhausen’s Criterion:
For sustained oscillation at constant amplitude, the poles must be on the jω axis
which states that the open-loop gain around the loop is 1 and the phase around the
loop is 0 or some multiple of 2π.
• To find the poles of the closed-loop system, one can equate this
expression to zero, as in .
 
 
 
   1
out
in
V s G s
V s G s H s


   1 0G s H s 


    1G j H j       1G j H j       2G j H j n   and
Department of Electronic Engineering, NTUT
 G s
 H s
 outV s inV s
6/60

Design Method: Two-port Reflection
• Two-port Reflection
'
11 1G S  
'
22 1L S  
If the two-port network is oscillating at one port, it must be simultaneously
oscillating at the other port.
Sine the resonator is passive, thus 1G 

Design Method: Negative Resistance
• One-port Negative Resistance:
    0DR R  
    0DX X  
     Z j R jX   
     D D DZ j R jX   

Current Limiting
• If the overall resistance is negative, then the oscillation
amplitude will continue to grow indefinitely. In a practical
circuit, this is, of course, not possible.
• Current limiting (power rails, or nonlinearity) eventually limits
the oscillating magnitude to some finite value effect of the
negative resistance in the circuit until the losses are just
canceled, which is equivalent to reducing the loop gain to 1.
v
growth
t
limited

Implementations of Feedback
• Feedback (or Gm ) is usually provided in one of three ways:
 Colpitts oscillator:
Using a tapped capacitor and amplifier to form a feedback loop
 Hartley oscillator:
Using a tapped inductor and amplifier to form a feedback loop
 Gm oscillator:
Using two amplifiers in a positive feedback configuration
G
amplifier
G
amplifier
amplifier
G
L
buffer

Common Oscillator Configurations
bi
ci
C
E
B
1C
2C
3L
bi
ci
C
E
B
1L
2L
3C
bi
ci
C
E
B
1C
2C
3L
bi
ci
C
E
B
1C
2C
3L
Colpitts Hartley
Clapp Siler

Amplifier Configuration (Colpitts or –Gm)
• The Gm oscillator has either
 A CC amplifier made up of Q2 , and Q1 forms feedback
 A CB amplifier consisting of Q1, and Q2 forms feedback
• Colpitts and Hartley oscillators can be made either CB or CC.
C L
1Q
2Q
1C
2C
L
1Q
1C
2C
1Q
L
CB Colpitts CC ColpittsGm Oscillator
CC
CB

Loop Analysis (I)
• Loop analysis gives information about the oscillator :
(1) Determine the frequency of oscillation
(2) The amount of gain required to start the oscillation
1C
2C
L
1Q
Common base
2C
1C
LpR
er
ev
c m ei g v
At the collector,  1 1
1 1
0c e m
p
v sC v sC g
R sL
 
      
 
At the emitter, 1 2 1
1
0e c
e
v sC sC v sC
r
 
    
 
1 1
1 1 2
1 1
0
01
m
p c
e
e
sC sC g
R sL v
v
sC sC sC
r
 
    
               
 
cv
ev
cv
13/60

Loop Analysis (II)
• The conditions for oscillation:
where ωconr is the corner frequency of the HPF formed
by the capacitive feedback divider.
 1 1 2 1 1
1 1 1
0m
p e
sC sC sC sC sC g
R sL r
  
         
  
 1 23 2 1
1 2 1 1 2
1
0m
p e p e e
L C C LC L
s LC C s LC g s C C
R r R r r
   
               
1 2
1 2 1 2
1 1
e p
C C
C C L r R C C L

 
  
 
p
L
R
Q
L

   
2
2 0
0 0 0
1 2 1 2
1 1
1 1
e p p e L conr
L L
r R C C R r C C Q
 
   
  
     
 
 1 2
m
L
C C
g
Q
 

Tells us what value of gm
(and corresponding value
of re) will result in
sustained oscillation. For
a real oscillator gm would
have to be made larger
than this value to
overcome any additional
losses not properly
modeled.
, , and
14/60

Capacitor Ratios with Colpitts (I)
• The capacitive divider (C1 ,
C2 , and re) affects oscillation
frequency and feedback gain,
which acts like a HPF.
 
