[time-nuts] questions on uncompensated crystal oscillators

Magnus Danielson cfmd at bredband.net
Wed Jul 5 14:58:36 EDT 2006


From: SAIDJACK at aol.com
Subject: Re: [time-nuts] questions on uncompensated crystal oscillators
Date: Wed, 5 Jul 2006 14:12:48 EDT
Message-ID: <3b9.4f1fe3a.31dd5b20 at aol.com>

> Hi John,

Hi,

> great idea to compensate for temperature using the varicaps in a  
> free-running system.

Recall that AT-cut crystals have a third-degree temperature-to-frequency
curve.

> The basic design was for locking an MPEG video stream to a  broadcaster's 
> 27MHz master clock using the MPEG time stamps in a digital PLL  loop.

Actually, when running in such an application you can autotrim very cheaply.

The naive approach would be just to lock it (through a software loop) and then
use holdover (i.e. just keep the last frequency correction prior to breaking
the loo).

A little more advanced approach would be to correlate the needed correction
with that of the temperature. You can either do a lock-up table or better yeat,
figure out the 4 unknowns in the third-degree equation

f = a*T^3 + b*T^2 + c*T + d

and then use that as a separate correction method. Whenever you are tracking,
you also record the required total correction for a few different temperatures.
When you have recent measures for four different temperatures, you can solve
the equation.

T may be any unit really, so it hasn't have to be scaled to match "real"
scales. It should just be whatever the temperature measurement cranks out.

> When free-running (eg playing back video from a hard disk etc), the  
> temperature compensation should work great to keep the 27MHz within specs.

Indeed. By providing the temperature compensation in combination with tracking,
the necessary tracking-dance will be lowered.

Cheers,
Magnus



More information about the time-nuts mailing list