Project Oscillator1
Project Oscillator1
2. DESIGN DESCRIPTIONS
2.1. RESONATOR DESIGN [1, 2]
The first part of the design process is to determine the required unloaded Q of the
resonator to achieve the phase performance of <-100dB/Hz at 10 KHz. The simulation
shown in Fig.1 simulates the phase noise of an oscillator for a given loaded Q, Noise
figure and frequency. The resulting simulation plot for a VCO at 200MHz with a loaded
Q of 30 is shown in Fig. 2.
1
Radar Subsystem Project
Now that we have design criteria for the resonator loaded Q (in this case > 30) we could
decide on the topography of the resonator.
Fig.2: Resulting simulation of the ADS schematic shown in Fig. 1, showing a predicted
phase noise of -94dBc/Hz at 10 KHz for a loaded Q of > 30.
The unloaded/loaded Q of the resonator depends on the values of the L/C ratio of the tank
circuit and on the unloaded Q’s of the individual components. Typically, ceramic
microwave capacitors such as the ATC100a range (by American technical ceramics) have
unloaded Q’s of between 200 and 1000 with capacitor values up to 10pF. Unfortunately
inductor Q’s are considerably less, for example the Coilcraft 0805CS series have
unloaded Q’s of between 50 – 60 and thus are the limiting factor on resonator loaded Q.
2
Radar Subsystem Project
At resonance the transmission phase is zero and the network is loss less (except for the
resistance of the inductor). The series resonator impedes signal transmission while the
parallel network allows signal transmission. The main problem with such a simple
resonator is achieving a required Q, for example if we want a Q of 30 we would need the
following series inductor & capacitor at 200MHz:
The problem with this circuit is that the loaded/unloaded Q and phase of the network are
interdependent and therefore it is better to lightly couple the resonator by using a
capacitor so that the unloaded Q of the resonator is effectively separated from the load
impedance load, which would normally greatly reduce the Q of the resonator.
The addition of the coupling capacitor Cc to the resonator, will allow us to change the
loaded Q of the circuit without greatly effecting the resonant frequency. This
arrangement is shown in Fig. 4. In addition if frequency control is required then the
varactor network shown can be added in parallel with the tuned circuit.
3
Radar Subsystem Project
Fig. 4: Modified L-C resonator with coupling capacitor Cc added to isolated the
unloaded Q of the resonator with the applied load. In addition the varactor network can
be added to give frequency tuning. The bandwidth of the frequency controlled network
is adjusted by the capacitor in series with the varactor.
The easiest way to determine the maximum Coupling capacitor Cc is by way of the
simulation shown in Fig. 5.
Note that an arbitrary inductor value has been chosen and the Capacitor C2 is used to
resonate it at 200MHz. It is better to have a higher value of resonating capacitor so that
the effect of the Coupling Capacitance (Cc) on the resonator frequency is minimized.
Fig. 5: ADS simulation to vary the value of Cc to determine the resonator loaded
Q. Note the S-parameter block needs the ‘Group Delay’ parameter checked.
4
Radar Subsystem Project
The resulting simulation plot is shown in Fig. 6. The plots show us that if we require a
loaded Q of > 30 then Cc must be less than 2.5pF. But we have to remember that Cc will
also effect the phase looking into the resonator which we have to equal to the reflection
amplifier phase but with opposite sign. In addition the varactor network will effect the
loaded Q and resonator frequency depending on the coupling to the varactor. This
coupling determines the capacitance swing and hence tuning bandwidth of the resonator.
So a narrow tuning network gives the better Q than a wideband network.
5
Radar Subsystem Project
Now that we have determined our basic resonator design we need to determine the values
of the varactor network, by choosing a varactor and the correct coupling capacitor to give
us the correct delta C for the resonator. From the varactor data sheet we can note the
capacitance vs voltage curve and thus generate a varactor model for use in our
simulations.
Fig. 7: Typical equivalent circuit of a varactor diode. The series inductor Ls and
parallel capacitor Cp are package parasitics, typical values are Cp ~ 0.1pF and
Ls = 1.5nH.