1 1
1 2 1 21 1
e e cor
c e
cor
j
v j rC C
v j r C C C C j

 


 
  
    
       
 
L pR
1C
2C er
ev
cv
Frequency
1
1 2
C
C C
Gain
0A
cor
1
0
1 2
C
A
C C

  1 2
1
cor
er C C
 

1
tan
2 c
 


  
   
 
If the frequency of operation is well
above the corner frequency ωcor , the
gain is given by the capacitor ratio
and the phase shift is zero.
90o
0o
Phase
Frequency
cor

Capacitor Ratios with Colpitts (II)
• re is transformed to a higher value through the capacitor
divider, which effectively prevents this low impedance from
reducing the Q of the LC resonator.
• The resulting transformed circuit as seen by the tank
2
2
,tank
1
1e e
C
r r
C
 
  
 
L pR
1C
2C
,tanker
cv
1 2
1 2
1
T
C C
LC CLC


 
(make C2 large and C1 small to get the maximum
effect of the impedance transformation)

Negative Resistance
• Negative resistance of CB Colpitts oscillator
• Input impedance:
• A necessary condition for oscillation:
This is just a negative resistor in series with the two capacitors.
where rs is the equivalent series resistance on the resonator.
2i m
e
v
i g v j C v
r

 

   
1
m
e
g
r
;
2
ii
v
j C


 
1
i m
ce
i g v
v
j C




1 2
1 m i
ce i
g i
v i
j C j C 
  
   
  
2
1 2 1 2
1 1i ce m
i
i i
v v v g
Z
i i j C j C C C

  
 
    
2
1 2
m
s
g
r
C C

ii
iv
cev
v

er
1C
2C
mg v







where , , and
17/60

Negative Resistance for Series/Parallel Circuits
• Since the resonance is actually a parallel one, the series
components need to be converted back to parallel ones.
• However, if the equivalent Q of the RC circuit is high, the
parallel capacitor Cp will be approximately equal to the series
capacitor Cs , and the above analysis is valid. Even for low Q,
these simple equations are useful for quick calculations.
2C
1C
LpR
er
xv
m xg v
cv sr negR
L TC

Example
• Assume L = 10 nH, Rp = 300 Ω, C1 = 2.5 pF, C2 = 10 pF, and
the transistor is operating at 1 mA, or re = 25 Ω and gm = 0.04.
Using negative resistance, determine the oscillator resonant
frequency and apparent frequency shift.
( |negative resistance| > original resistance,
the oscillator should start up successfully)
This is a frequency of 1.2353 GHz, which is close
to a 10% change in frequency. Further refinement
should come from a simulator.
1 1
7.07 Grad/s
10 nH 2 pFTLC
   

2 pFTC 
 
22
1 2
0.04
32
7.07107 Grad/s 2 pF 10 pF
m
s
g
r
C C

     
 
1 1
2.2097
7.07107 Grad/s 32 2 pFs T
Q
r C

    
 
   2 2
par 1 32 1 2.2097 188sr r Q       
0 1.1254 GHzf 
 par 2
2
2 pF
1.66 pF
1 1 1 2.20971
sC
C
Q
  

par
1 1
7.7615 Grad/s
10 nH 1.66 pFLC
   


Negative Resistance of Gm Oscillator
• Assume that both transistors are biased identically, then gm1 =
gm2 , re1 = re2 , vπ1 = vπ2 , and solve for Zi = vi /ii .
• Input impedance:
• Necessary condition for oscillation:
where Rp is the equivalent parallel resistance of the resonator.
1 1 2 2
1 2
i
i m m
e e
v
i g v g v
r r
   

2
i
m
Z
g


2
m
p
g
R

ii
iv 1 1mg v


1v
2v
1er
2er
2 2mg v





Minimum Current for Oscillation (I)
• Using a 5-nH inductor with Q = 5 and assuming no other
loading on the resonator, determine the minimum current
required to start the oscillations of 3 GHz if a Colpitts
oscillator is used or if a –Gm oscillator is used.
 To find the minimum current, we find the maximum rneg by taking the
derivative with respect to C1.
The maximum obtainable negative resistance is achieved when the two capacitors
are equal in value, C1 = C2 = 1.1258 pF, and twice the Ctot.
 
