time-nuts@lists.febo.com

Discussion of precise time and frequency measurement

View all threads

Re: [time-nuts] NTP server using an OCXO, GPS chip and Raspberry Pi

HM
Hal Murray
Sat, Apr 25, 2020 10:19 AM

David Taylor said:

With those recommendations, and noting the delays in RPi I/O (at least in
the earlier RPi cards where Ethernet is over USB), I would suggest that an
OCXO is overkill.  No reason to deviate from the Raspbian OS.

That's an important point.  Thanks David.  I'm embarrassed that I didn't
mention it.

On Pi 2 and Pi 3, the Ethernet is on USB which adds a layer of jitter to the
timing you can get using NTP.

--
These are my opinions.  I hate spam.

David Taylor said: > With those recommendations, and noting the delays in RPi I/O (at least in > the earlier RPi cards where Ethernet is over USB), I would suggest that an > OCXO is overkill. No reason to deviate from the Raspbian OS. That's an important point. Thanks David. I'm embarrassed that I didn't mention it. On Pi 2 and Pi 3, the Ethernet is on USB which adds a layer of jitter to the timing you can get using NTP. -- These are my opinions. I hate spam.
AK
Andreas Kempe
Sun, Apr 26, 2020 7:25 PM

On Sat, Apr 25, 2020 at 03:19:54AM -0700, Hal Murray wrote:

David Taylor said:

With those recommendations, and noting the delays in RPi I/O (at least in
the earlier RPi cards where Ethernet is over USB), I would suggest that an
OCXO is overkill.  No reason to deviate from the Raspbian OS.

Deviating from Raspbian isn't really anything other than personal
taste. I happen to like the BSDs. I'm not opposed to Raspbian as such
if it were to simplify things.

That's an important point.  Thanks David.  I'm embarrassed that I didn't
mention it.

On Pi 2 and Pi 3, the Ethernet is on USB which adds a layer of jitter to the
timing you can get using NTP.

Thank you both for this important insight! The only reason I really
mention a Raspberry Pi is because my originial idea was to feed the
PPS signal into a GPIO pin and the Raspberry is very easy to get a
hold off and has GPIO pins.

Cordially,
Andreas Kempe

On Sat, Apr 25, 2020 at 03:19:54AM -0700, Hal Murray wrote: > > David Taylor said: > > With those recommendations, and noting the delays in RPi I/O (at least in > > the earlier RPi cards where Ethernet is over USB), I would suggest that an > > OCXO is overkill. No reason to deviate from the Raspbian OS. Deviating from Raspbian isn't really anything other than personal taste. I happen to like the BSDs. I'm not opposed to Raspbian as such if it were to simplify things. > > That's an important point. Thanks David. I'm embarrassed that I didn't > mention it. > > On Pi 2 and Pi 3, the Ethernet is on USB which adds a layer of jitter to the > timing you can get using NTP. > Thank you both for this important insight! The only reason I really mention a Raspberry Pi is because my originial idea was to feed the PPS signal into a GPIO pin and the Raspberry is very easy to get a hold off and has GPIO pins. Cordially, Andreas Kempe
PT
Petr Titěra
Mon, Apr 27, 2020 1:21 PM

On 25.04.2020 12:19, Hal Murray wrote:

David Taylor said:

With those recommendations, and noting the delays in RPi I/O (at least in
the earlier RPi cards where Ethernet is over USB), I would suggest that an
OCXO is overkill.  No reason to deviate from the Raspbian OS.

That's an important point.  Thanks David.  I'm embarrassed that I didn't
mention it.

On Pi 2 and Pi 3, the Ethernet is on USB which adds a layer of jitter to the
timing you can get using NTP.

What is considered as too much jitter in this case? I see this in my
home setup:

  remote           refid      st t when poll reach   delay   offset 

jitter


---=============
SHM(0)          .SHM.            0 l  34  64  377  0.0000 -57.9552
1.6973
*SHM(1)          .GPSD.          0 l  13  16  377  0.0000  -0.0016
0.0004
+192.168.3.246  .PPS.            1 u  17  64  377  0.3834  -0.0064
0.0149
-192.168.3.247  .PPS.            1 u  18  64  377  0.3545  -0.1689
0.0069
+192.168.3.102  .PPS.            1 u  15  64  377  0.1966  -0.0729
0.0187

Server 192.168.3.246 is RPi3.

Petr Titera

On 25.04.2020 12:19, Hal Murray wrote: > > David Taylor said: >> With those recommendations, and noting the delays in RPi I/O (at least in >> the earlier RPi cards where Ethernet is over USB), I would suggest that an >> OCXO is overkill. No reason to deviate from the Raspbian OS. > > That's an important point. Thanks David. I'm embarrassed that I didn't > mention it. > > On Pi 2 and Pi 3, the Ethernet is on USB which adds a layer of jitter to the > timing you can get using NTP. > > What is considered as too much jitter in this case? I see this in my home setup: remote refid st t when poll reach delay offset jitter =============================================================================== SHM(0) .SHM. 0 l 34 64 377 0.0000 -57.9552 1.6973 *SHM(1) .GPSD. 0 l 13 16 377 0.0000 -0.0016 0.0004 +192.168.3.246 .PPS. 1 u 17 64 377 0.3834 -0.0064 0.0149 -192.168.3.247 .PPS. 1 u 18 64 377 0.3545 -0.1689 0.0069 +192.168.3.102 .PPS. 1 u 15 64 377 0.1966 -0.0729 0.0187 Server 192.168.3.246 is RPi3. Petr Titera