[time-nuts] Testing frequency using NTP

David M. Witten II wittend at wwrinc.com
Mon Nov 3 01:09:00 UTC 2008


Poul-Henning Kamp wrote:

#--

Actually, the best hardware for the job is the Soekris NET4501
which can timestamp a signal to a quarter microsecond.

That's why we use it for NTP servers in the first place.

All the software you need is in the FreeBSD kernel, and there
is a handy userland program to get the measurments into ascii
format (the earlier mentioned ppsapitest program).

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

#--

I am familiar with your articles relating to the use of the Soekris 
NET4501 and its characterization when used with FreeBSD for timing 
purposes.  I am fortunate to have one on hand, but have some questions.

Can I still expect to see similar performance using more recent versions 
of FreeBSD than you originally used?

Will a usable build of FreeBSD 7.x run in the memory available on the 
older Soekris boards? (64 MB RAM)

I understand that there are features that made boards incorporating the 
133 MHz AMD ElanSC520 processors suitable for this purpose.  Are these 
features still present in the newer AMD Geode LX based boards?

Though I know that you have worked with Soekris products a lot, is there 
any reason that the same techniques cannot be applied to the WRAP and 
ALIX boards from PCEngines?  or perhaps an One Laptop Per Child (OLPC) XO?

I would appreciate any information that you can provide me.


Dave Witten



More information about the time-nuts mailing list