# How to make sure this oscillator will be stable?

In my textbook, the Wien-Bridge oscillator and the Twin-T oscillator needs a tungsten lamp as the self-adjusting resistance:

The lead-lag circuit connected to noninverting input has a gain of 1/3 on its center frequency. Since the lamp started with low heat (less resistance,) the starting close loop gain is big. As the output oscillation builds-up, the lamp heats up and thus close loop gain decreases until it reaches the point where gain is 3 so that $$\A_{V(CL)}*\frac{1}{3}=1\$$. This ensures stable oscillation. (Well, that's how I understood it.)

How can I make sure that the tungsten lamp resistance will remain constant?

Won't the oscillation decay immediately if the resistance of the lamp increases even just a little bit since $$\A_{V(CL)}*\frac{1}{3}\$$ will now be less than one?

• Learn the temperature coeff of R vs power and temp but there are better ways to use this with a zener diode and R for soft limiting gain – Tony Stewart Sunnyskyguy EE75 Aug 14 at 4:57
• @Tony Stewart Sunnyskyguy EE75 Yes I know about temp coefficient, but what I mean is that in real situation won't there be parasitic resistances or some external factors(like temperature of the environment) that can affect the heat of the lamp which will cause change in resistance? And yes I will try to read zener diode for soft limiting later – Iwatani Naofumi Aug 14 at 5:04
• Using a lamp is an ancient way of regulating amplitude, subject to temperature drift and aging of the lamp. – user105652 Aug 14 at 5:54
• How is it done today? Do we use other adjustable resistance for the lamp or we do not use the Wien-Bridge Oscillator entirely anymore(use another entirely different kind of configuration) ? – Iwatani Naofumi Aug 14 at 6:11
• I have several Wien bridge solutions using LEDs or Zeners with >10xR in series for feedback such that when the diodes begin to conduct the impedance matches the 10R or more that reduces the gain 10% or less depending on your tolerance stackup of loop gain to unity positive feedback. 1% is possible . The % controls the THD in the resulting sine and then the OSC starts up slow. But if a Cap is connect to V+ instead of gnd it starts full amplitude then soft limits to the desired Vpp – Tony Stewart Sunnyskyguy EE75 Aug 14 at 12:18

The idea of the stabilization is as you wrote. The immediate decay problem doesn't exist if nothing can suddenly make the temperature of the filament higher. In properly designed generator the filament is insulated from the environment so that in usual environments the heating caused by the output signal surely makes the filament substantially hotter than the environment. Thus the environment cannot warm it, it only causes cooling which is in balance with the heating caused by the output signal.

The filament has thermal mass. It makes its response slow when compared to single pulses of the output signal. That makes very low distortion possible.

But as already said by others, the lamp detoriates gradually like incandescent lamps do so the output amplitude changes as the lamp gets older. And the ambient temperature affects the output amplitude because the needed output power for certain filament temperature depends on the temperature of the environment.

I guess (haven't tested) mechanical knocks also can be detected in the output signal.

• As long as the lamp is operated around 10% of its rated power or less, deterioration isn't a practical issue. – Brian Drummond Aug 14 at 12:36

Practically, measure the resistance of the lamp while cold, and (via V and I) when it just starts to glow (hot), and at some intermediate V (warm). Keeping it that far below rated power will practically eliminate ageing.

Then design the NFB network such that nominally, gain = Av(CL) at the warm resistance, with V(warm) across the lamp. Also ensure for all resistor tolerances, gain > Av(CL) when cold and gain < Av(CL) when hot.

That includes tolerances on resistors R in the frequency selective +ve feedback path if you are adjusting them to adjust frequency. As dual gang pots are pretty poor tolerance, I have resorted to gearing 10-turn 1% wirewound pots together for frequency control ... nowadays a dual MDAC like AD7528 might be more feasible.