22
1 1
562.9 fF
2 3 GHz 5 nH
tot
osc
C
L 
  
 
1 2
1 2
tot
C C
C
C C


1
2
1
tot
tot
C C
C
C C


neg 2 2 2 2
1 2 1 1
m m m
tot
g g g
r
C C C C C  
  
neg
2 2 2 3
1 1 1
2
0m m
tot
dr g g
dC C C C 

   1 2 totC C

Minimum Current for Oscillation (II)
 Now the loss in the resonator at 3 GHz is due to the finite Q of the inductor.
The series resistance of the inductor is
 Therefore, rneg = rs = 18.85 Ω. Noting that gm = Ic /vT ,
 In Gm oscillator, there is no capacitor ratio to consider. The parallel
resistance of the inductor is
A Gm oscillator can start with half as much collector current in each transistor as a Colpitts
oscillator under the same loading conditions.
 2 3 GHz 5 nH
18.85
5
s
L
r
Q
  
   
   
2 22
1 2 neg 2 3 GHz 1.1258 pF 25 mV 18.85 212.2 AC TI C C v r         
 2 3 GHz 5 nH 5=471.2pR LQ      
m C Tg I v 2 2 25 mV 471.2 106.1 AC T pI v R     

Basic Differential Topologies
• Take two single-ended oscillators and place them back to back.
1C
1C
CCV
1Q 2Q
2 2C
biasV
biasI biasI
L
CCV CCV
1Q 2Q
L
1C 1C
2 2C
biasI biasI
CCV
L
C
1Q 2Q
biasI
Copitts CB Copitts CC Gm

Modified CC Colpitts with Buffering
• Oscillators are usually buffered (use emitter follower) in order
to drive a low impedance. Any load that is a significant
fraction of the Rp of the oscillator would lower the output
swing and increase the phase noise.
• CC oscillator is modified slightly by
placing resistors in the collector.
The output is then taken from the
collector. Since this is a high-
impedance node, the resonator is
isolated from the load. However, the
addition of these resistors will also
reduce the headroom available to
the oscillator.
CCV
L
1C
1Q 2Q
biasI
1C
biasI
2 2C
LR LR
CCV

Several Refinements to the Gm Topology (I)
• Decouple the base from the collector with
capacitors to get larger swings.
• The bases have to be biased separately.
• Rbias have to be made large to prevent loss
of signal at the base. However, these
resistors can be a substantial source of
noise.
biasV
L
C
1Q
2Q
biasI
biasR
CCV
biasV
biasR
cpC cpC

Several Refinements to the Gm Topology (II)
1Q 2Q
pL
CCV
C
sL
biasV
biasI
• Use a transformer to decouple the collectors
from the bases.
• Since the bias can be applied through the
center tap, no need for the RF blocking.
• A turns ratio of greater than unity is chosen,
there is the added advantage that the swing
on the base can be much smaller than the
swing on the collector to prevent transistor
saturation.

Several Refinements to the Gm Topology (III)
• Since the tail resistor is not a high
impedance source, the bias current will
vary dynamically over the cycle of the
oscillation (highest when voltage peaks
and lowest during the zero crossings).
• Since the oscillator is most sensitive to
phase noise during the zero crossings, this
oscillator can often give very good phase
noise performance.biasV
L
C
1Q
2Q
tailR
biasR
CCV
biasV
biasR
cpC cpC
Time
Amplitude
 1ci t  2ci t
AVEI
dcI
27/60

• Using a noise filter in the tail can lead to a
very low-noise bias, thus low-phase-noise
designs.
• Another advantage is that, before startup,
the transistor Q3 can be biased in
saturation, because during startup the 2nd
harmonic will cause a dc bias shift at Q3
collector, pulling it out of saturation and
into the active region.
• Since 2nd harmonic cannot pass through
Ltail, there is no ‘‘ringing’’ at Q3 collector,
further reducing its headroom requirement.
biasV
C
1Q
2Q
tailL
biasR
CCV
biasV
biasR
cpC cpC
3Q tailCbiasV
Several Refinements to the Gm Topology (IV)
28/60

The Effect of Parasitics on the Frequency
• The first task in designing an oscillator is to set the frequency
of oscillation and hence set the value of the total inductance
and capacitance in the circuit.
• To increase output swing, it is usually desirable to make the
inductance as large as possible (this will also make the
oscillator less sensitive to parasitic resistance). However, it
should be noted that large monolithic inductors suffer from
limited Q. In addition, as the capacitors become smaller, their
value will be more sensitive to parasitics.

