RE: [chrony-users] Refclock Integration Advice

["Chang, Benjamin" <Benjamin.Chang@xxxxxxxxxxxxxxx>]

Regarding the interrupts, perhaps? I am using a RHEL linux system, so I guess I'll look into that. I guess I don't know typical PPS lengths. Out of curiosity, what are typical pulse lengths for a PPS?

Regarding the minpoll and maxpolls of my ntp server, I have it set at -6 because  I want my system to be synced as close as possible to the ntp server.. I noticed that the measured offset decrease with higher polling frequency.. I am be misunderstanding how polling works, but doesn't an ntp just respond to whenever its polled? So if my system polls my ntp server every 1/64th a second, itll respond just as fast? I am aiming for ns resolution. I don't believe I am using the gps unit with my ntp server. I am just using its freestanding clock (with a rubidium oscillator).

Thanks,
Ben


-----Original Message-----
From: Bill Unruh <unruh@xxxxxxxxxxxxxx> 
Sent: Monday, November 30, 2020 12:37 PM
To: chrony-users@xxxxxxxxxxxxxxxxxxxx
Subject: Re: [chrony-users] Refclock Integration Advice


On Mon, 30 Nov 2020, Miroslav Lichvar wrote:

> On Mon, Nov 30, 2020 at 03:03:10PM +0000, Chang, Benjamin wrote:
>> So I managed to switch it to a 50% duty cycle PPS, and it appears to take every pulse now. The LastRX floats between 0 and 2 (seemed like 20 us is too short a pulse).
>>
>> However, now when I look at the measured offsets, it seems to greatly vary from 100s of ns to 10s of us. Shouldn't it be consistent? According to the datasheet of my PPS (the syncserver S650), it should have ~15 ns accuracy and resolution. When I scoped it on an oscilloscope, it was consistent and not varying.

Almost certainly problems with the interrupt handling by the computer. It takes a while for the operating system to get to the interrupt, if other things (disk reads, etc) are hogging the interrupt system. 20us seems a bit long, but it will depend on your system. 
>>
>> Here is my conf file setup
>> My ntp server setting: minpoll -6 maxpoll -6

That still makes absolutely no sense. What in the world do you expect to change in 1/64 of a second. Certainly there is no new time signals coming from the gps unit.

>> My PPS: poll 0 prefer trust pps width 0.5
>
> If you use a serial port for the PPS, the stability will depend on how 
> stable is the interrupt delay. A stability in tens of microseconds is 
> not unusual. You may get better results if you disable all 
> power-saving states and set a fixed frequency for the CPU.
>
> If you increase the chrony poll to 3 or 4, the PPS samples will be 
> filtered using a median filter and that should help with the outliers.

--
To unsubscribe email chrony-users-request@xxxxxxxxxxxxxxxxxxxx
with "unsubscribe" in the subject.
For help email chrony-users-request@xxxxxxxxxxxxxxxxxxxx
with "help" in the subject.
Trouble?  Email listmaster@xxxxxxxxxxxxxxxxxxxx.


--
To unsubscribe email chrony-users-request@xxxxxxxxxxxxxxxxxxxx
with "unsubscribe" in the subject.
For help email chrony-users-request@xxxxxxxxxxxxxxxxxxxx
with "help" in the subject.
Trouble?  Email listmaster@xxxxxxxxxxxxxxxxxxxx.