The big thing is, this amplitude regulation has very low stability margin. So low, it needs some nonlinearity amplitude compression in the opamp to get stable, that is what "helps" it when going near clipping. If you look at it from the regulation feedback loop theory perspective, you find two inertial elements in the regulation feedback loop: One is the filament thermal inertia and the second is the response of the oscillator circuit when the lamp resistance changes, becoming even slightly 3 means the amplitude is increasing till some compression effect brings the gain back to the exact 3 (assume the RC positive feedback is exact and there is no phase shift in the amplifier). And vice versa, even slightly smaller than 3 means the amplitude starts decreasing till it disappears. And because each inertia means a 90deg phsse shift component in the response function, having two of them means a supposedly negative feedback (higher amplitude -> higher lamp resistance -> lower gain -> amplitude getting lower) may actually become nearly positive at a certain frequency range. Just because the lamp, as well as the Wien bridge oscillator) have some noninertial components in its response (static V-R curve of the lamp, amplitude compression in the opamp), the regulation feedback phase shift does not reach the 180deg to cancel the intended negative feedback, the amplitude regulation loop is still somewhat stable (it does not generate other oscillation forming AM on the output). But even when stable, it exhibits rather strong resonance, which then amplifies any disturbance (e.g. the filament motion, or even just the circuit noise). Note we are speaking about the generated sinewave envelope as the variable to be regulated here. In fact this circuit was originally designed with vacuum tubes, exhibitting quite significant low order distortion. This distortion on one hand still does not generate that much of higher harmonics (compare to e.g. opamp or diode clipping), yet provides very significant amplitude compression characteristics needed to have the thermally controlled feedback loop stable, so the generated signal free from amplitude fluctuations. Using a diode clipper to vary the gain makes the amplitude very stable (as the only inertial element in the amplitude regulation loop is the main Wien bridge oscillator response), but is causes the generated signal to get distorted. Or you may combine the diodes and bulb to get some compromise, but it is all the time just a compromise between amplitude stability (so side band noise around the generated signal in the spectrum), vs distortion (harmonisc). To get away from this compromise, most advanced variants use some form of DC voltage controlled gain (a FET in the feedback as a variable resistor, analog multiplier circuit,...) controlled from rectified and filtered output. This way you may shape the dynamic response of the amplitude feedback by the feedback filter components so well it won't pass anything on the generated signal (so does not cause harmonic distortion), while still maintaining very good amplitude regulation stability margin, both at the same time. Yes, then we are talking about the amplitude regulation circuit to become way more complex than the main Wien bridge generator core itself, but that is what it took to get a clean sinewave before the modern digital synthesizers became usable...
Nice demonstration! I wonder what happens when you slow down the variable lamp resistance even more with a cap in parallel to it. By smoothing/integrating over how many sine periods do you think is the most stable operation achieved?
If you put a cap in parallel therre will be no stabilisation since it is the AC gain at the oscillation frequency that must be reduced. The cap will negate that effect.
Great experiment makes me wonder whether you could utilize a thermistor with a similar to the bulb coefficient. Maybe thermally coupled with a small resistor or just a low-value thermistor on its own.
I have read that thermistors have been used. But I think most modern serious attempts use fairly complicated JFET circuits for gain stabilization. I think every approach requires tweaking.
Microphonics from touching it, vibrating the lamps filament. The wien bridge really got my attention when trying to get a decent sine wave at audio frequencies. Turns out it's not that easy!
I mentioned it in a reply to another video of yours , that using multiple lamps adds thermal mass , hence slowing down the changes resulting in a more stable and cleaner output. I've ordered in a LOT of different lamps ( over $100!) And now trying to get a selection of op amps when I get some testing finished I'll let you know what I find. Thanks for your videos!
really you need not use only the lamp in the divider leg. consider adding an R in series with the lamp. Choose Rf and (Rin + [Rlamp@peak current| + Rextra) so that the gain with nominal lamp resistance value is 3. The idea is too limit the extents of the gain from ~=2.9 to 3.1 the purpose of the lamp then is NOT to cover a wide range of gain compensation but only a percentage. This should prevent the large swings and settling.
Resistor stability will be so critical that Low distortion say about -100 db below fundamental wil N OT be obtainable , that is why Hewlitt used a light bulb !
![Felix "Unfair" | [Stray Kids : SKZ-PLAYER]](http://i.ytimg.com/vi/Oswujxm2Ag0/mqdefault.jpg)
The big thing is, this amplitude regulation has very low stability margin. So low, it needs some nonlinearity amplitude compression in the opamp to get stable, that is what "helps" it when going near clipping.
