Message	Date	From	Subject
1976	2008-08-20 10:37:26	Glen Leinweber	Oscillator topology Questions
We desire very good performance from RF oscillators: -good frequency stability -very low sideband noise -no pulling from the load -constant amplitude output. -few spurious frequency outputs So why do we so often use a single active device in our oscillator? The single device must do the following: -have gain in a feedback loop -lightly load the resonating components -provide power to a load -stabilize amplitude through limiting or AGC Why do we see so few oscillators with multiple stages to split up some of these tasks, each doing its job in an optimal way? We DO tend to use extra stages as a load buffer. But we DON'T tend to split limiting/AGC and feedback gain as separate stages. In EMRFD, a crystal Butler oscillator uses two transistors, and a separate diode limiter. Ulrich Rhode's crystal oscillator uses extra transistors, and a diode detector for AGC feedback. We should be trying these ideas with VFOs using L-C resonators. Most of our single-stage oscillators employ limiting as the amplitude-control method. Their gain is non-linear, like a class-C amplifier. EMRFD cautions that such oscillators are actually mixers, so that any low-frequency noise appears as sidebands on the desired output frequency - not good! Any noise around any of the harmonics can mix too. Yet we see few linear gain-stage LC oscillators that employ AGC amplitude control. Such an approach should kill sideband noise due to mixing. Is AGC amplitude control so difficult? Are Class-C gain stages actually lower-noise than linear gain stages? I'm quite certain that at least some of my conventional non-linear oscillators are unstable because of the following: When switching from conducting to non-conducting state (or visa-versa) one or more VHF parasitic LC resonances that exist in ALL oscillator circuits gets shock-excited, and rings for some time. If it rings long enough (into the next half cycle, or longer) it adds noise and instability to the main oscillator frequency. Some oscillators go to great lengths to stifle these VHF resonances, as in the K7HFD oscillator, which requires ferrite beads on almost all transistor leads. A linear-mode oscillator should be much less susceptible to these shock excitations. Ferrite may still be required to support the desired HF frequency rather than the VHF frequency however.
1978	2008-08-20 17:13:04	Allison Parent	Re: Oscillator topology Questions

1981	2008-08-20 18:00:40	victorkoren	Re: Oscillator topology Questions
Oscillator design is not so simple and straightforward and there is no single design that will suit every set of requirements. For example some advantages for single amplifying stage oscillator: 1) Each active stage adds noise. Potentialy lowest noise oscillator is a single stage oscillator that oscillates at high power, to get the signal as high as possible above the active element noise floor. A multiple stage oscillator can potentialy add more noise to the signal. 2) Each active stage delays the signal it processes, this increases its input to output phase shift. The closed loop phase of the oscillator path adds-up to 0 degrees (Barkhausen law). if the active stages have significant phase shift, the oscillation frequency will not be exactly on the resonant frequency of the LC circuit (or xtal or any resonator) and not be optimal. There will be higher sensitivity of the oscillation frequency to the amplifiers phase shift - frequency drift and higher phase noise. A single amplifier will have low phase shift and better stability and phase noise. 3) Amplitude stabilization through limiting has advantages. in principle noise added to the signal has both amplitude and phase components. Using amplitude stabilization by limiting the signal removes the amplitude noise, leaving only the phase noise. Using a well filtered DC supply and using an active component with low noise at low frequencies, solves the problem of modulating the oscillator signal by low frequency noise. Adding negative feedback at the DC bias circuit lowers even more the low frequency noise, lowering the close to carrier phase noise. Engineering is the science of choosing the right compromise, so the exact topology to use depends on the requirements (low or high frequency, VCO or manualy variable or single frequency, LC or xtal, power used, area

We desire very good performance from RF oscillators:
-good frequency stability
-very low sideband noise
-no pulling from the load
-constant amplitude output.
-few spurious frequency outputs

So why do we so often use a single active device in
our oscillator? The single device must do the following:
-have gain in a feedback loop
-lightly load the resonating components
-provide power to a load
-stabilize amplitude through limiting or AGC

Why do we see so few oscillators with multiple stages
to split up some of these tasks, each doing its job in
an optimal way? We DO tend to use extra stages as
a load buffer. But we DON'T tend to split limiting/AGC
and feedback gain as separate stages.
In EMRFD, a crystal Butler oscillator uses two
transistors, and a separate diode limiter. Ulrich Rhode's
crystal oscillator uses extra transistors, and a diode
detector for AGC feedback. We should be trying these
ideas with VFOs using L-C resonators.

Most of our single-stage oscillators employ limiting as
the amplitude-control method. Their gain is non-linear,
like a class-C amplifier. EMRFD cautions that such
oscillators are actually mixers, so that any low-frequency noise
appears as sidebands on the desired output frequency -
not good! Any noise around any of the harmonics can
mix too. Yet we see few linear gain-stage LC oscillators
that employ AGC amplitude control. Such an approach
should kill sideband noise due to mixing.
Is AGC amplitude control so difficult? Are Class-C
gain stages actually lower-noise than linear gain stages?

I'm quite certain that at least some of my conventional
non-linear oscillators are unstable because of the following:
When switching from conducting to non-conducting state
(or visa-versa) one or more VHF parasitic LC resonances
that exist in ALL oscillator circuits gets shock-excited,
and rings for some time. If it rings long enough (into the
next half cycle, or longer) it adds noise and instability
to the main oscillator frequency. Some oscillators go to
great lengths to stifle these VHF resonances, as in the
K7HFD oscillator, which requires ferrite beads on
almost all transistor leads. A linear-mode oscillator should
be much less susceptible to these shock excitations. Ferrite
may still be required to support the desired HF frequency
rather than the VHF frequency however.