SyncServer GPS (GT-8031) alternatives?

[N1ZYY]

Hi there!

I recently bought a Symmetricom SyncServer S350 for what I naively thought was a great deal, only to learn what has been discussed here a few times: the GPS week number rollover issue has left it unable to get time from GPS. In trying to figure out what's going on, I found EEVblog a few times. The older posts weren't specifically about this issue, so I wanted to start a new thread rather than reviving an old post.

The GPS board I have has the Furuno GT-8031H. Someone in a previous thread had linked to https://furuno.ent.box.com/s/fva29wqbcioqvd6mqxn5rt976dkaxudj (embedded PDF) which shows that it ran into trouble 2022/09/18, but it also notes "The receivers can update the date up to year 2079 by entering the date with ZDA command." It does not elaborate, but it looks like that's an NMEA command. I inferred that if I was able to connect and set the date, it would then correctly keep it updated.

This post: https://www.eevblog.com/forum/testgear/possible-to-make-a-quality-universal-timercounter-using-$12-pic-countergpsdo/msg1353557/#msg1353557 attached the hardware specs and protocol specs, so I briefly set out to see if I could figure out how to connect to it.

I realized (photo attached) that the GPS module is the Furuno mentioned, but it's on a Pentar SA101 board. There's an extra IC on the underside of the board, and from what I found online, it sits between the SyncServer and the GT-8031 and makes it appear to be an Oncore GPS speaking Motorola's binary protocol. That would seem to throw a wrench in my plans to connect and send the "ZDA" command.

A copy of the old Symmetricom field service bulletin from 2020 warning about the issue that was shared with me notes:

It is possible to reprogram the Furuno GPS receiver,
but only if the chip is removed from the PCB, reprogrammed and reattached to the
PCB. There is no way this operation can be conducted in the field.

I was initially skeptical of this, but based on what I later learned about the Motorola emulation, I think they really did have to desolder the Furuno unit. It's not clear if there was actually an alternative firmware, or if they were just setting the reference time (via the ZDA command referenced above). I'm also not sure if there's a battery to store that clock setting, or if it would be at risk of reverting. A more recent service bulletin does not offer this service, and just advised users to disconnect the GPS antenna permanently, and consider upgrading to an S650.

In https://www.eevblog.com/forum/metrology/symmetricom-syncserver-s350-furuno-gt-8031-timing-gps-gps-week-rollover/msg4544237/some people report that the GT-8736 is a drop-in replacement. However, there's mention of it not working fully, and in any event, the GT-8736 looks to have been discontinued and is not available anywhere either.

I'm curious if anyone has gotten further, either with trying to get the GT-8031 working, or with finding a GPS receiver board that works properly as a replacement. Right now it's just providing stratum 2 NTP to my home LAN which is not what it was meant to do. 

[N1ZYY]

Oh, one other thing -- while I was trying to figure out how I might connect to the GPS with minicom or something, I looked at the NTP docs for the OnCore driver: https://doc.ntp.org/documentation/drivers/driver30/

However, there was no /dev/oncore.*, and ntp.conf whows 127.127.45.0, which would seemingly correspond to a Spectracom PCI card: https://doc.ntp.org/documentation/drivers/driver45/

That's not present in /dev/ either. I'm guessing they're doing something custom here and the upstream NTP docs don't cover everything?

[GigaJoe]

maybe ....

in overall GNSS time server usually need to synchronize events happening in a different places in the same time
consider NTP in overall has a multiply software and hardware layers to calculate a time in the best case you have a deviation of fraction milliseconds, on recipient side
to deal much more precisely are complex hardware equipment as direct time source  accessible as data in a memory.

as an example 24 hours Stratum 2 from S1 ,over UDP


requirement for logging are 1mSec so it in overall fit my purpose ...

i may suggest to use 10Mhz for overall frequency usage for your lab
and 1 PPS signal will fit a dedicated NTP server that sync over Internet , but use 1PPS as a precision time second source ( consider linux\freebsd with minimal working process , and PPS signal usually feed to com port)  you may get about 10-6 precision , ( but again i have no clue how to use it on another device as over Ethernet \ UDP and software layer all this evaporated)

[N1ZYY]

Thanks for taking a look at my rambling post.

I should have been a little clearer about my goal, though. I'm specifically looking to see whether anyone has gotten the GPS in these units working after the September 2022 impact of the GPS week number rollover (offset in these Furuno units from the actual April 2019 GPS rollover), or found a replacement GPS board that works in them. This is kind of dipping my toe in the time metrology pool; like you say, NTP only goes so far. The SyncServer has a PTP option, and I've got some PTP-capable NICs it'd be fun to experiment with.

I like your suggestion as a workaround, though. As I understand it, the 1PPS signal from GPS is still firing even if its idea of the date is off by a couple decades, and I've seen some implementations that combine combine NTP for coarse time with PPS to get a very accurate time. I don't currently have a need for the 10 MHz reference, but I'd expect even the undisciplined Rubidium oscillator to be pretty good.

[GigaJoe]

frankly , i have no idea of PTP linux like implementation.  I'm wildly guess it kinda unrealistic , due to internal multiply processes in common operating system as it will disturb a sensitive timekeeping.

so at least you can play with old timestamp

( i made couple years ago 10M gps for 10M source, and as side option display with time, i have to adjust on software level an incorrect time date,  but cut the corners ,and it bit drifted,  to my surprise when  turn it on a few month ago date completely shifted - and i really don't want to reprogram again, as it fairly simple, but my laziness won at 100% )