Note: The buffer amplifier is untuned.

Thus, we need to avoid reduced conduction angle

modes such as Class AB or Class B. There is no low impedance termination for even-
order harmonic frequencies. Harmonic termination is needed to form the half sine wave
current waveform for Class B. A push-pull design with appropriate transformer could be
used if Class B is desired.

We must work within the limitations of the device chosen for the buffer.

IC,max
BV,CEO
Tmax

These can typically be found on the data sheet. A safe design does not run all the way to
the limits of the device, so a reasonable margin needs to be provided. It may not be wise
to exceed 75% of the recommended maximum values.

For the Class A buffer amplifier, the peak values of voltage and current are:

VCE, max = VCC + ICQ RL

IC, max = 2 ICQ

Let’s design a buffer.

Specs: 7 dBm output power (5 mW; 0.71V in 50 ohms); Source impedance = 100 ohms.
Voltage output from oscillator = 0.5V.

The device has: BVCEO = 12V; IC, max = 100 mA

1. Determine ICQ.

We start with RL = 50 ohms.

Po = ICQ2 RL/2 = 0.005 W ICQ = 14.2 mA.

This gives us a maximum current of 28.4 mA, well within the limits of the device.

2. Determine RE.

This will be determined by the gain required.

− RL
Av = 0.71/0.5 = 1.42= − g m RL =
re + RE
 gm = 0.55; re = 1.8 ohms. So, RE = (RL/Av) – re = 33 ohms.

Check the input resistance: RB||RE (β+1) >> RS. Make sure the bias resistors (RB) don’t
load down the oscillator or reduce the input amplitude significantly.

3. Determine VCC,min, VCC,max. This is the range of acceptable power supply voltages
which avoid clipping and breakdown.

VCC, min = VCE,sat + ICQ (RL + 2RE) = 2.1 V

VCC, max = 0.75 BV,CEO – ICQ RL = 8.3 V

4. Check for temperature rise

Worst-case Power Dissipation = PD = VCC ICQ for Class A

Choose VCC = 5V. PD = 71 mW. (This will drop to approximately 66 mW at full

output, but you must design for the worst case where the oscillator may fail to produce an
output.)

Tmax = RTH PD + Tambient

where RTH is the Thermal Resistance (degree C/Watt)

Maximum temperature is usually 150 C. If you exceed this, then you must reduce VCC or
ICQ.

5. Stability. Finally, you must check for stability over a wide range of frequencies. Use
the S parameter simulation mode in ADS and plot k and mag delta vs frequency. If there
are potential instabilities, you must deal with them using techniques that were employed
with the power amplifier designs last quarter.
Resonator Design for Low Phase Noise Oscillators

Phase noise can be estimated by a simplified version of Leeson’s equation:

 2kTF  ω 
2
10 log  
0
  (1)
 Psig  2Qu ∆ω  
 

This is a noise power (single sideband) to carrier power ratio normalized to a 1 Hz

bandwidth. Units are dBc/Hz.

We saw that Vres = 2 IBIAS Rp is an approximation of the voltage across the resonator for
the Colpitts oscillator.

Qu = Rp/Xp = Bp/Gp for a simple parallel LRC resonant circuit.

So, how can we modify the circuit to improve the noise to carrier ratio? From the
equation (1), increasing Psig and Qu would help. The former increases carrier power,
whereas the latter reduces noise generated by losses in the resonator and also does a
better job of bandpass filtering the noise. We increase signal power by increasing voltage
or current.

2 2
Psig α Vres or I BIAS or R 2p

Inductor:
Let’s assume that the inductor limits the Qu. This is typically true at frequencies below
10 GHz or so. How can we increase Qu or Rp at a given frequency?

Qu = ωL/Rs for a series LR.

S. Long January 19, 2005 1

For a coil inductor (whether on a magnetic core or in air) with N turns, we can assume
that the series resistance increases in proportion with length of the wire, therefore with N.
The inductance will increase in proportion to N2. Thus, Qu increases in proportion to N.

X 2 ω 2 L2
What about Rp? R p = = and thus should increase as N3.
Rs Rs

This is an ideal version and in reality may not scale as favorably. Skin effect will cause
Rs to increase with frequency for example. Also, if we consider a spiral inductor on an
RFIC, the inner turns have less area than the outer ones. They contribute less inductance
per turn than a solenoidal coil, so the scaling is not obvious and requires E/M simulation
to determine Qu and Rp.

Another consideration is what Q definition should be used? There are several:

Q = ω L / Rs

ω × Energy Stored
Q= (2abc)
Power Dissipated

dθ
Q = − ω0
dω ω = ω
0

Analysis of oscillators has shown that the last definition, the phase slope with frequency,
correlates best with phase noise performance. While these definitions may predict the
same Q for simple resonators (series or parallel RLC), they do not agree very well with
more complicated topologies or with transmission line resonators.

