[time-nuts] New WWVB format...

[Chris Albertson]

>         DSP would be good, although I also think an microcontroller
> implementation could be interesting.  Atmel's ARM MCUs look like they'd
> be good candidates for this sort of thing.


The first stepis to simply use a regular PC, maybe running Linux.
This is the easiest and fastest platform to develop on.  It gets
harder and takes longer if you use a smaller and more esoteric
platform like a DSP or FPGA.  Using a quad core Intel chip is gross
over kill but it allows for quick development.  Later with working
software you have actually measurements in hand and can pick a "right
size" processor.

The wrong way to do it is to select hardware before you even have the
basic sketch of the software algorithm. The best way is "software
first".  You don't even need an RF front and.  You can simulate what
the D/A converter would see based on the transmitter spec and feed
that to your software before you even put up an antenna.   I worked on
a project where we actually did that.  We built a simulation of the
transmitter first.   At first the sim was very crude and simple.
Later we added noise, multi path  and so on.  This allowed the radar
receivers to be tested without special hardware and most importantly
it allowed regression testing of the software after every upgrade.

If you need a stating point a time code generator that outputs the
"old" WWVB signal is available as part if the NTP source code
distribution in a /test directory.  Is was written to help test NTP's
time code reference clock.  NTP has a way to connect a time code that
is at baseband to audio "sound card" and use it as a reference.  The
test software produced the "old format" baseband.    I'm saying this
so that no one wastes time reinventing wheels.


Chris Albertson
Redondo Beach, California