[time-nuts] Minor NTP Issue?

Hal Murray hmurray at megapathdsl.net
Mon Nov 1 20:00:12 UTC 2010


>      remote           refid      st t when poll reach   delay   offset  jitter
> ==============================================================================
> *NET4501-1       .GPS.            1 u   36   64  377    0.724    0.022   0.100
> +NET4501-2       .GPS.            1 u   22   64  377    0.801    0.038   0.132
> +PRAECIS         .GPS.            1 u   14   64  377    0.740    0.097   0.034
> -TS2100-1        .GPS.            1 u   50   64  377    4.995   -0.362   0.030
> -TS2100-2        .GPS.            1 u   17   64  377    4.966   -0.331   0.035

I can think of two reasons for something like that.

1) Asymmetric network (or processing) delays:
  NTP measures the round trip time, assumes it is symmetric, then calculates 
the clock offset from there.

What is your setup like?  Are all the boxes plugged into the same switch?  
What are ping times to the TS2100 from a PC right next to it?

If the TS2100s are located "far away", can you swap a TS2100 with a NET4501? 

2) The clock really is off:
  This seems a bit unlikely, but not too crazy.  Maybe Datum didn't have a 
time-nut on staff and nobody ever looked closely to see how well it worked.


  
> I seem to recall that the standard package came with 50' or
> 150' of cable, but even that pre-factored delay wouldn't equate to
> ~300-400us would it? 

The speed of light in air/vacuum is 1 ft per ns.

On PC boards, it's half that.  Outer layers are slightly faster than inner 
layers.

Good coax is rough 5/8 to 2/3 the speed in air.

(The speed of light is determined by the dielectric constant of the medium.  
Low loss coax is mostly air.)

Fiber is close to coax.  For back of the envelope, I use the mile to km 
conversion factor: 5000 ns per km.



-- 
These are my opinions, not necessarily my employer's.  I hate spam.







More information about the Time-nuts_lists.febo.com mailing list