Puzzling clock offset spikes - update 2

[Paul Theodoropoulos]

On 7/8/2018 13:04 PM, Gary E. Miller wrote:
> Yo Paul!
>
> On Sat, 7 Jul 2018 19:33:36 -0700
> Paul Theodoropoulos <paul at anastrophe.com> wrote:
>> et voila! repeating double spikes gone, and overall error intervals
>> across nearly all metrics much tighter. Still variable, still
>> assorted spike-like glitches from time to time, but I think I'm
>> probably at the limit of what's attainable on a $35 SBC with
>> peripherals.
> Good news, but why do I get such different resuls than you?
Well, right now it's hard to tell what's different, due to the glitched 
graphs (per below).
>> Oh - and previously on ethernet, my Raspi was burning about 2.4 to
>> 2.7 watts continuously.
> Odd.  I'll have to retest.  I thought my power usage was much closer to
> 5 watts.
I'd be surprised if it's that high.  I'd expect your device to be 
burning up at that rate - I was at 60 deg C at 2.5w.

>> Link to my ntpviz graphs -
>> https://ntpsec.anastrophe.com/1hr/index.html
>> - pay no attention to graphs other than 1hr and 3hr
> You just did something that screwed the data...
Yeah, that was not expected. Shortly after posting, I was running cgps 
continuously, to see if I could adjust the GPS antenna a bit better - 
and while that was running, the unit spontaneously rebooted. There's 
been a few glitches since then too, which I can't entirely explain. I'm 
slightly worried my MicroSD may be going, so I have a couple more on the 
way. I'm also contemplating adding an RTC to my device, simply because 
it's annoying as hell running 'last' and seeing every reboot having 
occurred Dec 31, 1969. :)

Here's a link to screenshot from right around when I posted that, to 
give a better idea in the nearterm -

https://imgur.com/haWm8sv


-- 
Paul Theodoropoulos
www.anastrophe.com