How can we improve Psig? Any increase in IBIAS or Vres will help, but our device may
begin to clip or breakdown. There are limits to how much voltage and current a device
can handle safely. Simply increasing Rp will increase Vres, but may reach these limits in
the device. So, a better approach is to modify the topology so that more power can be put
into the resonator without exceeding the limits of the transistor. There are many such
attempts in the literature of oscillators to do this, but two approaches will be briefly
reviewed here.

1. Tapped inductor.
2. Clapp oscillator.

S. Long January 19, 2005 2

1. Tapped Inductor Oscillator.
Vres

N
VC

C1 1

C2

IBIAS

Here we see a modified version of the Colpitts. The tapped capacitor provides the
feedback as before, but now there is a tapped inductor, an autotransformer with turns
ratio N.
• This has the effect of reducing the collector voltage, VC = Vres/(N+1).
• It also reduces the load resistance at the collector. RL = Rp/(N+1)2.

If the tapping ratio was 1:1, then VC = Vres/2 and RL = Rp/4. Thus, we can now increase
IBIAS by a factor of 4, increasing the Psig by 16 times, without increasing the voltage
swing at the collector. Of course, the noise contribution of the transistor generally also
increases with IBIAS, so the net result is a 4 times improvement in phase noise to carrier
ratio (6 dB). Scaling to higher N values may be beneficial, but it depends on how Rp
scales with N. This can only be determined accurately by measurement or E/M
simulation of the inductor.

S. Long January 19, 2005 3

2. Clapp Oscillator Z(jω) VCC

C1
L
RE

C2 IBIAS
RS

CC

In this case, the resonator is modified by adding a capacitor CC in series with the inductor
L. This allows the inductance to be increased (Qu, Rp higher) from its Colpitts value, L0.
To keep the resonant frequency the same, the imaginary part of the impedance of the
series LRC resonator must be the same as that of L0. If we increase the inductance by a
factor P1,

j (ω L0 P − 1/ ωCC ) = jω L0 . (3)

From this, we can determine CC:

1
CC = . (4)
ω 2 L0 ( P − 1)
To determine what effect the series RLC resonator has on the load presented to the
collector, RL, we must first find Z(jω), then Y(jω), then determine RL from 1/Re{Y(jω)}.

Define Qu = ω L / R s = ω PL0 / R s (5)

1
W. Hayward, Introduction to Radio Frequency Design, Ch. 7, American Radio Relay League, 1994.

S. Long January 19, 2005 4

Qu is the unloaded Q and R s corresponds to the series resistance of the Clapp inductor,
L = PL0.

Eliminating CC, we get for Z:

ω PL0 ω L0 ( P + jQu )
Z ( jω ) = + j (ω PL0 − 1/ ωCC ) = (6)
Q u Q u

Y = 1/Z, so
1  P Q u  
RL = = ω L0  +  = ω L0Qu / P (7)
Re (Y ( jω ) ) 
 Qu P 
because P/Qu << 1.

To determine what effect this has on the oscillator, we can use the assumption above
about the how series resistance and inductance scale with number of turns2. If we want
an inductor P times larger than the Colpitts inductor, we need to increase the number of
turns by a factor of P . Thus,

L = PL0 R s = P Rs . (8)

ω PL0
Then, Qu = = Qu ( Colpitts ) P . (9)
P Rs

First observation: The effective unloaded Q of the Clapp resonator is higher than the
Colpitts when the scaling of resistance is included in the calculation. It can be shown that
the phase slope also increases by the same amount.

What about RL? Equation (7) described RL in terms of the unloaded Q of the Clapp
inductor, L. But, if we are comparing with the Colpitts, we should refer RL to Qu of L0,
not L. For the Colpitts, the parallel equivalent resistance is Rp.

R p = X 2 / Rs = ω L0Qu
RL = ω L0Qu / P = R p / P

2
It is important to remember that all inductors might not scale in this fashion.

S. Long January 19, 2005 5

Observation #2: The loading that the collector sees with the Clapp resonator is reduced
by a factor of P below the Colpitts. Therefore, IBIAS can be increased by a factor of
P while maintaining the same collector voltage.
Observation #3: How much more energy is stored in the Clapp resonator than in the
Colpitts?

Vres is increased by the presence of a third capacitor, CC, in series with C1 and C2 3.
Z(jω) VCC

C1
L
RE

C2 IBIAS
RS
Vres

CC

Energy = ½ CTOTAL V2res

Let the series capacitance of C1 and C2 be represented by C12.

C1C2
C12 =
C1 + C2
Then,
C12CC C1C2CC
CTOTAL = =
C12 + CC C1C2 + C2CC + C1CC

C C 
Vres = V2  12+ C 
 CC 

3
T. H. Lee, Design of CMOS Radio Frequency Integrated Circuits, Cambridge, 1998.

S. Long January 19, 2005 6

Then,
1  C + CC  2
Energy = C12  12  V2
2  CC 

If we then increase IBIAS by a factor of P , V2 remains the same as in the unmodified

Colpitts case, but the energy has increased by the factor

 C12 + CC 
 
 CC 
as does the Psig.

S. Long January 19, 2005 7