[time-nuts] HP 1938 revisited

Richard (Rick) Karlquist richard at karlquist.com
Sun Apr 12 23:32:35 UTC 2009

wje wrote:
> Fluke.l (China) was selling a number of 1938's on Ebay. I snagged one 
> just to have a piece of HP history.
> It works just fine, but I've noticed something a little strange.
> Comparing the 1938 to both my cesium and GPS standards, there's a 
> distinct periodic 1ns phase shift every second. Seems to smoothly 
> advance  1ns for 500 ms, then retard back to the original phase point 
> over the next 500ms. Question: is this to be expected? I'm assuming this 
> is from the AFC loop, but I would have expected it to be better damped.

The AFC loop (as opposed to EFC) is a purely analog loop in which there
is no mechanism for a 1 Hz oscillation.

Now if you are talking about the EFC, that is another story.  Depending
on how you are driving the EFC, you can pickup noise from any number
of sources.

I vaguely remember there was an LED that flashed at 1 Hz if everything
was working OK.  You might see if that is the source of what you are

> I've searched both this group and the HP group; there really doesn't 
> seem to be a great deal of info about these, other than schematics and 
> some nice variance plots on leapsecond. Just what is the serial port and 
> the PIC data lines useful for, if anything?

The serial port can be used to change the oven temperature set point and
IIRC monitor various oven parameters.  Probably nothing you want to
play with if you want a working oscillator.  If you want to experiment,
I have released the software to talk to the serial port.  It should
be archived somewhere.  In production, we ramped the temperature up
and down and found the exact turnover temperature for each individual
oscillator and set the oven to the turnover.  This required that we
make crystals that actually had a turnover.  Many 10811 crystals did
not have turnovers, only an inflection point.

Rick Karlquist N6RK
Designer of E1938A

More information about the time-nuts mailing list