Simple wien bridge oscillator with temperature compensation

[akis]

Picture 1 shows a basic wien bridge oscillator. Gain must be 3 or over in order to start oscillations. In practice gain is slightly over 3 to ensure oscillations will occur.

However the oscillations reach the rails and then clip, so you get a square wave (almost) and lots of harmonics as a result (which is not very useful).

The old wien bridge used a filament lamp in the feedback loop, where as the filament heats up the resistance increases thus reducing gain. This stabilises the circuit and it does not clip, and you get a nice sinusoidal output.

Picture 2 shows the same oscillator with a diode arrangement in the feedback loop which serves to reduce gain, similarly to the filament lamp's operation, and it is a very common circuit which I have seen in many search results.

I was running temperature stability tests on some transistors and heatsinks using a 3KW room heater fan heating up the components to over 60 C or cooling them down quickly on the "cool" setting. Room temperature was 14.5 C. So my tests ran from 15C to 70C.

I noticed that my wien bridge oscillator which was placed a few inches away was greatly affected by the temperature rise. The output swing would drop, and it even stopped oscillating altogether. The problem is with D1/D2 which as they heat up conduct more and reduce the gain.

I then came up with the circuit in figure three. With the addition of R7/D3/D4 I have measured the output to be almost perfectly stable between 15 C - 65 C.

 

[DF96]

Did you try simply increasing the value of the 15k in series with the diodes? Provided the total resistance here is more than 40k it ought to oscillate. A pair of back-to-back diodes ought to give you more than 25k with small signals, unless they are leaky or the frequency is so high that stray capacitance is significant. I suspect the latter as you are using about 200kHz. What type of diodes did you actually use?

 

[Elvee]

The problem with 1N4148-types diodes is their gold doping to improve speed.
It does something for speed, but unfortunately, it also ruins about every other characteristic.

And nowadays, all the so-called general-purpose signal diodes all undergo that kind of treatment: BAW62, 1S2473, etc etc

If you can get get hold of some OA200 or low digit 1N-series, they won't display the same behavior.

Alternatively, you can use the CB of a small transistor, preferably a HF one.

The fix you found may work, but it isn't a very good one as it introduces a degree of indetermination

Here is the behavior of the 1N4148 at around 200KHz.
You can see that the conductance increases with temperature, gently at first, more dramatically when you pass 100°C (grey an blue curves, 110 and 150°C).

By contrast, the BF240 shows a diminished overall conductance with temperature.

 

[Elvee]

Another option could be the use of a more deterministic type of clipper:

 

[Osvaldo de Banfield]

Or a simple incandescent lamp as from years has been used.

 

[akis]

The problem with 1N4148-types diodes is their gold doping to improve speed.
It does something for speed, but unfortunately, it also ruins about every other characteristic.

And nowadays, all the so-called general-purpose signal diodes all undergo that kind of treatment: BAW62, 1S2473, etc etc

If you can get get hold of some OA200 or low digit 1N-series, they won't display the same behavior.

Alternatively, you can use the CB of a small transistor, preferably a HF one.

The fix you found may work, but it isn't a very good one as it introduces a degree of indetermination

Here is the behavior of the 1N4148 at around 200KHz.
You can see that the conductance increases with temperature, gently at first, more dramatically when you pass 100°C (grey an blue curves, 110 and 150°C).

By contrast, the BF240 shows a diminished overall conductance with temperature.

I have run more tests using schottky diodes (BAT48), germanium diodes, and a variety of transistors BC548C, BC550C, BC560C, MPSA18, BC640. Unfortunately I did not have any BF240 to try with. All the diodes were affected almost equally by temperature. The transistors were less sensitive to the temperature rise, but produced a funny looking curve on the scope, and also required some temperature compensation. So for the time being the best results in terms of curve shape and temperature stability (up to 75C) is the schematic I showed above.

 

[akis]

Another option could be the use of a more deterministic type of clipper:

OK this is too complicated for me I have not built it on breadboard to test it, but is it temperature stable?

For what its worth there is a document on op-amp oscillators here http://www.ti.com/sc/docs/apps/msp/journal/aug2000/aug_07.pdf

but it does not talk about temperature stability. I quote from it "The Wien-bridge oscillator has few parts, and its frequency stability is good. Taming the distortion in a Wien bridge oscillator is harder than getting the circuit to
oscillate". The last part means that the circuit will either not oscillate or with a slight change in the feedback it will oscillate to clipping, and it is hard to maintain the curve within the rails.

 

[Elvee]

I have run more tests using schottky diodes (BAT48), germanium diodes.

These are the worst possible choice.
A suitable type would be band-switching diodes for TV tuners: BA143, BA182, etc.
PIN diodes might work, but should preferably be avoided

 

[KSTR]

The problem with 1N4148-types diodes is their gold doping to improve speed.
It does something for speed, but unfortunately, it also ruins about every other characteristic.

And nowadays, all the so-called general-purpose signal diodes all undergo that kind of treatment: BAW62, 1S2473, etc etc

From my colleage I learnt recently the same thing (he has semicon die-level expertise). He thinks that a JFET typically is an excellent diode, too, because it's pn-junction must be very pure, no doping.

 

[Elvee]

He thinks that a JFET typically is an excellent diode, too, because it's pn-junction must be very pure, no doping.

Yes indeed, but it is the deluxe option for this application.

 

[akis]

I have tried a number of FETs, BF244B, 2N5459, 2N5462. I have settled on the BF244B. I have discovered that when I start the heater fan to heat the circuit to 80C, the 1N4148 diodes behaved a bit tempermentally, depending on how they were mounted on the breadboard and which direction they were facing, which indicates to me that there is a capacitance issue on top of the voltage drop issue as the diodes heat up. With the FETs the circuit was much less sensitive to position, orientation etc. So I will use two pairs of FETs in place of the two pairs of diodes.