The general equation for calculating the capacitance of the varactor is:
6
Radar Subsystem Project
Where:
For the Philips BB405 varactor diode, Rs is given as 0.75ohms with capacitances of
18pF@1V, 11pF@3V and 2pF@28V. For the above equation the capacitance values
equate to a value of γ = 0.74 with CJ = 35pF (valid for control voltages 1 to 28V) . The
varactor model was substituted for the fixed capacitor and simulated over a range of
tuning voltages to determine the frequency range and tuning constant of the resonator.
The easiest way to determine the value of the coupling capacitor is to generate a
spreadsheet and enter values of Varactor coupling capacitor as shown in Table 1. From
the table we can see that if we pick a varactor coupling capacitor of 6pF then we should
achieve our tuning bandwidth of ~ 3MHz/V. Note the addition of the resonator coupling
capacitor Cc will alter these values slightly and some adjustments may need to be made.
7
Radar Subsystem Project
We can now put all the components together and simulate the frequency response, Q and
tuning range of the resonator before designing the reflection amplifier. For the next
simulation shown in the varactor sub-model has been added, which is shown in Fig. 8 and
a symbol was generated for use in the resonator model shown in Fig. 9.
8
Radar Subsystem Project
The simulated sweep of Q/frequency vs control voltage (Vcntrl) is shown in Fig. 11. We
can see from the top plot of input return loss (S11) that we need to design a reflection
amplifier with at least 8dB of reflection gain to overcome the losses of the resonator.
Usually the reflection amplifier is design with greater gain > 3dB extra to give some
margin.
9
Radar Subsystem Project
The next section will now deal with the design of the reflection amplifier part of the
colpitts oscillator design.
10
Radar Subsystem Project
Fig. 12: Simplified schematic diagram showing the reflection amplifier part of the
Colpitts oscillator. The capacitor network of C1 and C2 would provide positive
feedback from the emitter to base.
The steady state loop equations were expanded out, (as shown in the Colpitts derivation
tutorial) to yield an expression for the negative impedance of this type of amplifier, in
terms of the capacitors used for feedback and the trans conductance of the active device:
As we assumed that XC1 << hie, then C1 should be as large as possible and in order for
these feedback components to dominate over the parasitic capacitance’s of the transistor
C2 should be as large as possible too.
11
Radar Subsystem Project
By simulation of the Spice model – BFR92 at various collector currents, the input
resistance (hie) and AC current gain (hfe) were determined, in order to calculate the
values of gm and G as given in the previous equations. Fig. 13 below shows the HP ADS
circuit used to analyse the required parameters and Table 2 shows the resulting values of
hie and hfe, for a current of 5mA (Rload = 1150 ohms):
Fig. 13: HP ADS circuit used to analyse the necessary parameters required calculating the
feedback capacitors C1 & C2. To calculate hfe & hie the values of Ic, Ib, vin & ve were
analysed - hfe = Ic/Ib and hie = (vin-ve)/Ib. By RF decoupling the emitter of the transistor,
(by use of C1) the input impedance at 200MHz was predicted, by running an S-parameter
simulation on the circuit.
12
Radar Subsystem Project
By running the simulation with different values of Rload a range of parameters are
calculated so that the colpitts capacitors can be calculated for a particular current. The
table of various Rload settings is shown in Table 3.
Using the values shown in Table 3, we could estimate the maximum series capacitance
of the combination of C1 and C2 i.e. Cm at a picked current of 5mA
Note the input impedance of the resonator Rs can be found be re-simulating the
resonator with a Zin block and for our example Rs = 17 ohms
13
Radar Subsystem Project
Using initial values of C1 and C2 = 30pF, the reflection amplifier was analyzed using
HDADS and the circuit shown in Fig. 14. The two feedback capacitors were optimized to
produce a reflection gain > 12dB and resulted in capacitor values of 33pF for C1 and C2.
This circuit included the spice model of the BFR92, which required biasing to a voltage
supply via the RF bias components for the simulation to work. A
S-parameter analyser box, was added to the schematic to allow the reflection magnitude
and phase to be plotted while, the addition of the DC analyser box allowed DC bias
conditions to be predicted at each node.
Fig. 14: Circuit used to analyse the reflection amplifier, with the additions of
the ‘ideal’ bias components. An S parameter simulation is run to determine
the frequency response of the reflection amplifier’s return gain, together
with the corresponding value of reflection phase.