Oscillating Frequency Summary
1 2 1
1 2
1
osc
C C C C
L C
C C C



 
 
 
  
1 2 2
1 2
1
osc
C C C C
L C
C C C



 
 
 
  
1
2
2
osc
C
L C C

 
 
  
 
1C
2C
L
1Q
1C
2C
1Q
L
CB CC Gm
C L
1Q
2Q
30/60

• VCO is an oscillator of which frequency is controlled by a
tuning voltage.
• VCO is a simple frequency modulator
Voltage Controlled Oscillator (VCO)
Vtunevcof
tuneV
tuneV
tuneV
 oscs t
31/60

Making the Oscillator Tunable
• Varactors in a bipolar process can be realized using either the
base-collector or the base-emitter junctions or else using a
MOS varactor in BiCMOS processes.
CCV
CCV
LL
1C 1Q 2Q 1C
R
1BR
2BR
varC varC
CB
Subs
CB
Subs
Tuning port
CCV
biasI
L
varCvarC
conR
conV
1Q 2Q
CCV
biasI
L
varCvarC
LR
conV
1Q 2Q
CCV
LR
1C
biasI
1C

• Frequency Range
• Frequency tuning characteristics
 Tuning sensitivity (Hz/V) :
 Linearity
VCO Sensitivity and Tuning Linearity
VK f V  
vcof
tuneV
,0tV
0f
maxf
minf
,mintV ,maxtV
v
f
Ideal (perfect)
Piecewise good
Piecewise good
Poor

Important Figures
• Output power (50 Ohm)
• Frequency stability: frequency drifting
• Source pushing and load pulling figures
• Harmonics
• Phase noise in frequency-domain (or Jitter in time-domain)

Small-Signal Simulation – Use OscTest
• It is based on small-signal operation and only for prediction.
• When the simulation fails, try to reduce the value of Z in the
OscTest simulator (do not use 1 or 0, there would be a problem of
convergence, lower impedance values usually seem to work better) or to
reverse OscTest direction.
Resonator Active Part
(include load network)
Varactor: Voltage-controlled capacitor
OscTest
OscTest is a controller base on S-parameter
simulation to determine if the circuit oscillates.

Small-Signal Simulation – Oscillating Condition
• S(1,1) is the predicted loop gain by OscTest. By showing it on
the polar plot, you can find that the locus goes beyond the
circle with radius of 1 (S11>1)。
• The blue point is (1+j0), which is the target when steady
oscillating occurs.
-1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2-1.4 1.4
freq (500.0MHz to 4.000GHz)
S(1,1)
m1
m1
freq=
S(1,1)=1.172 / 0.975
1.410GHz
Setup the dataset named: osc_test, and data
display named: osc_basics.
Show S(1,1) on
a Polar-plot
When the x-axis value of
1.0 is circled by the
trace(because S11 > 1), it
means that the circuit
oscillates. This is the
purpose of the OscTest
component.
S11 > 1

Small-Signal Simulation – Oscillating Condition
• S(1,1)=1 25o @885 MHz
• S(1,1) > 1 above 885 MHz
• S(1,1)=1.1 0o @1.445 GHz
• S(1,1)=1.08 6.6o @1.8 GHz (Target frequency)
1.0 1.5 2.0 2.5 3.0 3.50.5 4.0
-20
-10
0
10
20
30
-30
40
freq, GHz
phase(S(1,1))
m4
m5
m4
f req=
phase(S(1,1))=0.005
1.445GHz
m5
f req=
phase(S(1,1))=-6.604
1.795GHz
1.0 1.5 2.0 2.5 3.0 3.50.5 4.0
0.6
0.7
0.8
0.9
1.0
1.1
1.2
0.5
1.3
freq, GHz
mag(S(1,1))
m2 m3
m2
f req=
mag(S(1,1))=1.013
885.0MHz
m3
f req=
mag(S(1,1))=1.009
3.982GHz
Around 1.8 GHz (Marker m5), the phase is not 0o, but this is OK at
this time. The harmonic-balance simulation will be performed later.
S11 > 1 above 880 MHz
The device is unstable and
has a chance to oscillate.
Possibility!

