[time-nuts] Thought experiment on a low cost timing board

Tim Shoppa shoppa at trailing-edge.com
Sat Feb 25 21:31:05 EST 2006

John Pettitt <jpp at cloudview.com> wrote:
> Tim Shoppa wrote:
> > My gut feeling: back up a little bit. Figure out how to do what you
> > want without a PCI bus, without gold fingers, without BGA's, etc.
> >
> >   
> I had several goals in mind when I asked the initial question:
> 1) a low cost high stability ntp stratum 1 clock board - something that 
> when added to a sub $100 gps would yield a really stable time source for 
> ntp.  To do this it really needs to let the main cpu read the counters 
> in a predictable, consistent time which is why I wanted PCI (3v because 
> I have soekris boxes).   

Your random PC-clone motherboard does admirably when hooked up to
a halfway decent PPS source.  Jitter in the low microseconds or better.
There's probably a constant skew in the same range but measuring
that... ? A microsecond is just 1000 feet of cable!

Now you said "high stability", does that mean that you want to
holdover time well when GPS PPS isn't available?

> 2) it would also be nic to be able to add a low cost frequency synth to 
> the design.

Combine that with holdover and you have, say, a Z3801A with the 10MHz
driving a AD9951 etc. Not fancy tech and not something you need to
be messing with PCI bus or BGA's or any of that crap. (Unless you 
just really really want to mess with it, which I'm starting to think
you do!)

> What I've seen so far there are designs out there based on various cpu 
> cores and or f/c pga chips that are close to what I'm thinking.

Well, from my perspective you want a cheap PC-clone and a Z3801A
and a DDS synth locked to the 10MHz OCXO.

> What prompted this in the first place was the horrible temperature 
> sensitivity my soekris boxes exhibit - I can only keep them to within 
> 10us of the gps 95% of the time and the occasional 100us excursion is 
> not uncommon.

Gak. Is this relative to PPS from GPS? That's awful. 100us is like
an instantaneous excursion of many PPM. My cheap PC-clone motherboards
go up and down by a single PPM over a 24-hour period.

> I want to be stable to the limit of the gps I'm using - 
> for no other reason than the belief that I can do it for way less than 
> commercial devices using open source software and low cost hardware.  In 
> other works I'm hacking for the sake of it :-)

Some GPS's have PPS outputs that do sick things. I'm thinking back
to my Trimble SV6, whose PPS would go off by hundreds of microseconds
just because it didn't see 4 sats. But that's like early or mid-90's
tech and it wasn't a timing receiver!

As for "hacking for the sake of it", my tendency these days is to
keep away from PC's unless it's a last resort.


More information about the time-nuts mailing list