14
Radar Subsystem Project
Fig.15 shows the frequency response of the reflection amplifier part of the Colpitts with
the feedback capacitors both set to ~ 33pF. The aim was to maximize the ‘reflection gain’
at the centre frequency of 200MHz, by varying the values of C1 and C2. This return gain
must end up being greater than the return loss of the resonator and in addition, the phase
of the resonator must be opposite to the input phase of the reflection amplifier, to ensure
oscillation will occur. The circuit was optimized to maximize the gain of the reflection
amplifier. This occurred with values of capacitors C1 = 33pF and C2 = 33pF, resulting in
a reflection gain of 13.8dB, with a corresponding reflection phase of –67 degrees.
Fig. 15 CAD analysis of the reflection amplifier with the input bias inductor
tuned for best reflection gain. The plot on the left shows return ‘gain’ in dB
and the plot on the right shows the return phase in degrees both at 200MHz.
With these values of gain there was a ~ 6dB gain margin when the resonator
was coupled to the amplifier
The final phase and input return loss plots of the resonator set to the middle of the band ie
200MHz (with a control voltage of 4.3V) is shown in Fig. 16.
15
Radar Subsystem Project
Fig. 16 shows the phase of the resonator with the control voltage of the varactor
set to give a frequency of 200MHz. The resulting phase is +155 degrees with a
return loss of 9.4dB
4. OSCILLATOR SIMULATION
When the two circuit elements (resonator & reflection amplifier) are connected together
that the loop phase will be zero. Analysis of the resonator and the reflection amplifier,
yielded reflection phases of +155 and –67 degrees respectively.
However, the reflection phase of the resonator already includes, the coupling capacitor
required to give the correct loaded ‘Q’. For oscillation to occur there must be >1
magnitude at the zero phase point. Ideally this should occur at maximum reflection gain
for best performance.
The easiest way to evaluate this capacitance & frequency of oscillation is to perform a S-
parameter measurement with the resonator and reflection amplifier connect together as
shown in Fig. 17.
16
Radar Subsystem Project
Fig. 17 S-Parameter simulation of the completed Colpitts oscillator. The OscTest block
has been placed between the resonator & reflection amplifier in order to determine the
magnitude & gain at a particular frequency. That’s to say we require a magnitude
greater than 1 at zero phase at the operating frequency for oscillation to occur.
With the OscTest simulator block connected between the resonator and reflection
amplifier we can simulate the combined phase and magnitude at the oscillation
frequency. The result of the simulation is shown in Fig. 18.
17
Radar Subsystem Project
With the circuit verified for oscillation at 200MHz, we could now simulate the two
circuits together using nonlinear analysis to obtain phase noise, output power, harmonics
and DC bias values. The simulation shown in Fig. 19 uses a Harmonic Balance simulator
to simulate phase noise, output power and frequency performance. A number of
parameters need to be set in the harmonic balance simulator box ie:
18
Radar Subsystem Project
Fig. 19: Harmonic balance ADS schematic of the Colpitts oscillator. The Colpitts
feedback capacitors and resonator tuning capacitor have been optimized for frequency
and oscillator output power. The simulator uses the ‘Oscport’, to inject RF into the
system to allow prediction of oscillated frequency, power and phase noise.
19
Radar Subsystem Project
Fig. 20: Predicted phase noise plot showing phase noise in dBc/Hz against carrier offset
frequency in Hz.
The red line is the simulation result with the blue line showing the required specification. The
oscillator was designed to meet the phase noise specification with a 20 dB margin. In practice
the close to carrier phase noise
I.e. <1 KHz will be at least 10dB worse than the simulation due to the device flicker
noise which is not defined in the BFR92 spice model. Any phase noise below about
–150dBC/Hz was unlikely to be measured as this level fell below the measuring system
noise floor.
20
Radar Subsystem Project
Fig. 21: Harmonic level output simulation showing harmonic output power (dBm)
against the harmonic index. The marker is set to the oscillator fundamental frequency
(index=1). The relatively high second harmonic level required that the following
buffer amplifier was fitted with some form of harmonic filter either as part of the
output matching circuit or a stand alone low-pass filter.
21