[ANSS-netops] Altus/K2 timing issue

[Jon Rusho]

We've run into some glitches with getting the fix merged into the 
Earthworm svn repository.   The fix is pretty simple--check to see if 
the data is timestamped between Nov 1998 and July 2018 and add an offset 
if it is.   The code snippet for outptthrd.c is below, with the existing 
lines before and after.   This has been tested in k2ew version 2.43, on 
Ubuntu 14 32-bit.   We're still trying to get it to work with the latest 
& greatest 64-bit version of Earthworm.

----------------
       starttime = (double)(timestamp + K2_TIME_CONV) + (double)msec * 
0.001;

    /* check for 1024 week rollover glitch 2018/2019   jr 01jul2018 */
       if (starttime > 911143558 && starttime < 1530403744 )
       {
         starttime += 619315200;   /* add 1024 weeks of seconds */

       }


       /* Check packet time; reject packets older than 1981, which would 
mean
----------






On 7/2/18 2:49 PM, Jon Rusho wrote:
> We've done some testing with an impacted K2 using software-based time 
> correction and a Nanometrics Centaur.   Both were fed the same 
> square-wave input signal and compared.
> Attached are two tracings from Swarm.   The left one is from the K2 
> with software correction, the right from the Nanometrics.   The red 
> vertical line shows the same time in both windows.
>
> For Earthworm users, I'm working with ISTI to get the code change put 
> into the svn repository.
>
> Note: the reason this rollover glitch hit early is based on the last 
> firmware revision from Kinemetrics.   The firmware fix for the 1999 
> rollover was put out 40 weeks before the actual rollover.   That fix 
> expired 40 weeks before the upcoming actual rollover in 2019.    Also, 
> this doesn't seem to impact all K2 and Etnas.   20 of our K2's were 
> hit, but another 20+ are fine.
>
>
> --jon
>
> On 7/2/18 9:44 AM, Jon Rusho wrote:
>> UUSS experienced an early roll-over of the GPS 1024-week counter in 
>> many of our Kinemetrics K2 instruments this weekend.   Looking at 
>> IRIS, this appears to have also impacted other networks.   The 
>> problem manifested itself as waveform data being timestamped in 1998 
>> instead of 2018.
>>
>> I'm testing a software fix in the k2ew Earthworm modules.   For data 
>> that comes in timestamped between 1998 and 01-July-2018, I shifted 
>> the time by 1024 weeks (619315200 seconds).
>>
>> My question to all of you, is this the correct shift or am I missing 
>> something?
>>
>>
>> More info on the GPS rollover: 
>> https://spectracom.com/resources/blog/lisa-perdue/2018/gps-2019-week-rollover-what-you-need-know
>>
>>
>>
>> Thanks
>> --jon
>>
>>
>>
>>
>
>


-- 
===================================
Jon Rusho
Seismic Network Engineer
University of Utah Seismograph Stations

jon at seis.utah.edu
801-585-5523