Large-Signal Simulation
• Harmonic-Balance (HB)
OscPort
Enable the oscillation analysis
with “Use Oscport” method.
Oscport HB simulation
attempts to find the correct
oscillating frequency using
loop gain and current
(Barkhausen’s Criteria).
3

Setting Up the Parameters
• V:
It is the initial voltage. Leave it when you don’t know the initial value, the HB will
run AC analysis to get that.
• NumOctaves:
The frequency range in which you want to search the oscillating condition. In this
example, Freq[1] = 1 GHz, and NumOctaves = 2. This means that the frequency
range is from 0.5 GHz to 2 GHz.
• Steps:
It is set to show how many points you like to run the prediction with in the range per
octave. In this example, the simulator use 20 points from 0.5 GHz to 2 GHz to search
the condition. If you are designing a high-Q oscillator, the step should be finer.
• FundIndex
FundIndex = 1 means that the oscillating frequency Freq[1] is found by OscPort
prediction, but not as your preset Freq[1] = 1.0 GHz. You should tell OscPort that
which frequency to predict by using the frequency index.
• HB Simulator Settings:
Usually, we use the order of harmonicas equal to 3, 7, 15, 31. When the DC term is
included, the data memory would be 4, 8, 16, 32, that are numbers by 2N and they are
good for computation. StatusLevel = 3 to show more information when the
simulation is running, and turn on the OscMode while the OscPortName is assigned
to the OscPort name.

Simulated Results
• freq[1] is the oscillating frequency, which is 1.806 GHz in the
example.
• Use dBm( ) function to show the spectral values.
Eqn loop_current=real(ICC.i[0])
Eqn osc_freq=freq[1]
loop_current
-0.011
osc_freq
1.806E9
m6
harmindex=
dBm(Vout)=7.318
1
1 2 3 4 5 60 7
-30
-20
-10
0
-40
10
harmindex
dBm(Vout)
m6
m6
harmindex=
dBm(Vout)=7.318
1
harmindex
0
1
2
3
4
5
6
7
freq
0.0000 Hz
1.806 GHz
3.611 GHz
5.417 GHz
7.222 GHz
9.028 GHz
10.83 GHz
12.64 GHz
harm_power
<invalid>
7.318
-2.208
-17.501
-17.061
-27.317
-27.815
-35.340
Eqn harm_power=dBm(Vout[0::1::7])
2 4 6 8 10 120 14
-30
-20
-10
0
-40
10
freq, GHz
dBm(Vout)
Fundamental Frequency (oscillation frequency)
Use dBm( ) to show the signal power
(Note: x-axis is “harmonic index”)
Use plot_vs( )to show the signal
power versus frequency.
(Note: x-axis is now “frequency”)

Simulated Results
• Use ts( ) function to show the time-domain waveform
-600
-500
-400
-300
-700
-200
ts(Vres),mV
-400
-200
0
200
-600
400
ts(VB),mV
-600
-500
-400
-300
-700
-200
ts(VE),mV
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.10.0 1.2
-0.5
0.0
0.5
-1.0
1.0
time, nsec
ts(Vout),V

Tuning Sensitivity: KV (MHz/V)
Pass the variable “Tune_Step” to dataset Plot oscillating frequency v.s. Tuning voltage
“Osc1”
We already know the
oscillating frequency
is around 1.8 GHz by
previous simulation.
Sweep the tuning
voltage of the varactor.
42/60

Simulated Results
Eqn osc_freq=freq[1]
m7
indep(m7)=
plot_vs(freq[1], Vtune)=1.806E9
4.000
m8
indep(m8)=
6.500
1 2 3 4 5 6 7 8 90 10
1.75
1.80
1.85
1.90
1.95
2.00
2.05
1.70
2.10
Vtune
freq[1],GHz
m7
m8
m7
indep(m7)=
4.000
m8
indep(m8)=
6.500
Eqn Tuning_Sensitivity=diff(freq[1])/Tune_Step[0]
1 2 3 4 5 6 7 8 90 10
6.0E7
8.0E7
1.0E8
1.2E8
1.4E8
1.6E8
4.0E7
1.8E8
Vtune
Tuning_Sensitivity
1.75E9 1.80E9 1.85E9 1.90E9 1.95E9 2.00E91.70E9 2.05E9
6.0E7
8.0E7
1.0E8
1.2E8
1.4E8
1.6E8
4.0E7
1.8E8
osc_freq[0::1::(tune_pts-1)]
Tuning_Sensitivity
Eqn f_pts=sweep_size(osc_freq)
f_pts
41
tune_pts
40
Eqn tune_pts=sweep_size(Tuning_Sensitivity)
Eqn Tuning_Sensitivity_band=(m8-m7)/(indep(m8)-indep(m7))
Tuning_Sensitivity_band
3.904E7
m7
1.806E9
m8
1.903E9
Oscillating frequency v.s. Tuning voltage Calculate tuning sensitivity from
makers m7 and m8
Calculate sensitivity by using
diff() function.
Note: Since no “padding” with diff(),
there will be 1 point less than freq[1]
points.
Sensitivity v.s. Vtune Sensitivity v.s. Frequency

