AWR eBooks

TECHNICAL MEMORANDUM Page 9 of 12 further investigation is required. The oscillator tuning versus voltage relies on the varactor model and the coupling capacitor. The hyper-abrupt tuning varactor part number is MHV514-21 from Micrometrics. Micrometrics supplied the Gamma and capacitance versus voltage data. The data was entered into the simulation as discrete points and a polynomial regression was used within the program to approximate a continuous function. An error tolerance may be specified to indicate if the polynomial deviates from the discrete point data by more than a specified percent; in this case, the error was less than 2%. VI. OSCILLATOR SIMULATION ISSUES Simulation of RF and microwave oscillators present several limitations with respect to accuracy, particularly under conditions where the active device is operated over a wide dynamic range of current and/or voltage. More specifically, bipolar junction and field effect transistor noise parameters vary dramatically with bias, and unfortunately, are usually extracted under 'nominal' bias conditions. While parametric extraction at 'nominal' bias conditions is quite adequate for small signal and some large signal amplifier simulations, it is entirely inadequate for accurate oscillator simulation. An additional consideration is the statistical variation of parametric device data typically associated with manufacturing processes and other tangential considerations, e.g., environment. Notwithstanding the benefits and improvements in harmonic balance simulation algorithms, accurate oscillator simulation requires parametric data of the active device over the entire range of operational current and voltage. The extraction of a single set of parameters to represent the active device under static and dynamic operation is a formidable task. From experience with the push-push oscillator of this investigation and other less complex oscillator circuit topologies, the fundamental limitations in simulation accuracy are summarized: • Inaccurate BJT and FET device model parameters and parameter variation; particularly noise parameters • Resistive and reactive loading of the fundamental and harmonics • Highly nonlinear device operation, e.g. see the dynamic load line excursion • Device manufacturing variance 3 The Chang reference has been added to the June, 2019 revision. In addition to accurate model data over large regions of the operating voltage and current i.e. cut-off to saturation, harmonic balance algorithms exhibit a profound impact in oscillator simulation performance. VII. CONCLUSIONS Conventional oscillator theory of operation has been demonstrated using a simple, common collector oscillator configuration. The common collector oscillator has been extended and shown to be a basic element of the push-push oscillator circuit topology. Highly nonlinear operation, model device parameter extraction and simulation algorithm are fundamental limitations to accurate oscillator simulation. Notwithstanding accuracy limitations, computer simulation is shown to be an effective educational tool. THE 2019 PERSPECTIVE As mentioned in the preface, this technical documentation was originally written in 2005 to address oscillator tuning issues within a radar subsystem. The exercise created a better understanding of the push-push oscillator circuit topology, and to that extent, the simulation software provided a valuable contribution, notwithstanding the BJT device model limitations. In retrospect, there are several elements that merit consideration: • The active device model remains a significant issue with respect to simulation accuracy – particularly phase noise prediction. • The Spice model selection remains an appropriate choice for oscillator simulation. • Discrete BJT and FET oscillator design is becoming – or has become – the province of MMIC manufacturers. Over the past 15 years, the author's oscillator design activities have entailed MMIC device selection, i.e., VCOs, PLLs and transmit/receive subsystems. • The unique low phase noise property of the push-push oscillator circuit topology is likely the higher 'Q' due to reduced loading at the fundamental frequency, as suggested by Gris [3.], and coupled oscillator properties as described in Chang [7.] 3 . X-Band Push-Push Oscillator Simulation and Measurement 9 www.cadence.com/go/awr

• Cover