[ntp:questions] How bad is USB? (was: Re: Can the line audio out of HF radio be used to sync ntp. Trying to get a cheap ($) radio method.)

[David J Taylor]

Hal Murray wrote:
[]
> USB is polled by the host.  That gives it a bad reputation.
>
> But that polling is done by hardware.  The basic time scale is 1 ms.
> If you don't get your low-latency stuff sorted out on a serial port
> you can easily introduce delays longer than 1 ms.
>
> If you want microseconds, USB won't work.
>
> I've been experimenting with low cost GPS gadgets.  I haven't found
> anything great.  A main problem is that the software on most of
> them is setup for navigation rather than timekeeping.  Some/many
> of them send the serial text from a 100 ms timer so the USB
> delays are lost in the noise.
>
> Note that the Garmin 18x has this problem where the 18 didn't.
>
> I haven't tried grabbing PPS over USB.

Thanks, Hal.  I now a test setup with a GPS 18 LVC (not the "x" version) 
connected as a ref-clock via a Sitecom USB-to-serial converter.  The 
results are here, being updated every 30 minutes:

  http://www.satsignal.eu/ntp/NTP-on-Windows-serial-port.html#usb

I hope it's using the PPS, but I don't know the status bits well enough. 
Here's what I get:

__________________________________________
C:\>ntpq -c rv
associd=0 status=0415 leap_none, sync_uhf_radio, 1 event, clock_sync,
version="ntpd 4.2.5p161 at 1.1825-o Apr 02 11:57:05.06 (UTC-00:00) 2009 
(9)",
processor="x86", system="Windows", leap=00, stratum=1, precision=-22,
rootdelay=0.000, rootdisp=0.992, refid=GPS,
reftime=cdb02b98.382eb6a0  Sat, May  9 2009 17:19:36.219,
clock=cdb02ba5.6d93cbf6  Sat, May  9 2009 17:19:49.428, peer=25071,
tc=4, mintc=3, offset=0.514, frequency=11.026, sys_jitter=0.039,
clk_jitter=0.038, clk_wander=0.008
__________________________________________


I'm suspcious about "sync_uhf_radio", but of my two other ref-clocks with 
GPS, the one with the atom driver says: "status=2144 leap_none, sync_33", 
and the one without the ATOM driver says: "status=0415 leap_none, 
sync_uhf_radio".  Maybe they are both wrong?

Cheers,
David