[hpsdr] Gibraltar fanning out clock and 1pps to HORTON/CASMIR/MERCURY/..

[Alex]

Analog Devices have some nice 2 port to 6 port clock distribution modules. I was going to use one when I was trying to squash everything into 1 card for CASMIR.

----- Original Message ----
From: Lyle Johnson <kk7p at wavecable.com>
To: Eric Blossom <eb at comsec.com>
Cc: Achim Vollhardt <avollhar at physik.unizh.ch>; hpsdr at hpsdr.org
Sent: Monday, July 10, 2006 11:19:21 AM
Subject: Re: [hpsdr] Gibraltar fanning out clock and 1pps to HORTON/CASMIR/MERCURY/..

***** High Performance Software Defined Radio Discussion List *****

Hello Eric!

>> I would like to remind people about the coherent system approach Bob and
>> others mentioned some while ago:
>>
>> IF we manage to derive all signal path clocks from a single source
>> (let's assume a 10 MHz GIBRALTAR for now), this system will be able to
>> PRECISELY time any recorded signal to the sub-100 ns level!
> 
> Coherent processing will be very useful.  However, sub-100 ns is way
> too sloppy if you're dealing with 100 MHz sampling clocks.  That could
> be 10 samples!  Accounting for aperture jitter at those rates, the
> specs get even tighter.
> 
> It seems to me that for any serious coherent processing, clock
> distribution is *not* going to happen across the ATLAS bus, but rather
> through equal length coax or LVDS/LVPECL in some kind of star configuration.
> 
> If I'm wrong, please set me straight, and explain to me how you plan
> on handling a system with say 4 MERCURY's, or 4 HORTONs for that matter.

I suspect that for most uses, distribution via the backplane will be 
sufficient.  We need to think about the way in which we input the 10 MHz 
clock to minimize jitter to reasonable levels.

For really precise applications -- such as multi-node coherent systems 
-- we'll need something better.  Perhaps one or more 50-ohm output ports 
is the right way, or perhaps an LVDS or similar connector is appropriate.

I suspect we should be able to get much, much better than 10 nsec on the 
Atlas, and the increasing from slot to slot should be on the order of 
100 picoseconds.  One could put two Mercuries in the backplane, on one 
either side of Gibraltar, and have very little skew between the two 
boards,and a fairly stable -- and measurable - offset from Gibraltar.

If we use a 50-ohm system, we can use splitters to get as many nodes as 
we need, as long as we calibrate the splitter.  We could conceivably use 
an extra port from it to measure the phase shift and then compensate for it.

Exciting stuff!

73,

Lyle KK7P

_______________________________________________
HPSDR Discussion List
To post msg: hpsdr at hpsdr.org
Subscription help: http://lists.hpsdr.org/listinfo.cgi/hpsdr-hpsdr.org
HPSDR web page: http://hpsdr.org
Archives: http://lists.hpsdr.org/pipermail/hpsdr-hpsdr.org/




 1152548063.0