[time-nuts] Soekris Net4801 vs NET4501 as Statum 1 timeserver

[Poul-Henning Kamp]

In message <002b01c5b5bf$8eda7670$0201010a at barney>, "Geoff Horner" writes:
>Hi all,
>
>Does anyone know which of the two Soekris boards makes a better statum 1 
>time server.

The net45xx series does because you can make hardware timestamps
with it:
	http://phk.freebsd.dk/soekris/pps

>Anyone used a OCXO in either?

Yes, both.  Finding OCXO's with the oddball frequencies are the
hardest bits.

>I have seen phk's results of the 4501, although 
>that was tweaked at fair bit using an Rb oscillator.

The timestamping performance would be the same without an Rb,
even without using an OCXO you will get much better performance
than with anything that needs interrupts to timestamp.

Be aware that the normal NTPD can't really use this for anything,
the PLL is not tuned for such precision work.

I have writte from scratch NTP server called NTPns which is geared
towards stratum-1 work with atomics etc, and you can have a copy
if you want.

This is the stability display from a net4501 with an ISOtemp OCXO131
OCXO and a Oncore m12+T GPS:

Source m12_0: votes 6.000000 flags <UTC> los 0/15 update 1 SELECTED
        stratum 0 delay 0.000000000 dispersion 0.000000088 refid [47 50 53 00]
        last_ts 1126340906.000000006 last_delta 0.000000007
     1: 11111111 -0.000000046 1.391e-07
     2: 11111111 -0.000000059 1.456e-07
     4: 11111111 -0.000000052 7.344e-08
     8: 11111111 -0.000000031 1.525e-08
    16: 11111111 -0.000000036 4.834e-09
    32: 11111111 -0.000000030 8.111e-09
    64: 11111111  0.000000044 6.030e-10
   128: 11111111  0.000000007 4.264e-10
   256: 11111111  0.000000011 3.015e-10
   512: 11111111  0.000000030 1.317e-10
  1024: 11111111  0.000000028 8.538e-11
  2048: 11111111 -0.000000013 1.013e-10
  4096: 11111111 -0.000000012 4.363e-11
  8192: 11111111 -0.000000024 3.043e-11
 16384: 11111111  0.000000006 6.043e-12
 32768: 11111111 -0.000000035 4.342e-12
 65536: 11111111 -0.000000033 2.434e-12
131072: 11111111  0.000000002 6.655e-13
262144: 11111111  0.000000009 4.418e-13
524288: 11111111 -0.000000002 2.961e-13

First column: integration time
Second column: precense bitmap for 8 samples
Third column: average offset [sec]
Fourth column: Allan variance after discipline.

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.