TO:      DAVE RANSOM

FROM:    MIKE GARDNER

SUBJECT: PC CLOCKS


Dave:

Last June/July I promised to try to find my notes on how to make your
PC's clock frequency super accurate.  Well, with the move & all I
finally found my notes.  What's needed is to change the circuit from a
SERIES to a PARALLEL Resonant circuit.  In most (old) PC's this ckt is
done with a few spare gates biased such that they form an oscillator
('kinda on the edge of data-book parms, I guess).  The change below
also adds a trimmer cap to tighten things down even more.  The final
ckt is phenom'ly acurate.  I easily tuned it in to 8 signif. digits
(it was fairly stable too).  Still, see my 16 July note on frequency
variation with PS voltage...  The ckts follow:


	Old (Series) Ckt         New (Parallel) Circuit
	
	        _ 32,768Hz              _ 32,768Hz
	       /                       /
	ĴĴĿ      ĴĿ
	           27pf                            
	                         27     9-35       
	        \/\Ĵ        pf     pf\/\Ĵ
	           2.2M                     22M  
	                          68K              
	      \/\
	                                      
	                                      
	                                      
	to                to       = gnd
	RTC                      RTC                 


	Key: Ĵ Crystal  Ĵ Capac.  \/\ Res   Gate/Op Amp

(The Gate circuit used in the HP Vectra is a 74LS04 Hex Inverter.)

Note that the resistor changes from 2.2M to 22M, and that a trimmer cap
of 9-35pf (cheap & easily found) has been added.  To calibrate, attach a
counter to the output ("to RTC") & adjust for 32.7680 KHz.  If the
trimmer cap won't pull the freq in, replace the crystal.  These small
"pencil" crystals are commonly available at less than $2.  (NB: This
whole design assumes you can only find Parallel-ckt xtals; otherwise...)

Clearly there are variations possible on this circuit, depending on how
the design was laid out originally & what components are accessable.
The key here, however, is that the trimmer cap now goes to ground rather
than being placed in series with the xtal.

That should otta do it...  Let's get out those soldering irons, now!

Mike

PS:  Since I've still got a few bytes, I'll repeat some warnings:

The biggest source of error with this new design will be in powering
the machine on/off.  If left on, the accuracy will be within a fraction
of a sec/yr!  When the machine is powered off, the osc. will slow down
by SEVERAL Percent!  (You could design a circuit to avoid this prob
but vendor's don't; with only 3 volts available it's difficult.)
With the improved ckt, the main source of inaccuracy is now just the
fraction of the time the machine is powered on!  Changing this ratio
will change the clock "accuracy".
