[ntp:hackers] Agilent 5071/a

todd glassey tglassey at earthlink.net
Fri Mar 23 07:20:43 PST 2007


Hal -  more commentary below.

Todd Glassey

----- Original Message ----- 
From: "Hal Murray" <hmurray at megapathdsl.net>
To: "Fernando P. Hauscarriaga" <fernandoph at iar.unlp.edu.ar>
Cc: "NTP Hackers Mailing List" <hackers at ntp.isc.org>
Sent: Thursday, March 22, 2007 3:24 PM
Subject: Re: [ntp:hackers] Agilent 5071/a


>
>> Does anyone ever hacked the hpgps refclock driver for using it with
>> the  Agilent 5071/a Primary Frequency Standard?
>
> I haven't heard of any work in that area.

CertifiedTime pushed this as part of the deployment of the remote 'Certifed 
Timing Center' and its all been done. There is a CommonView based practice 
for initializing and setting the clocks already that was globally tested... 
I.e. NIST operated our systems in Tokyo for CertifiedTime for three years 
without glitch. The biggest hassle was when DHL lost one of my two 5071a's 
enroute to Tokyo. Wound up that it was sitting in a DHL depo in Narita 
Airport. Aggggggghhhhh.

>
> How does a 5071A get synced up to UTC?

Its a CommonView based process called the NIST Global Time Delivery Service 
...

>
>
> You might try connecting up the PPS signal and using the ATOM/PPS driver.

Len Cutler and the Agilent crew built a 1pps-Serial Interrupt Converter that 
I would think Symmetricom might offer as well.

I have several of them Greg if you folks want one as a prototype.

>  It
> needs a prefer-ed driver that is close to get the right second.  Normally
> that would come from the serial channel from the 5071A or Z3801A or ... 
> but a
> system on the net might work if you have a reasonable network connection. 
> At
> least to get off the ground.  It obviously breaks if either the other 
> system
> or your network connection dies and may not be good enough if your link is
> sometimes busy.
>
>
>> In addition, hpgps responds with yyyymmmddhhmmss string and 5071a
>> respongs  only hhmmss and for getting the date you have to ask for
>> Julian Date and  make the conversion.
>
> Julian Date conversion is easier than year-month-day.  It's just a 
> constant
> offset and multiply by seconds/per/day.
>
> Watch out for the midnight glitch.

Amen!

> The standard trick is to read date, time,
> date, and try again if the dates don't match.  (Or ignore the problem and 
> the
> filter will take care of it.)
>
>




More information about the hackers mailing list