Over-tuned : Varactor Breakdown
2 4 6 8 10 12 14 160 18
1.7
1.8
1.9
2.0
2.1
1.6
2.2
Vtune
freq[1],GHz
m7
m8
m7
indep(m7)=
4.000
m8
indep(m8)=
12.000Sweep Vtune up to 18 V
The diode is breakdown
above 12 V (acts like a
resistor), it no longer acts
like a variable capacitor.
Diode = Varactor
Maximum oscillating
frequency is 2.13 GHz

Source Pushing
• Sweep the supply voltage from 5 V to 20 V with a 0.25 V step.
Change the supply voltage
to a variable “Vbias”Sweep the supply voltage “Vbias” from 5 V to
20 V while Vtune is now held constantly at 4 V.
(In practice, Vtune is set to a voltage that oscillator oscillates
at “target” center frequency.)

Simulated Results
• Source pushing is measured around the normal supply voltage
(12 V in this example). This example shows that the source
pushing figure is 21.77 MHz/V.
m9
indep(m9)=
plot_vs(freq[1], Vbias)=1.825E9
13.000
m10
indep(m10)=
11.000
6 8 10 12 14 16 184 20
0.5
1.0
1.5
0.0
2.0
Vbias
freq[1],GHz
m9m10
m9
indep(m9)=
13.000
m10
indep(m10)=
11.000
Eqn Source_pushing=(m9-m10)/(indep(m9)-indep(m10))
Source_pushing
2.177E7
Plot freq[1] v.s. Vbias to
show the source pushing
results. Here, use makers
and equations to calculate
the pushing figure around
Vbias = 12 V. As we can see,
this oscillator has the source
pushing figure equals to
21.77 MHz/V.

Load Pulling
• Frequency pulling figure illustrates the frequency variation
subjection to load impedance (normally, the load impedance is
50 Ω).
vw1: VSWR sweep start
vw2: VSWR sweep stop
nvw: num. of VSWR sweep
real part of load
sweep load
Image part of load
Sweep load for different constant VSWR circles in Smith chart.
Save these variables in dataset

Load Pulling
• Move maker m12 to change VSWR. In this example, we
observed the frequency variation corresponding to VSWR=1.2.
• Find the peak frequency change to 1.806 GHz (df_peak).
m12
indep(m12)=
vs([0::sweep_size(VSWRval)-1],VSWRval)=2.000
1.200
Eqn refl=rload+j*iload
Eqn vswr_k=(nvw[0,0]-1)*(indep(m12)-vw1[0,0])/(vw2[0,0]-vw1[0,0])
Eqn VSWR=vswr_k*(vw2[0,0]-vw1[0,0])/(nvw[0,0]-1)+(vw1[0,0])
Eqn LoadRefl=mag(refl[::,1])
Eqn df_peak=max(abs(freq[vswr_k,::,1]-1.806e9))
df_peak
3.202E7
Load Pulling Figure @ VSRW=1.200
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.91.0 2.0
VSWR
m12
m12
indep(m12)=
1.200
Write down these equations for load pulling figure measurement
@certain VSWR value. (You can change VSWR by scrolling marker m12)
Find peak frequency that deviates
from center frequency 1.086 GHz.
phi (0.000 to 2.000)
refl[vswr_k,::]
m12
indep(m12)=
1.200
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.91.0 2.0
VSWR
m12
m12
indep(m12)=
1.200
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.80.0 2.0
0.65
0.70
0.75
0.80
0.85
0.60
0.90
phi ( *pi radians)
mag(Vout[vswr_k,::,1])
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.80.0 2.0
1.7800G
1.7900G
1.8000G
1.8100G
1.8200G
1.8300G
1.7700G
1.8400G
phi ( *pi radians)
freq[vswr_k,::,1],Hz
Eqn refl=rload+j*iload
Eqn vswr_k=(nvw[0,0]-1)*(indep(m12)-vw1[0,0])/(vw2[0,0]-vw1[0,0])
Eqn VSWR=vswr_k*(vw2[0,0]-vw1[0,0])/(nvw[0,0]-1)+(vw1[0,0])
Eqn LoadRefl=mag(refl[::,1])
Frequency variation for VSWR = 1.20
Eqn df_peak=max(abs(freq[vswr_k,::,1]-1.806e9))
df_peak
3.202E7
Load Pulling Figure @ VSRW=1.200
Constant VSWR circleVout amplitude variations
Frequency variations
use @VSWR in the text to show the number

