[GPS_Standard] No joy here

[Dave Platt]

On 02/14/2012 07:23 PM, Chris Howard w0ep wrote:
> 
> How do you make a graph like that?  do you just use the log file data?

Indirectly, yes.

I get the data into a log file by one means or another... at the
moment I have a perl script communicating with the serial port,
doing the sampling-interval adjustments, and sending everything
which comes from the FLL controller into the Linux "syslog"
mechanism.  The system's syslog daemon writes it out to a file,
and time-stamps each line.

Periodically, I run another perl script which reads the file,
tears apart the lines into the various data fields and the
date/time information from syslog, does the long-term
averaging computation, and writes all of the data out in
a comma- or tab-separated format.

This converted data file then goes into the "gnuplot"
program, via a script which specifies the entries on each
line to be plotted, the scales, and the plotting styles.
Currently I have it creating a .png file, and I can also
create PDF.

The compute-and-generate script runs once an hour, so I
can keep an eye on how the frequency standard is behaving
and judge the effect of my various tweaks to the parameters.

> I have a separate supply for my +/- 5 volt DAC.
> I used the jumpers  so that I'm not
> using the on-board 5 volt regulator for my + Rail.

Unfortunately, that doesn't actually isolate you from
loading and temperature effects on the main +5 rail.
If you take a look at the schematic:  the PIC microcontroller
produces a pulse-width-modulated signal on pin 24 (this
alternates between ground, and the PIC's +5 supply
voltage).  This is low-pass filtered by R4/C6, and then
goes into U5, which buffers it and (optionally) performs
a level-shift for a bipolar-rail setup such as yours.

The use of the separate voltage rails doesn't start to
take effect until the signal reaches U5, which is after
it has been low-pass filtered (averaged) by R4/C6.
If the PIC's +5 rail moves around due to changing loads,
the PWM signal voltage will, as well, and thus the
low-pass-filtered signal going into U5 will change in the
same proportion.  The use of separate voltage rails for U5
gives you extra flexibility in meeting the tracking-voltage
requirements of your DAC, but it's not a panacea.

Avoiding this effect would have required more circuitry,
I think.  It would be necessary to have the PWM signal
"jump across" into a different (separately-regulated) power
domain while it was still digital, before the low-pass
filtering.  An extra logic gate, a comparator, or even
just a single-transistor inverter could probably do this.

In the circuit as it is today, the averaged DAC voltage can
shift around due to changing loads on the 7805 (limited
load rejection) or temperature (the 7805 has a tempco of
around -85 parts per million per degree C - it's by no means
a precision device).

Even with all of those perturbing effects, the VE2ZAZ design
works very nicely... it's a real kick to be getting better
than 1 part per billion of frequency accuracy from a little
box sitting in my garage shop!  Bert, thanks again for
creating and sharing this design!