Like this (experiment):

$ ./t.pl
UTC original = Wed Mar 22 20:28:27 2000
UTC adjusted = Wed Nov  6 20:28:27 2019

$ cat t.pl
#!/usr/bin/perl

use POSIX qw(strftime);
use Time::Local;

my @original = (27,28,20,22,3-1,2000-1900,0);

print "UTC original = ", strftime("%a %b %e %H:%M:%S %Y",@original),"\n";
my $jdt = timegm(@original);
my $ajdt = $jdt + (1024 * 7 * 86400);

my ($sec,$min,$hour,$mday,$mon,$year,$wday,$yday,$isdst) = gmtime($ajdt);
print "UTC adjusted = ", strftime("%a %b %e %H:%M:%S %Y",($sec,$min,$hour,$mday,$mon,$year,$wday,$yday,$isdst)),"\n";

@Tamás: Does the above adjustment show the correct date/time (Wed Nov  6 20:28:27 2019 UTC)?

Regards, Mark.

On 7 Nov 2019, at 2:13 PM, Mark Webb-Johnson <mark@webb-johnson.net> wrote:


The issue is that we don’t have week numbers from the GPS (and presumably week to date is non-trivial).

We get the date/time from the NMEA RMC message, then pass it into this to convert to a julian date/time:

/**
 * utc_to_timestamp:
 *  convert GPS date & time (UTC) to timestamp
 *   date: "ddmmyy"
 *   time: "hhmmss"
 */
static unsigned long utc_to_timestamp(const char* date, const char* time)
  {
  int day, month, year, hour, minute, second;

  day = (date[0]-'0')*10 + (date[1]-'0');
  month = (date[2]-'0')*10 + (date[3]-'0');
  year = (date[4]-'0')*10 + (date[5]-'0');

  hour = (time[0]-'0')*10 + (time[1]-'0');
  minute = (time[2]-'0')*10 + (time[3]-'0');
  second = (time[4]-'0')*10 + (time[5]-'0');

  return
    (JdFromYMD(2000+year, month, day) - JDEpoch) * (24L * 3600)
      + ((hour * 60L + minute) * 60) + second;
  }

Perhaps sufficient to add 1024 * 7 * 86400 to that?

Regards, Mark.

On 7 Nov 2019, at 1:44 PM, Stephen Casner <casner@acm.org> wrote:

Mark,

You don't need the condition "date <2019-11-03", just always add 1024
weeks to the current time value coming from the modem.

                                                       -- Steve

On Thu, 7 Nov 2019, Mark Webb-Johnson wrote:

Damn:

https://www.cika.com/soporte/Information/GSMmodules/GPS-week-rollover_Simcom.pdf <https://www.cika.com/soporte/Information/GSMmodules/GPS-week-rollover_Simcom.pdf>

I think we can do a fix in the SIMCOM module that if the date <2019-11-03 then offset it somehow? Any ideas?

Regards, Mark.

On 7 Nov 2019, at 1:09 PM, Stephen Casner <casner@acm.org> wrote:

This page mentions the problem:

https://techship.com/news/gps-week-roll-over-issue-during-epoch-restart/

Quoting there:

  For the Qualcomm MDM9x15 series (4G/LTE Cat 3) and MDM6200
  (3G/HSPA+) series chipsets there is a offset used, so the GPS week
  roll over will occur 1054+1023 weeks since launch of the GPS, and
  more precisely in beginning of November 2019 (UTC 23:59:42
  November 2, 2019)

  [snip]

  3G/HSPA+ modules based on Qualcomm MDM6200 series chipsets:
  Huawei EM820W / MU609 series
  Telit DE910 series
  Simcom SIM5320 series

This sounds like the hardware/firmware won't be fixed.  Perhaps the
solution for OVMS is just to put an offset as a workaroud.  That will
work for the next 1024 weeks.

                                                      -- Steve

On Wed, 6 Nov 2019, Stephen Casner wrote:

Sorry, I did not read carefully.  The item you mentioned is just an
antenna.  The GPS implementation is in the modem module, correct?
That's where the correction is needed.

                                                      -- Steve

On Thu, 7 Nov 2019, Tamás Kovács wrote:

I use a universal gps connected to ovms, not the car own gps. (
https://www.fasttech.com/products/1020200)

Stephen Casner <casner@acm.org> ezt írta (időpont: 2019. nov. 7., Cs, 0:42):

Tamás,

This looks like the GPS sensor did not handle the rollover of the
10-bit week-number field in April, 2019.  So the timestamps are off by
1024 * 7 * 24 * 60 * 60 seconds.  Here is the US government alert about
this problem:


https://www.us-cert.gov/sites/default/files/documents/Memorandum_on_GPS_2019.pdf

This problem was recently discovered for the Garmin GPS 18x LVC
sensors in the Tesla Roadster 2.5 (but not in the earlier 1.5 version
like I have, which has the older GPS 18 LVC sensor):


https://teslamotorsclub.com/tmc/threads/caution-annual-service-car-power-down-gps-issue.163328/

There is a firmware update to fix this for the GPS 18x LVC.  Perhaps
there is one for the GPS sensor in the Mitsubishi as well.

                                                      -- Steve

On Wed, 6 Nov 2019, Tamás Kovács wrote:

I have a problem with utc time, the time is correct but the date is not
valid. I use 3.2.005-99-g88cfeef9 version with modification of Mitsubishi
i-MiEV.

OVMS# me li utc
m.time.utc                               953756768Sec
I (900675) housekeeping: 2000-03-22 21:27:06 CET (RAM: 8b=82812-86452
32b=27308)
I (981115) webserver: HTTP POST /api/execute
I (981115) webcommand: HttpCommandStream[0x3fa32958]: 1984428 bytes
free, executing: time
OVMS# time
Time Zone:  CET-1CEST,M3.5.0,M10.5.0/3
UTC Time:   2000-03-22 20:28:27 UTC
Local Time: 2000-03-22 21:28:27 CET
Provider:   gsm-nmea

PROVIDER             STRATUM  UPDATE TIME
*gsm-nmea                  2       0 Wed Mar 22 20:28:27 2000


--
Üdvözlettel:
Kovács Tamás_______________________________________________
OvmsDev mailing list
OvmsDev@lists.openvehicles.com
http://lists.openvehicles.com/mailman/listinfo/ovmsdev



--
Üdvözlettel:
Kovács Tamás
_______________________________________________
OvmsDev mailing list
OvmsDev@lists.openvehicles.com
http://lists.openvehicles.com/mailman/listinfo/ovmsdev

_______________________________________________
OvmsDev mailing list
OvmsDev@lists.openvehicles.com
http://lists.openvehicles.com/mailman/listinfo/ovmsdev