Rasp Pi at +/- 1 us from GPS

MLewis mlewis000 at rogers.com
Fri Dec 1 04:39:30 UTC 2017


On 30/11/2017 9:53 PM, Gary E. Miller via devel wrote:
> Yo MLewis!
> On Sun, 26 Nov 2017 14:33:26 -0500
> MLewis via devel <devel at ntpsec.org> wrote:
>> I thought people working on having a Rasp Pi as a Time Server
>> would/should be aware of pps-client.
> Sorry for a late reply, somehow I got unsubbed from devel@
> What is the advantage of pps-client?
One advantage is that I can feed the GPS messages to Lady Heather and 
get that sw's displays (I dedicate a monitor to the Pi for this). I also 
have LH setup to discipline the system clock to provide the whole second 
if it's different from system time.
Then the GPS pps goes to PPS-Client to discipline the system clock to TOS.
I'm setting up NTPsec to pick that up with local clock.

No pgsd installed. No need.
>> My Rasp Pi 3 is running software (pps-client) that uses GPS PPS and
>> seems to be meeting the goal of maintaining system time to +/- 1 us.
> Easy. gpsd and NTPsec can get you close with no special effort.  Tobreak
> 1 uS I have to temeprature stabilize the RasPi and GPS.
The graphs I've seen published typically show significantly more than 1 
micro-second jitter for the various implementations of Pi machines as 
time servers. The regulated thermal environment shown at 
https://blog.ntpsec.org/2017/03/28/Finding_zero_TC.html shows jitter of  
+ 5 us and - 20 us, which is way better than most, and 
https://blog.ntpsec.org/2017/03/21/More_Heat.html at + 4 us and (?) -4 
us is the best of the best I've seen - except for PPS-Client.

PPC-Client adjusts for some calibrated latency, getting a jitter of +/- 
1 us, but without any of the thermal stabilizing efforts. I don't know 
how to measure if the claims are true, but the messages it's producing 
supports that.

I was heading down a similar path to the "more-heat" and such, so I do 
have my GPS module in one thermal container and the Pi 3 in another (but 
not sealed yet - a moderate blast of canned air on the Pi's heatsinks 
amusingly sends the jitter spiking).

So PPS-Client seems to have the precision down, but perhaps not the 
accuracy. (One enhancement I'm think of coding and proposing to the 
author of PPS-Client, is antenna length.)

>> - pps-client -v displays the "status printout", showing that in
>> around ten minutes it was reporting jitter as single digit us.
>> avgCorrection is typically 0.00000, -0.016667 or -0.033333,
>> occasionally 0.016667 or 0.033333 us.
> The granularity of the RasPi kernel clock is just under 0.200 uS.  So I
> assume that is off the isntalled RTC?
That's from the messages PPS-Client produces. Same as the second example 
of message output I see at https://github.com/rascol/Raspberry-Pi-PPS-Client

Is there any way to access the Pi 3 1.2G/1.1G for 883 ps or 909 ps?


More information about the devel mailing list