Re: [chrony-users] GPS/PSS hardware

[ Thread Index | Date Index | More chrony.tuxfamily.org/chrony-users Archives ]


I have tested the intrrupt delays on my system (assert a pin on a parallel
port at time read from the system clock, and read the clock again when the
interrupt is triggered) and get aour 1-2 usec delay. That is far far far less
then a baud clock (which even at 9600 is 100 usec.) Now it may be that there are serial ports that act as you say, but I guess I
have not seen them.




On Tue, 8 May 2018, Stephen Satchell wrote:

On 05/08/2018 06:06 AM, Bill Unruh wrote:


On Tue, 8 May 2018, Nicolas Embriz wrote:

Hi, I would like to build/test a stratum-1 server, I have found articles about how to do that with a raspberry PI and boards like https://store.uputronics.com/index.php?route=product/product&path=60_64&product_id=81 But wondering if there is hardware/devices that could be attached via USB and connect to an existing server.

USB is possible but not ideal. The interrupts are soft interrrupts (ie done by sending signals on the usb in software rather than some hardware line.) Best is if your system has some hardware line (eg serial port or parallel port) to
use. Of course that depends on what kind of accuracy you want. If it is ms
then a serial to usb converter will certainly work, and even just the GPS
signal on the usb might be fine. If it is sub-ms you want, then a hardware
interrupt is needed.
Or your desktop server might already have a serial port, and you can use that.

Or use an NTP appliance connected on a local link via Ethernet, and let the NTP/chronyd standard filtering and latency calculations allow (potentially) sub-microsecond. A lightly-loaded LAN would also work, with a slight degradation in synchronization.

One thing to note: PPS into a serial port will be affected by the interrupt request time from the UART. Some implementations of the NS16550 in ASIC cell libraries can delay interrupts by some number of baud clock ticks. This introduces what *should* be a fixed latency. But, from a system viewpoint, it's not. Latency would be affected by other things going on in the system: queued interrupts from other sources, interrupt routines running with interrupts disabled.

So, using a serial port, you will get systemic jitter from PPS.

Parallel port? Because the printers are slow async devices, they are normally put LAST in interrupt priority. Hence, jitter.

On appliances: I've not seen in the specifications where they declare the system-induced jitter from their primary source of time. I think I need to find a WWV-based NTP server, and compare apples to oranges on my network. Another VLAN...

--
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.


Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/