Oscillator Phase Noise
 V f
f
1f
 V f
f
1f
 v t
t
1
1
f
 v t
t
1
1
f
Time Domain Frequency Domain
Jitter Phase noise
mf
49/60

Phase Disturbance Due to Thermal Noise (I)
• Modeling the noise with the phasor diagram
nP
sP
sP
Phase disturbance
Amplitude disturbance
FkTB
avsP
Noise-free amplifier
f
0f 0 mf f
1 Hz1 Hz 1nRMS
FkT
V
R
2nRMS
FkT
V
R

avs
avsRMS
P
V
R

The input phase noise in a 1-Hz bandwidth at any frequency
from the carrier produces a phase deviation.
0 mf f
Phasor Diagram
50/60
avsRMS avsV P R
FkTR FkTR

(noise from )
Phase Disturbance Due to Thermal Noise (II)
• RMS phase deviation
avsavsRMS
nRMS
peak
P
FkT
V
V
 1

avs
RMS
P
FkT
2
1
1 
avs
RMS
P
FkT
2
1
2 
2 2
1 2RMS total RMS RMS
avs
FkT
P
       
m
12 nRMSV
2 avsRMSV
peak
(total phase deviation)
   02 cososc avsRMSv t V t t     
mf(noise from )
mf
51/60

Lesson’s Phase Noise Model (I)
• The spectral density of phase noise :
 Due to Thermal Noise
 Consider Flicker Noise (modeled)
  2
m RMS
avs
FkTB
S f
P
   
1)(BdBm/Hz174 kTB
(due to theoretical noise floor of the amplifier)
1)(B1)( 






m
c
avs
m
f
f
P
FkTB
fS
noise floor flicker noise
 mS f
Noise-free amplifierPhase modulator
avs
FkTB
P
mf
 mS f
cf

Lesson’s Phase Noise Model (II)
• The oscillator may be modeled as an amplifier with feedback
 
0
1
2
1
m
L m
L
Q
j




 
  
 
22
0 B
QL


   0
1
2
out m in m
L m
f f
j Q

 

 
    
 
   
2
0
,2
1
1
2
out m in m
m L
f
S f S f
f Q
 
  
    
   
 , 1 c
in m
avs m
FkTB f
S f
P f

 
   
 
Noise-free
amplifier
Phase
modulator
 ,in mS f
Output
Feedback

Resonator
Resonator
equivalent low-pass
 in mf
 0
2
in m
L m
f
j Q



 
 
 
 out mf
 mL 
53/60

Lesson’s Phase Noise Model (III)
• Lesson’s phase noise model:
 
2
0
2
1 1
1 ( )
2 2
m in m
m L
f
L f S f
f Q

  
    
   
 , 1 c
in m
avs m
FkTB f
S f
P f

 
   
 
Open-loop
Closed-loop w/ Resonator
 
2
2
3 2 2
1 1
1
2 4 2
o c o c
m
avs m L m l m
FkTB f f f f
L f
P f Q f Q f
  
     
   
Up-convert 1/f noise
Thermal FM noise
Flicker noise
Thermal noise floor
54/60

Lesson’s Phase Noise Model (IV)
Low-Q oscillator
Phase perturbation
1
mf 
3
mf 
2
mf 
0
mf
mf
mf
Resulting phase noise
cf
cf 0 2f Q
High-Q oscillator
Phase perturbation
1
mf 
1
mf 
0
mf
3
mf 
mf
mf
Resulting phase noise
cf
cf0 2f Q
55/60

