[ntp:questions] Using different timebase for ntpd

Greg Dowd GDowd at symmetricom.com
Tue Mar 24 16:17:20 UTC 2009

Most if not all commercial ntp appliance manufacturers have some sort of
hardware support, either RTOS, custom clock, better oscillator or even
modified Ethernet devices to support hardware time stamping.  Truetime
was doing hardware gated ntp packets in the 90s.  I tried  modifying one
of our SyncServers and gave a presentation at ietf tictoc bof on
precision frequency transfer over packet based networks. Dave Mills also
did some experimentation with this.  While I used mode 3/4 with a
lagging timestamp for the followup function, Dave restricted that
operation to symmetric modes. I have another presentation from a sync
conference last year comparing the two protocols.  Let me know if you
are interested and I'll dig them up.  Also, commercial product that does
hardware timestamping is available from us (Symmetricom).  I'm sure
there are others out there.  Brilliant Telecom has a ntp box that must
use hardware timestamping.  Maybe Meinberg as well.  Of course, where
the rubber really meets the road is recovering a signal.  Most of the
products I mentioned above are optimized for generating a synch signal.
Their performance as client devices varies widely.

Symmetricom TimeHub NTP Blade
Symmetricom SSU-2000 NTP Blade
These are expensive solutions but they are both accurate and stable.  

In your case, you are not talking about hardware timestamping but
software timestamping using a hardware clock.  It is a good first step
but os/stack jitter is still likely to be dominant.  Also, ntp puts a
few more constraints on protocol operation and clock recovery than ptp.
While this lets it operate in more widely varying network environments,
it often restricts accurate clock recovery.  If you get to hardware
timestamping, there are a few tricky things that arise.  Depending upon
which 1588 profile you use (I use annex D), both ntp and ptp are layer 3
protocols.  That means a hardware timestamp is really a layer violation.
PTP defines the timestamp as the beginning of the first bit following
the preamble.  NTP doesn't define it (I've tried) but Dave has a page on
his website that talks about timestamping.  He uses a different approach
(which corresponds to ietf ippm and ITU.T Y.1540) where the timestamps
are at the start of transmission and the end of receipt.  These two
approaches have different strengths and weaknesses.

BTW, IEEE has just in the process of creating a study group to define
Ethernet timestamping.  This is being driven due to the 802.1AS
community request.

-----Original Message-----
From: questions-bounces+gdowd=symmetricom.com at lists.ntp.org
[mailto:questions-bounces+gdowd=symmetricom.com at lists.ntp.org] On Behalf
Of Patrick Loschmidt
Sent: Tuesday, March 24, 2009 3:47 AM
To: questions at lists.ntp.org
Subject: Re: [ntp:questions] Using different timebase for ntpd


Thanks all for the answers so far.

> David Woolley wrote:
> As I understand it, he is replacing the kernel clock's role, so he 
> needs to replace all the system calls that read and discipline the 
> kernel clocks by ones that read and discipline the special hardware.

Correct. I have a PCI network card which runs it's own internal clock. I

made a literature research and couldn't find anyone implementing NTP 
with hardware support for timestamping. Since I can modify the FPGA 
based card, my first step is to get NTPd discipline the clock on the 
card, that's why I have to replace all the system calls to the kernel 
clock ...

Richard B. Gilbert wrote:
> It might be useful to know exactly what hardware is in use.  Some of
>  fancier clocks have provision for GPS input.  I believe that both 
> Meinberg and Symmetricomm offer such.  I believe that both companies 
> offer driver software for these clocks allowing you to read and set
> and, perhaps, offering other functions as well.
Normally, I would just use a normal mainboard with a special oscillator,

but that doesn't solve my problem. As far as I could find out, nobody so

far modified the clock to be disciplined in NTP ...


questions mailing list
questions at lists.ntp.org

More information about the questions mailing list