[time-nuts] IEEE 1588 PTP devices

[Jeff Woolsey]

I ran across some Agilent LXI 1588 Demomstration [sic] Kits for far, far
less (beer money) than what
https://www.ebay.com/itm/Agilent-IEEE-1588-Demonstration-Kit-New-in-box/161954792190
wants. There are 4 BNCs on the back, labelled 1PPS, TT, TS, and TEST. 
With that, and Ethernet and serial and USB on the back, they looked
intersting enough to take one home and play with.  That's all I knew
about them.  I thought I might be able to inject PTP time with 1PPS from
my GPSDO into my NTP network.

I found that it had a webserver for configuring from a browser, and
that's the only way I've found (so far) to set the time.  So it's an
island of precision unto itself and any brethren it finds.  I tried to
get linuxptp to talk to it, but it could only listen (I guess the LXI
box thinks that itself is the best master clock).  I wondered how well
it would sync to another PTP box. So I went and picked up another one.

It seems they sync to each other pretty well.  I was hoping that the
1PPS BNC was an input.  Alas, it's an output.  (The GUI also told me
that the TEST input is an external 100MHz clock.) I put each one's PPS
on a scope and measure the difference as about -11ns to +40ns
(jittering). Tried the same thing on a 53131A (also Agilent).  It was so
noisy I couldn't trust/believe what it reported, though a single 22.8ns
did come up.  I couldn't do that on the 5316A, since its resolution is
100ns, so I had it report GPSDO 1PPS vs LXI GPS.

LinuxPTP can at least tell me the difference between NTP (host) time and
PTP time (but not when it is a master), and it takes a couple days to
roll over one second.  I'd been arguing with these things a while trying
to get it to timestamp an external event.  Only one seems to be able to
do that so far.  So I gave it the GPSDO 1PPS to TimeStamp, and data
started accumulating.   From the TS log, I can see that its clock is a
bit less than 3ppm slow (slower if it gets colder in the lab, er,
garage).  It reports everything to nanosecond precision.

typical timestamp: 1519980450.210347925   linuxptp reports difference
between PTP and host in nanoseconds, typical value 786979741 .  That
sums close to the next second.  The difference is positive and rising now.

==

Does anyone out there have any info on these things?  I did look through
time-nuts archives around 2012.

-- 
Jeff Woolsey {{woolsey,jlw}@jlw,first.last@{gmail,jlw}}.com
Nature abhors straight antennas, clean lenses, and empty storage.
"Delete! Delete! OK!" -Dr. Bronner on disk space management
Card-sorting, Joel.  -Crow on solitaire