Design Example of Phase Noise Limits (I)
• A 5-GHz receiver including an onchip phase-locked loop (PLL)
is argued to be implemented with the VCO requirements:
1.8V supply, <1 mW DC power, and phase noise performance of 105 dBc/Hz at
100-kHz offset. It is known that, in the technology to be used, the best inductor Q is
15 for a 3-nH device. Assume that capacitors or varactors will have a Q of 50.
 Assume a Gm topology will be used:
2 5 GHz 3 nH 15 1413.7pr L LQ       Parallel resistance due to the inductor:
Required capacitance:
 
22
1 1
337.7 fF
2 5 GHz 3 nH
tot
osc
C
L 
  
 
 
50
4712.9
2 5 GHz 337.7 fF
p
tot
Q
r C
C 
   
 
Parallel resistance due to the capacitor:
Equivalent parallel resistance of the resonator is 1087.5 Ω
Current limit: 1.8-V VCC and PDC< 1 mW: 555.5 μA
Peak voltage swing:  tank
2 2
555.5 A 1087.5 0.384 Vbias pV I R 
 
    

Design Example of Phase Noise Limits (II)
 Assume all low-frequency upconverted noise is small and active devices
add no noise to the circuit (F=1), we can now estimate the phase noise.
 
 
22
tank
0.384 V
67.8 W
2 2 1087.5
RF
p
V
P
R
  

337.7 fF
1087.5 11.53
3 nH
tot
p
C
Q R
L
   
This is 98.5 dBc/Hz at 100-kHz offset, which is 6.5 dB higher than the promised
performance. Thus, the specifications given to the customer are most likely very
difficult. This is an example of one of the most important principles in engineering.
RF output power:
Oscillator Q:
 
 
 
22
2 23
10
3 2 2
1.12 2 5 GHz1 1 1 1.38 10 J/K 298 K 1 Hz
100 kHz 1 1.79 10
2 4 2 2 67.8 W 2 11.53 2 100 kHz
o c o c
avs m L m l m
FkTB f f f f
L
P f Q f Q f

 


         
             
         
~ 0 far from carrier
97.5 dBc Hz@100 kHz offset 
dominant around carrier
57/60
1.4
1
98.5

Phase Noise Simulation
Phase Noise Simulation Setup
7 in most cases
Agilent suggests
Turn on
Turn on
Set noise (1) Set noise (2)

Simulated Results
• Plot the “pnmx” (PSD of the phase noise).
• This oscillator has the phase noise = 78.39 dBc/Hz@10 kHz,
98.34 dBc/Hz@100 kHz, and 118.08 dBc/Hz@1 MHz。
m11
noisefreq=
pnmx=-78.390
10.00kHz
m13
noisefreq=
pnmx=-98.340
100.0kHz
m14
noisefreq=
pnmx=-118.079
1.000MHz
1E1 1E2 1E3 1E4 1E5 1E61 1E7
-120
-100
-80
-60
-40
-20
0
-140
20
noisefreq, Hz
pnmx,dBc
m11
m13
m14
m11
noisefreq=
pnmx=-78.390
10.00kHz
m13
noisefreq=
pnmx=-98.340
100.0kHz
m14
noisefreq=
pnmx=-118.079
1.000MHz

Summary
• In this presentation, few kinds of popular oscillator topologies
were introduced. The CB and CC configurations are good for
high frequency operation while the CE is good for high power
application and has good buffering characteristics.
• The active device is configured as feedback loop to provide a
negative resistance for resonator.
• For a voltage-controlled frequency application, an oscillator is
usually designed with variable capacitors, or varactors, to
provide frequency-tuning capability.
• Lesson’s phase noise model gives an intuitive way to
understand the behavior of the phase noise generated from the
oscillator.

Voltage Controlled Oscillator Design - Short Course at NKFUST, 2013

More Related Content

What's hot (20)

Viewers also liked (7)

Similar to Voltage Controlled Oscillator Design - Short Course at NKFUST, 2013 (20)

More from Simen Li (12)

Recently uploaded (20)

Voltage Controlled Oscillator Design - Short Course at NKFUST, 2013