[time-nuts] Re: Manual for Precision Standard Time PST-1030 WWV Receiver?

[Erik E. Fair]

That WWV radio clock brings up some memories.

I deployed the second stratum 1 NTP server on the Internet in 1989 at Apple Computer with a PSTI model 1020 WWV radio clock attached to the VAX-8650 (4.3 BSD Unix) that was "apple.com" at that time, running xntpd. Someone else at Apple had bought that radio clock, played with it in an Advanced Technology Group (ATG) lab for a while, and then abandoned it. When I found out it existed and no one was using it, I asked if I could have it for the "Apple VAX" ...

https://www.ntp.org/documentation/drivers/driver3/

Notice: driver #3 - written by Dave Mills (R.I.P.) himself. I remember the leap year DIP switches that annoyingly had to be fussed with - I wanted something I could set & forget rather than have to set a calendar(1) reminder to perform maintenance on it periodically.

I took a look at an old (but likely unchanged) copy of ntpd source code I have, and the serial port setup for all "reference clock" drivers is in ntpd/ntp_refclock.c with the specific stuff (polling command, interpretation of clock output) in ntpd/reflcock_pst.c - the comments in the code are substantially the same as in the URL above. The serial port settings are available in a range of Unix kernel API flavors: BSD, System V, TermIOs, and even some Windows thing. You should be able to get a sense of the radio clock interaction model from this code - though it looks like Dave opted for KISS: 320 lines of C code in this version. The Generic NMEA-183 driver (#20) source is 6x bigger.

I'd bet that the model 1030 is pretty similar to the model 1020.

The Apple VAX was on the top (3rd) floor of Mariani 1 (20525 Mariani Avenue, Cupertino, CA; the Apple Computer HQ building for a time) in a small machine room dedicated to it, along with the Cisco router & other network gear to connect to the Internet. There was a west-facing window overlooking the parking lot, and WWV signal was good enough in the room that I didn't have to run an antenna outside.

On July 11, 1989, Dave Mills goofed a fuzzball software release, and the fuzzballs stopped serving time to everyone other than themselves. I noticed, poked around that night (Dave fixed the bug the next day), and had an idea:

https://www.clock.org/clock.org.html

Others have had variants of this idea and actually implemented some of it at ntp.org, but none so far that I've seen have tried to optimize for minimum hops between NTP servers/clients (minimize remote clock jitter induced from variable network latency) and dynamically configured NTP peering for resiliency against disaster-induced disconnection, something I can tell you about because in October 1989, Apple (and I) suffered the Loma Prieta 6.9m Earthquake ...

It's interesting that the "positioning, navigation, timing" (PNT) community is pushing for the revival of LORAN as a resiliency measure against GPS/GNSS failure - I hope that Congress funds that to full deployment of eLORAN.

MacOS 7 got a third-party NTP client in a "control panel" from Pete Resnick in 1990, distributed for free - though I caught up with Pete at an IETF meeting and paid him a $5 shareware fee in cash in the terminal room. When they saw me do that, a number of other engineers with Macs in the terminal room ponied up at the same time.

https://www.macintoshrepository.org/2248-network-time

After I was laid off from Apple in 1997, Apple deployed a much more professional NTP server constellation: time.apple.com (and time.{euro,asia}.apple.com), and subsequently added an SNTP client in MacOS 8.5 (1998) with those DNS hostnames encoded as defaults (configured grossly by continental sales region - likely part of Apple's "localization" efforts in software) so that all Macs would keep good time.

	Erik Fair