Oscillator Fun!

[vxp036000]

So, I have two oscillators designed to operate at exactly the same frequency, but with vastly different topologies.  Which would you choose and why?  I'm still learning, so there are no right answers :-)

[nctnico]

So, I have two oscillators designed to operate at exactly the same frequency, but with vastly different topologies.  Which would you choose and why?

Neither if you don't tell what you need an oscillator for 

[vxp036000]

Lol!  That's part of the fun.  I'm thinking along the lines of a general purpose RF source with clean spectral output.  Something like a UHF VCO (I'm going to ignore dedicated ICs for the time being).

So, I have two oscillators designed to operate at exactly the same frequency, but with vastly different topologies.  Which would you choose and why?

Neither if you don't tell what you need an oscillator for 

[jimmc]

I've never liked the common base circuit as shown, the circulating current in the tuned circuit has to flow through the power supply leading to poor performance.  Simply moving the lower end of C1 from ground to the supply rail improves things.

I can't see the second circuit as a Colpitts.

My own favourite Colpitts is the common collector eg
http://how-to.wikia.com/wiki/How_to_build_an_oscillator_circuit.
This has the advantage of having one end of the inductor earthed which makes replacing it with a resonant line or adding varicap tuning much easier.

See fig 7 here for an example of a Colpitts running at UHF http://www.skyworksinc.com/uploads/documents/200316A.pdf

Jim

[EmbeddedEric]

Cool timing, I actually just finally got my JFET based Oscillator up and running last night and wrote about it this morning. Another topology for you to evaluate.

http://embeddederic.blogspot.com/2012/05/rf-isnt-black-magic.html

[nctnico]

Lol!  That's part of the fun.  I'm thinking along the lines of a general purpose RF source with clean spectral output.  Something like a UHF VCO (I'm going to ignore dedicated ICs for the time being).

So, I have two oscillators designed to operate at exactly the same frequency, but with vastly different topologies.  Which would you choose and why?

Neither if you don't tell what you need an oscillator for 

AFAIK the biggest trick is to get an oscillator compensated for temperature changes. An interesting experiment is connecting 2 identical crystal oscillator modules to an oscilloscope and use the X-Y mode to show the phase and frequency difference. Just blowing some air on one on them is enough to see the changes.

[vxp036000]

I guess I should have clarified that in any practical implementation, both of these circuits would have a blocking cap from Vcc to ground and possibly even an RFC to provide isolation from the supply.  No signal path through the supply line.

Your probably right about the second oscillator not being Colpitts, closer to Hartley.  The downside I see in the second architecture is that there is a spurious resonance at sqrt(2) w0.  I'll let you figure out why    So much for a clean output waveform. 

I've never liked the common base circuit as shown, the circulating current in the tuned circuit has to flow through the power supply leading to poor performance.  Simply moving the lower end of C1 from ground to the supply rail improves things.

I can't see the second circuit as a Colpitts.

My own favourite Colpitts is the common collector eg
http://how-to.wikia.com/wiki/How_to_build_an_oscillator_circuit.
This has the advantage of having one end of the inductor earthed which makes replacing it with a resonant line or adding varicap tuning much easier.

See fig 7 here for an example of a Colpitts running at UHF http://www.skyworksinc.com/uploads/documents/200316A.pdf

Jim

[vxp036000]

That's an interesting concept, buffering the feedback through J2.  The one thing I don't like is that this oscillator requires two transistors.  Is there an advantage to this over a conventional single transistor design?

Cool timing, I actually just finally got my JFET based Oscillator up and running last night and wrote about it this morning. Another topology for you to evaluate.

http://embeddederic.blogspot.com/2012/05/rf-isnt-black-magic.html

[vxp036000]

Here is another common textbook oscillator.  We'll see how it stacks up; haven't finished evaluating the open loop phase and gain yet.

[vxp036000]

What do you know, another poor topology.  Plotting the open loop phase and gain clearly shows a stray resonance at w0 / sqrt(2).  It will be interesting to see if the Hartley configuration for a common collector oscillator has better spectral purity.  Here is a bode plot, so you can see what I'm talking about.

[vxp036000]

In case anyone is interested, here is the general approach I use in evaluating oscillator configurations.  I'm working through a few Hartley oscillators to see if they're any better with respect to spurious outputs and phase stability.

[jimmc]

I guess I should have clarified that in any practical implementation, both of these circuits would have a blocking cap from Vcc to ground and possibly even an RFC to provide isolation from the supply.  No signal path through the supply line.

You misunderstand me a little, if you feed a current (I) to a simple parallel LC network at its resonant frequency then the current in both L and C will be Q times I.
Where Q is the quality factor of the tuned circuit. (Energy is exchanged between them during each cycle)

Take an example...
For a particular tuned circuit at resonance X_L = X_C = 10ohms
If the Q of the inductor is 100 and the Q of the Capacitor is very high (>>100)
then at resonance the tuned circuit will look like 1k resistor (Q.X_L = Q.X_C = 100x10)
Stick a volt across the tuned circuit an the current will be 1mA but the current through the inductor or capacitor will be 1v/10ohms = 100mA.

Again if you try injecting a small voltage (V) in series with the capacitor you will find the voltage across the tuned circuit is Q.V

My point was that moving C1 as I suggested would reduce the current passing through the supply decoupling by a factor of Q and also reduce the susceptibility to ripple at the resonant frequency by the same factor.

Jim

Oops...
Forgot to add that as well as moving C1 to the positive rail, C3 should also be moved. ie all refer all RF paths to the positive rail.

If you want thee RF ground to be the negative rail, switch to a PNP transistor and swap the rails or use the common collector circuit.

[vxp036000]

Hmm, that's an interesting point.  I'm tempted to build it up and see if practically speaking it makes much difference. 

You misunderstand me a little, if you feed a current (I) to a simple parallel LC network at its resonant frequency then the current in both L and C will be Q times I.
Where Q is the quality factor of the tuned circuit. (Energy is exchanged between them during each cycle)

Take an example...
For a particular tuned circuit at resonance X_L = X_C = 10ohms
If the Q of the inductor is 100 and the Q of the Capacitor is very high (>>100)
then at resonance the tuned circuit will look like 1k resistor (Q.X_L = Q.X_C = 100x10)
Stick a volt across the tuned circuit an the current will be 1mA but the current through the inductor or capacitor will be 1v/10ohms = 100mA.

Again if you try injecting a small voltage (V) in series with the capacitor you will find the voltage across the tuned circuit is Q.V

My point was that moving C1 as I suggested would reduce the current passing through the supply decoupling by a factor of Q and also reduce the susceptibility to ripple at the resonant frequency by the same factor.

Jim

[zeino]

Do you know why in * OscillatorFundamentals2.pdf" shared by vxp036000 when breaking the loop of the common emitter Oscilator it includes the RFC in the analysis? From which book or document is this analysis from. I am a little confused on how it breaks the feedback and includes the RFC in the frequency analysis. Can anyone help me on understanding it.