If you look at it from the regulation feedback loop theory perspective, you find two inertial elements in the regulation feedback loop: One is the filament thermal inertia and the second is the response of the oscillator circuit when the lamp resistance changes, becoming even slightly 3 means the amplitude is increasing till some compression effect brings the gain back to the exact 3 (assume the RC positive feedback is exact and there is no phase shift in the amplifier). And vice versa, even slightly smaller than 3 means the amplitude starts decreasing till it disappears.
And because each inertia means a 90deg phsse shift component in the response function, having two of them means a supposedly negative feedback (higher amplitude -> higher lamp resistance -> lower gain -> amplitude getting lower) may actually become nearly positive at a certain frequency range. Just because the lamp, as well as the Wien bridge oscillator) have some noninertial components in its response (static V-R curve of the lamp, amplitude compression in the opamp), the regulation feedback phase shift does not reach the 180deg to cancel the intended negative feedback, the amplitude regulation loop is still somewhat stable (it does not generate other oscillation forming AM on the output). But even when stable, it exhibits rather strong resonance, which then amplifies any disturbance (e.g. the filament motion, or even just the circuit noise).
Note we are speaking about the generated sinewave envelope as the variable to be regulated here.
In fact this circuit was originally designed with vacuum tubes, exhibitting quite significant low order distortion. This distortion on one hand still does not generate that much of higher harmonics (compare to e.g. opamp or diode clipping), yet provides very significant amplitude compression characteristics needed to have the thermally controlled feedback loop stable, so the generated signal free from amplitude fluctuations.
Using a diode clipper to vary the gain makes the amplitude very stable (as the only inertial element in the amplitude regulation loop is the main Wien bridge oscillator response), but is causes the generated signal to get distorted. Or you may combine the diodes and bulb to get some compromise, but it is all the time just a compromise between amplitude stability (so side band noise around the generated signal in the spectrum), vs distortion (harmonisc).
To get away from this compromise, most advanced variants use some form of DC voltage controlled gain (a FET in the feedback as a variable resistor, analog multiplier circuit,...) controlled from rectified and filtered output. This way you may shape the dynamic response of the amplitude feedback by the feedback filter components so well it won't pass anything on the generated signal (so does not cause harmonic distortion), while still maintaining very good amplitude regulation stability margin, both at the same time.
Yes, then we are talking about the amplitude regulation circuit to become way more complex than the main Wien bridge generator core itself, but that is what it took to get a clean sinewave before the modern digital synthesizers became usable...
This is great work ! thanks.
Nice demonstration! I wonder what happens when you slow down the variable lamp resistance even more with a cap in parallel to it. By smoothing/integrating over how many sine periods do you think is the most stable operation achieved?
If you put a cap in parallel therre will be no stabilisation since it is the AC gain at the oscillation frequency that must be reduced. The cap will negate that effect.
Near black on black might be very elegant, but it doesn't help in seeing the waveforms
Great experiment makes me wonder whether you could utilize a thermistor with a similar to the bulb coefficient. Maybe thermally coupled with a small resistor or just a low-value thermistor on its own.
I have read that thermistors have been used. But I think most modern serious attempts use fairly complicated JFET circuits for gain stabilization. I think every approach requires tweaking.
The oscillator with lamp was used in old HP generator. Therefore was a good circuit.
Microphonics from touching it, vibrating the lamps filament. The wien bridge really got my attention when trying to get a decent sine wave at audio frequencies. Turns out it's not that easy!
I mentioned it in a reply to another video of yours , that using multiple lamps adds thermal mass , hence slowing down the changes resulting in a more stable and cleaner output. I've ordered in a LOT of different lamps ( over $100!) And now trying to get a selection of op amps when I get some testing finished I'll let you know what I find. Thanks for your videos!
@@nopenadanowaynohow This sounds fascinating too !
really you need not use only the lamp in the divider leg.
consider adding an R in series with the lamp.
Choose Rf and (Rin + [Rlamp@peak current| + Rextra) so that the gain with nominal lamp resistance value is 3.
The idea is too limit the extents of the gain from ~=2.9 to 3.1 the purpose of the lamp then is NOT to cover a wide range of gain compensation but only a percentage. This should prevent the large swings and settling.
Resistor stability will be so critical that Low distortion say about -100 db below fundamental wil N OT be obtainable , that is why Hewlitt used a light bulb !
@@martinmartinmartin2996 note R in series with lamp, and this comment was 7 months ago,... I don't even recall the video.