linear oscillator analysis - will it oscillate ?

Apr 10, 2007 1 Replies
C

Hello,



I am trying to design a 24GHz transistor oscillator (feedback with transmission line resonator). I have only s-parameter files from the transistor available - so I can do only linear analysis to evaluate whether the oscillator will be unstable. I am using ADS for simulation



- with the modul "osctest" it is possible to evaluate the closed loop gain. According to Nyquist criterium the loop gain should be greater than 1 when the phase is zero. All the examples I have seen are using real transistor models , i.e. they are starting with the linear analysis and afterwards a harmonic balance simulation to confirm the oscillation. My loop gain is only "a little bit" greater than 1, actually 1.006 ! I have tried to change feedback characteristics and I have taken other transitor s-parameter, but the result is the same = only 1.006 of loop gain. What are your experience with the loop gain when using only s- parameters ? Does it make sense to build this oscillator on the basis of this simulation ?



Thanks for your feedback !



Regards Peter



Vote
Twitter Facebook LinkedIn
W

Hi Peter,

It comes very strange to me that two different transistors (probably with different Ft) have same loop gain under same load conditions. I think there is something wrong in the result.

If I really had a circuit with a (power) loop gain of 1.006 in a linear simulation, I would not build it. Normal deviation in component characteristics would result in a very unreliable circuit.

When I have to design an oscillator, I also start with the small signal model. It doesn't matter in what type of parameter system you do the analysis, as long as you know what you are doing. If I only have S-parameter data, I would use negative resistance method.

Did you try to simulate an amplifier based on the S-parameter file? If so, were the results as expected (for example did they matched the published data)?.

Did you consider the negative resistance (|input reflection coefficient| > 1) method for your circuit?

My experience with ADS dates back to the first introduction and I never used an "osctest" option.

When you could find package data valid up to 24 GHz, in combination with a spice model, you could use a SPICE simulator. Of course, you have to convert every structure to transmission line components and or C's and L's. If you are designing a common emitter oscillator, parasitic emitter induction has large influence on whether or not your transistor will oscillate.

Best Regards,

Wim PA3DJS

Vote
Twitter Facebook LinkedIn

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required