I'm designing this oscillator for a start-up product and am having trouble with the amplitude control. I'm using the simple feedback resistors. I seem to be needing ~20k for Rg and 82k for Rf which is not the gain=3 that seems to be theorized... but it works. My problem is that it is very much temperature dependent and the Rg must change from 18k-22k for the range of ambient temperatures for which it will be rated for.

I've considered using Thermistors but not sure about specifics. I know the Rg should be PTC or the Rf should be NTC but past that not much.

Also, do you guys think that a diode stabilized control will be more robust? should I ditch all this crazy thermistor stuff and go for a few mote components?

Thank you,

 

Use a JFET as a variable resistance element in the feedback path.

Google "AGC circuits" for examples of how this is done. This approach yields great amplitude stability and waveform purity.

 

Something is wrong if your gain needs to be much more than 3. If you are not using an op amp for your amplifier, you should be.

 

Something is wrong if your gain needs to be much more than 3. If you are not using an op amp for your amplifier, you should be.

I am using a TL704 opamp.

I'll look into JFET stability as well.

As you can see, I cannot use a pot to adjust for each oscillator as this will be a mass produced product.

 

Can you tell us more about the design criteria of this oscillator?
What does it need to do well?

 

Download this book from TI. It has a goos section on oscillators
www.ti.com/lit/an/slod006b/slod006b.pdf

 

Why are you using such low-value capacitors? The input capacitance to the op amp, plus the stray wiring capacitance, are of the same order of magnitude as your 30pF caps. If you are testing using a socket, that will add to it. A solderless breadboard will REALLY screw you up. This is the reason your gain has to be higher than 3. I simulated your circuit, and concluded that, with your gain-setting resistor values, the unwanted capacitance in parallel with C5 is about 14 - 15pF. This could be off by a few pFs either way.
Any fixed-resistor feedback scheme to set the gain will fail, unless you can tolerate considerable harmonic distortion. With a fixed feedback ratio, the circuit will either not oscillate, or, if you get the ratio just low enough to start and sustain oscillation, the amplitude will quickly rise until some nonlinearity is reached which lowers the gain slightly. With fixed resistors, this will be when the op amp saturates at one or both supply rails, causing distortion.

If you want stable amplitude with low harmonic distortion, you need a circuit that detects when the peak amplitude has exceeded a certain value (which is less than the saturation voltage of the op amp). The peak detector will hold the gain at the value required to sustain that peak value.

If you can tolerate some distortion, say -40dB, you can add a diode-resistor network to the feedback loop. This will probably be simpler, but will have more distortion that the peak detector scheme.

What amplitude do you need? How stable does it need to be?
How much distordion can you tolerate?

 

I appreciate the simulation Ron, I hadn't tried it and didn't realize that such extreme values would need closer inspection of the stray capacitance.

I'm working on a PCB so I have minimal capacitance, plus the frequency doesn't matter as much. I had designed for 25Khz but I measure 20Khz which is fine (i need >10Khz)

As far as harmonic distortion goes I don't know enough about it to design for or against it. It should be kept "low" whatever that might mean.

I thought the gain of 3 was only affected by the Rf and Rg feedback and not at all by the stray capacitance I have in the non-inverting leg.

Either way I'm thinking about designing for lowest common denominator with Rf=82k and Rg=19k. It seems that this assures oscillation will happen, I'd get some clipping if the ideal value should be more like 20 or 22k but at least I know for sure that it's oscillating since I havn't seen any scenario in testing that needs Rg<19k for oscillation to start. Now I just have to test how much "square" wave-ness can I tolerate.

Any extra thoughts on that? You guys' experience is really helping me out.

 

But why are you using 30pF? Why not use something like 100nF (1000pF) and 6.2k? Then your required gain resistors would be much closer to the calculated values.

As I said before, you will always get some clipping if your only gain reduction comes from op amp saturation.

What is the purpose of the oscillator?

 

You're definitly right about my frequency setting R and C values, they should be less extreme.

This oscillator is based on an old circuit for water conductivity sensor
with a bunch of changes of course. I'm using the oscillator and gain loop parts of the diagram given with a precision rectifier and cap for the microprocessor. I took out the zener stabilization and need a fixed resistor for the feedback loop instead of a pot.

I need a bit of clipping to ensure that every PCB that gets assembled is capable of producing the same max voltage to send through the water so that my DC signal at the end is proportional to the same 12V value across the board.

 

I am busy designing a wien bridge filter which will oscillate at a frequency of 1khz. if I do it on simulation it doesn't take everything such as temperature in to the equations. I've build it on bread board and as I got a perfect sine wave, the is absolutely no stability, i've tried to put in 2 diodes over the negative feedback for the circuit, which worked perfect on simulation, and on bread board it gives you perfect amplitude stabilization, but you don't get a perfect sine wave anymore, on the x axis on +0.7v and -0.7v the diodes switches of and then it follows the sine wave for that piece of the sine wave without diodes.
Is there any other way to get amplitude and temperature stabilization but still get a perfect sine wave.

 

Is there any other way to get amplitude and temperature stabilization but still get a perfect sine wave.

The textbook love talking about diodes for amplitude stabilisation, but they always seem to forget to point out that it distorts the signal rather a lot. JFETs can do slightly better, but they're still not as good as light bulbs or light dependent resistors. See page 43 onwards:
http://worldtracker.org/media/library/Electronics%20and%20Communications/Electronics%20ebook%20collection%20II/WILLIAMS,%20J.%20%281991%29.%20Analog%20Circuit%20Design%20-%20Art,%20Science,%20and%20Personalities.pdf

 

First, search the innergoogle for 'wein bridge oscillator schematic' to get a ton of sample circuits to chose from.

Second, a wein bridge oscillator is good for low distortion. But there are other sinewave oscillator circuits that are more stable and less finnicky, such as the phase shift oscillator.

Or you could generate a squarewave with a couple of gates and lowpass filter the doors off of it. The page linked in #10 shows that the output of the oscillator is attenuated by 99% before driving the sensor/gain loop. If you want to get away from all of that messy opamp goop, you can do the oscillator and filter in a single CMOS hex inverter.

ak