time-nuts@lists.febo.com

Discussion of precise time and frequency measurement

View all threads

Re: [time-nuts] PPS stats

HM
Hal Murray
Mon, Jul 6, 2020 7:26 PM

NTP is using the PPS and my stats look good, but when I run
    ntpq -c kerninfo
The pps frequency, stability, and jitter are all zero.

    dmesg | grep pps
and
     ppstest /dev/pps0
both indicate the kernel pps support is working.

Why isn't the kerninfo showing any info on the pps
frequency, stability, and jitter?

There are 2 modes of PPS.  ntpq -p doesn't show the difference.

The normal mode is that ntpd processes each PPS pulse as a data point, just
like a data point from a packet exchange with a NTP server.  (Not quite, there
is an extra level of filtering, but close enough.)

The second mode is that the kernel does everything after ntpd turns on a bit.
It's a kernel option.  Most distros don't include it - it conflicts with a
scheduler option that saves power.  (aka you have to build your own kernel)
Those slots get filled in when the kernel mode is running.

pll offset:            0
pll frequency:        -28.451
maximum error:        0.1055
estimated error:      3e-06
kernel status:        pll ppsfreq ppstime ppssignal nano
pll time constant:    6
precision:            1e-06
frequency tolerance:  500
pps frequency:        -28.451
pps stability:        0.012085
pps jitter:            0.007
calibration interval  256
calibration cycles:    1020
jitter exceeded:      13
stability exceeded:    0
calibration errors:    0

--
These are my opinions.  I hate spam.

> NTP is using the PPS and my stats look good, but when I run >     ntpq -c kerninfo > The pps frequency, stability, and jitter are all zero. >     dmesg | grep pps > and >      ppstest /dev/pps0 > both indicate the kernel pps support is working. > Why isn't the kerninfo showing any info on the pps > frequency, stability, and jitter? There are 2 modes of PPS. ntpq -p doesn't show the difference. The normal mode is that ntpd processes each PPS pulse as a data point, just like a data point from a packet exchange with a NTP server. (Not quite, there is an extra level of filtering, but close enough.) The second mode is that the kernel does everything after ntpd turns on a bit. It's a kernel option. Most distros don't include it - it conflicts with a scheduler option that saves power. (aka you have to build your own kernel) Those slots get filled in when the kernel mode is running. pll offset: 0 pll frequency: -28.451 maximum error: 0.1055 estimated error: 3e-06 kernel status: pll ppsfreq ppstime ppssignal nano pll time constant: 6 precision: 1e-06 frequency tolerance: 500 pps frequency: -28.451 pps stability: 0.012085 pps jitter: 0.007 calibration interval 256 calibration cycles: 1020 jitter exceeded: 13 stability exceeded: 0 calibration errors: 0 -- These are my opinions. I hate spam.