# Colpitts oscillator: how to reduce amplitude thermal drift?

I have built this circuit, supply ±12V:

I am using a LM6172 amplifier.

I have found the transfer function of the filter involved:

$$\H(j\omega)=\frac{1}{1 + \frac{R_3}{R_1} - L\omega^2(C_2+C_1\frac{R_3}{R_1})+j(\frac{L\omega}{R_1} - R_3C_1C_2\omega^2(L\omega-\frac{1}{C_e \omega}))}\$$

Upon my calculations, the gain provided by the op amp should be of ~1 but it didn't start and I had to increase the gain to 1.8 (R2/R1). Also, without the R4 resistor the op amp is acting strangely, saturates - very high peak to peak amplitude - and oscillates at very high frequency (10 times the one of the oscillator) and heats a lot.

My problem is that the amplitude of the oscillations - at 1MHz - is very unstable as temperature varies slightly.

I just have to blow on the ciruit to change the peak-to-peak amplitude by ~2V

I also tried with thsese values:

$$\R_1=1k\Omega,R_2=10k\Omega\$$

But I got the same behaviour.

I don't know if the problem comes from the resistors as they change of value with little temperature variations or if it comes from the op amp directly.

Maybe I have badly chosen the values of the components?

Is there a way to reduce this amplitude thermal drift of this circuit?

UPDATE: I just figured I was using carbon film resistors which have a high temperature coefficient. I will try to replace them by metal film resistors.Results: Didn't change anything.

• Don’t you know how to compute reactive impedance and load resistance? Make R1,R2 about 100 x bigger e.g 1M, 10M respectively. You are on the verge of insufficient gain for no oscillation, Jul 13, 2019 at 1:40
• Or 1000x bigger than 1k, 10k Jul 13, 2019 at 2:03
• That is why thermally stable oscillators use Crystal's and some use temperature controlled ovens. A purely analog oscillator cannot be temperature stable over a wide range.
– user105652
Jul 13, 2019 at 4:17
• @SunnyskyguyEE75 I don't understand how making R1 and R2 100 times bigger would make any difference since the op amp gain -absolute- is given by $\frac{R_2}{R_1}$. In the datasheet of the LM6172, it's specified that low feedback resistor values - about 1k ohms - are recommanded at high frequencies. According to the transfer function of the filter which seems to be correct after simulator tests, with those values, op amp gain is supposed to be about 1 because the filter is near gain resonance frquency. Jul 13, 2019 at 9:57
• @Sparky256 Thanks for the informations. My biggest problem is that the temperature variations are really tiny. I'd say 1 or 2°C are enough to make in big change in amplitude. Do you think I should try a crystal? Jul 13, 2019 at 10:01