[time-nuts] Improving the stability of crystal oscillators

Don Collie donmer at woosh.co.nz
Wed Oct 17 00:16:45 EDT 2007

Don Collie wrote:
> Gidday Bruce,
>    Yes, the temperature sensor [a diode or thermistor, etc] would be
> placed in thermal contact with [the plate attached to] the cold side
> of the pile [there would be a finned heatsink attached to the hot
> side, and an expanded polystyrene sheet between the two sides, as much
> as possible]. The voltage developed across the sensor would be fed to
> one input of a comparitor [or what have you - PID controller etc].
> There would be a voltage reference applied to the other input so as to
> make a conventional negative feedback control loop. If the loop gain
> was high enough, the system would operate in switch mode to maintain
> the cold side at a constant temperature. If the gain is lower the loop
> will operate in linear mode [ie : non oscillatory, but sluggish].
>    The advantage of a high gain loop is that there is no setting up
> necessary [self adjusting] - the thing will find the right
> temperature, and then cycle on/off to maintain it. [providing the
> temps involved are within the compliance of the system [Peltier,
> ambient temp.,etc]]
>     Does a PID control loop *always* out-perform an oscilliatory loop?
> [or does that depend on................................................?]
>    Another way to reduce the change of temperature at the crystal [and
> circuitry], would be to put a substance with thermal resistance
> between the plate [cold side], and a mass of copper or aluminium,
> which housed the crystal etc. This oven would have to be lagged.
> Please correct me if I`m wrong, but as the comparitor gain is
> increased, both the frequency of the temperature variation [at the
> sensor], and its amplitude are reduced.
>    A PID controller would be the *quickest* way to stabilise the
> temperature, but this is a secondary concideration : the primary
> concideration would be temperature stability - it wouldn`t really
> matter if it took an hour or two to stabilise].
>    If the object of the exercise is to make a *single* "oven", The
> current supplied to the Peltier by the comparitor could be arranged to
> be reversable, depending on whether the temp at the sensor was more,
> or less than the set temp. Thus it would be possible to maintain the
> 25 degree temp, as the ambient temp changed from [say] 0 to 70
> degrees. Not quite sure what to do about the [diode-like] barrier
> voltage of the Peltier pile.
>    Do crystals do better/last longer/less ageing at lower
> temperatures? - you would think so.
> Cheers,..............................................Don.

A temperature control system without an integral term will always have
an offset from the desired setpoint.
The offset will depend on the ambient temperature and the gain.

Depending on the amount of heat being pumped a convection cooled
heatsink will probably be inadequate.
At least a blown heatsink (or better yet a water cooled heatsink) may be
necessary if the heat being pumped is substantial.
For example, try using a 70Watt single stage Peltier module on an
convection cooled heatsink to cool the to surface to below freezing.
At first ice forms on the cold plate, but gradually as the heatsink
warms up the ice melts and the "cold" plate becomes fairly warm.

It may be possible to use a cpu heatsink with integral heatpipes, at
least some of these allow a reasonable thickness of insulation to be used.

The usual technique is to actually sense and control the Peltier device
Either a high frequency switchmode current controller with an output
filter to reduce the Peltier ripple current or an analog driver can be used.

Surely the oscillation frequency depends primarily on the thermal
transport delays within the system?
If self oscillating temperature controllers are so effective why are
they no longer used for precision temperature control?

If the control loop responds too slowly then it wont effectively correct
the effects of ambient temperature changes.

Placing a large thermal resistance between the cold plate and the oven
mass is somewhat counterproductive.
The oscillator will dissipate power that has to be removed, limiting the
maximum usable oven mass to ambient thermal resistance.
If this thermal resistance is too large most of the heat flow will be
via the oven thermal insulation at which point temperature control tends
to be somewhat ineffective.

Crystals operating at cryogenic temperatures were at one time found to
age very slowly.
However the effect of temperature on aging depends on the aging
mechanism(s) involved.

Peltier devices used to regulate the temperature of large baseplates
typically achieve a temperature stability of around 0.1C.

When used to control the temperature of smaller objects like laser
diodes a baseplate temperature stability of 0.001-0.002C has been achieved.
Objects of intermediate size like some ECDLs routinely achieve a
stability of 0.01C or better when a Peltier cooler is used.


----- Original Message ----- 
From: "Bruce Griffiths" <bruce.griffiths at xtra.co.nz>
To: "Don Collie" <donmer at woosh.co.nz>
Sent: Wednesday, October 17, 2007 4:35 PM
Subject: Re: [time-nuts] Improving the stability of crystal oscillators

> Don Collie wrote:
>> Dear Dr Bruce,
>>    I am partially depantsed [*only* partially]. Do I have your
>> permission to respond to your points on group? This way either you, or
>> I will be fully depantsed.
>> All the best,...........................................Don.
> Don
> That would perhaps be a good idea in that it may be educational to
> others and any delusions etc may be cured.
> Bruce 

More information about the time-nuts mailing list