[time-nuts] Improving the stability of crystal oscillators
bruce.griffiths at xtra.co.nz
Thu Oct 11 01:08:43 EDT 2007
Neville Michie wrote:
> I have made several ovens for oscillators over the years.
> The recipe is:
> get a piece of aluminium big enough to contain the oscillator,
> voltage regulator and first stage amplifier.
> With a mill remove the shapes of each component.
> Bolt a large power transistor, large power fets are best, to the
> outside of the block as a heater, and it is run
> off the unregulated input power. Judicious selection of a component
> decides the start-up current.
> Make a plate to cover the excavation for the components, and bolt it
> The circuit can be made with discrete transistors in the most
> unstable looking amplifier ever seen,
> alternating NPN and PNP transistors, connected directly to each other
> with load resistors.
> The main temperature sensor is a resistor bridge with a high value
> glass encapsulated thermistor.
> These are available a several trade houses. The amplifier is also
> temperature sensitive, but is within the thermal loop.
> The thermistor bridge gives a very large signal ~ 50mV per degree.
> Gain may have to be backed off if thermal
> oscillations do not die down, but the metal block acts as an
> integrator and the circuits are very easy to get
> high gain and sensitivity.
> The whole block is packed in two inches of foam insulation, my 1MHz
> oscillator only draws about 80 mA at 12 volts.
> The temperature is set to 40 C.
> The stability of the oscillator is very good, but as I have not yet
> got a disciplined oscillator going I dont know which is drifting,
> the HP 10811 in my frequency counter or the 1MHz oscillator. After a
> year, the difference is currently 0.3 ppm.
> cheers Neville Michie
A correctly tuned PID control loop should allow even tighter temperature
A "boostrapped" oven like that used by Wenzel should be even better.
Despite Wenzel's claims this type of oven isn't new it was used for
portable standard cell enclosures decades ago.
More information about the time-nuts