Hi

I've been trying to temperature stabilise a 555 square wave oscillator (astable) at around 20Hz.

I have chosen the SE555D as, I believe, it has the smallest negative temperature coefficient of all the 555s. I used 3 x 0.22 COG (NPO) SMD caps in parallel, a 39K fixed with a 3K3 and a 3K9 linear PTC thermistors in series. (see circuit diagram attached)
For a 50/50 square wave output Pins 6 and 2 are connected together the resistor chain is to pin 3 and the caps to earth, pins 7 and 4 are not connected, pins 5 and 1 have a 0.1 COG (NPO) SMD cap across and pin 5 has a 0.1 COG (NPO) SMD cap to earth. I have used a 1N4148WS in the power supply for reverse polarity protection.
This configuration gives me the stability I require - 20.05Hz at ambient, 19.92Hz, at +125 deg C. and 19.87Hz at -20 deg C on a single sided circuit board the size of a postage stamp.

I’m pretty sure if you play around with the PTC thermistor values you could get even better stability.

 

If you want better stability use a crystal controlled oscillator.

 

G'day MrChips
Never found a crystal controlled oscillator at 20Hz. When I did work with them (in radio transmitters) 40 years ago we had to have the crystals in 'ovens' to keep the frequency stable. Any improvement since then?
Stability is fine for this use - the recipient will tolerate around a 2 Hz deviation from the expected frequency input but not much more.
This is to drive an automotive circuit at battery voltage and must operate at the extreme range of temperatures found in the engine bay of a vehicle.
K.I.S.S. It uses only 10 components and $$$s are important in manufacturing and sales.

 

G'day MrChips
Never found a crystal controlled oscillator at 20Hz. When I did work with them (in radio transmitters) 40 years ago we had to have the crystals in 'ovens' to keep the frequency stable. Any improvement since then?
Stability is fine for this use - the recipient will tolerate around a 2 Hz deviation from the expected frequency input but not much more.
This is to drive an automotive circuit at battery voltage and must operate at the extreme range of temperatures found in the engine bay of a vehicle.
K.I.S.S. It uses only 10 components and $$$s are important in manufacturing and sales.

Yes, there have been improvements, like the development of integrated OCXO's (oven controlled crystal oscillator) that will give you excellent stability... for a price.
I suggest you use a simple crystal oscillator like this one, and then add a frequency divider circuit afterwards... that way you would get the 20 Hz you want, and with a high degree of precision too.

 

- the recipient will tolerate around a 2 Hz deviation from the expected frequency input but not much more.

I would be surprised that a 555 is not stable to within 10% of the target frequency, if you choose temperature-stable timing components. I know they drift around but I don't think it's anywhere near ±10%.

 

You don't need a temperature controlled crystal oscillator for what you need. A simple crystal oscillator will be stable to better than 10 parts per million. This is at least 1000 times better stability that a 555 timer circuit.

A 10.5MHz crystal divided by a 19-stage binary counter will get you very close to 20Hz.

A simpler solution would be to use a single chip MCU with any crystal and you can get exactly 20Hz, stable and accurate to 10ppm.