VCO that simulates but does not work in reality
I designed a fantastic VCO that consumes 1mA collector current and utilises a 1V supply. The VCO simulates perfectly, easily achieving the desired bandwidth and power output. However I have laid the oscillator out on a pcb only to find that in reality it doesnt oscillate at all.
When designing I have assumed ideal components. Would those of you with more experience always simulate with components that represent the non ideal even at UHF frequencies say 450-458MHz? i.e should I be modelling an inductor as actually an inductor in series with a resistor etc?
Secondly would those of you with more experience think this design is easily achievable without say a VNA.
I am hoping to ultimately design a UHF pager to operate from 450-458MHz.
Cheers
I guess, it can work with a small change to the circuit, e.g. capacitor ratios defining the feedback. For a realistic simulation,
the SOT23 package inductances should be included, other parasitics - depends on.
I think, it's much more easy to make the oscillator work by interactive design changes than to spend the effort of connecting
a VNA, if you have it. Those GHz transistor circuits are more likely to oscillate, if you don't want it to, than not to oscillate.
I suppose you checked the bias point.
I have been using the models provided with Agilent ADS. Do they not include modelling for component pins in their spice and S parameter models?
Cheers
With the S-parameters most likely, not necessarily with SPICE models.
Inductor should be modelled with its series resistance ( lower Q factor will decrease the tend of oscillation intention) and parallel parasitic capacitance.Also, every critical component should be replaced with an equivalent circuit that is provided by manufacturer.I mean, if you use -for instance- Murata components, you should also simulate with the equivalent models of the components.Ideal components will fake you out in terms of oscillation frequency and phase noise.
Generally, the SPICE models do not include parasitic package components, so you should add these parasitics by hand onto intrinsic spice model.This is very important at especially higher frequencies.
I recently had a similar problem. This can be very frustrating if you don't have the right tools to investiate. The oscillator design checked out with both linear (open loop) and non linear simulation on 4 different (ADS, Ansoft, Genesys, AWR) simulators. All gave results that were very close with excellent phase noise. All possible parasitics were included in the simulation. Used components with known s-parameters from vendors. However when a test circuit was constructed the oscillator failed to start.
Verified the open loop design on a VNA and found the loop gain to be much lower than in the simulation (1-2dB instead of 6-7dB). After tweeking some values the oscillator wanted to start but at a sub harmonic. After applying some selective negative feedback and some more tweeking, it eventually worked. Design values differed quite a bit from the practical ones.
BFR182 it will not work at 1V. At this voltage you can try to use a BFP740 from Infineon.
I agree with bigboss. You must not use ideal inductor models. An infinite Q inductor will alter simulated results significantly. SRF will also have an impact. It would be worthwhile including parasitics of a few key capacitors.
Keith
in oscillator design , it is not just the inductor , but all the passives must have very accurate models. to properly model the loss.
also the PCB interconnections must be taken into account , they add some inductive parasitic , so they can really affect the performance of the real hardware oscillator.
the pacakge parasitic of the transitor should be inculded as well.
these are the general guide lines in oscillator design.
Important note : I simulate the Oscillator layout with momentum in ADS and include it in VCO simulation, and i have got good resluts , and they are close to simulation.
Khouly