[time-nuts] UBLOX GPS board testing

[Magnus Danielson]

On 12/21/2010 10:26 PM, John Green wrote:
> I tried substituting the 1 PPS output from the Z3801 and comparing it
> to its own 10 MHz output and find the same jumpy behavior as I get
> with the UBLOX boards. Well, not exactly the same but pretty much. Now
> I am confused. I expected the 1 PPS to be in lock step with the 10
> MHz.

Sometimes counters has a problem when the start and stop signals occur 
too tight in time after each other. Running the Stop signal through some 
extra cable avoid the trigger issue and out pops stable readings. Since 
each meter of coax adds 5 ns of delay, it is a pretty handy method of 
getting out of trouble.

Do measure the PPS to Clock property of your GPS. If your counter needs 
it, use the cable trick above to get stable measurements.

A month ago Björn and I did some experiments together, found in my 
posting of 20 Nov titled "PPS and 5/10 MHz GPSDO time relationship":

8<---
Fellow time-nuts,

Björn and I have been having some fun during our get-together in his 
basement time-lab. I pulled with me some gear (CNT-90 and SR620) for him 
to play with, so a good warmup exercise was to measure the offset 
between the PPS and the clock output (5 MHz or 10 MHz).

The results was uhm... spread out... so we felt like sending you guys a 
report.

First out was a RAPCO 1804M which has a HCD 66 SC 5 MHz oven diciplined 
by an old Trimble SV6+ (?) GPS receiver. We popped the lid for other 
purposes... :) It had the 5 MHz rising edge 32,17 ns behind the PPS 
rising edge, with 100 ps RMS jitter. Quite noticeable offset but fair 
jitter.

The good old RAPCO was jumperable to be on "OSC" or "GPS" on the 
mysterious jumper LK9 and it was stuck hard on the "OSC" setting, but 
some physical exercise later we got it into GPS setting and it had a 
about 200 ns peak to peak sawtooth... nice and pedagogical exercise.

We then had a look at the Brandywine GPS-4 (mine on loan to Björn) and 
found it had fairly nice numbers... until we discovered it has a 
periodically reoccuring glitch of unknown system-source.

Natually we hooked in to Björns Thunderbolt and found the offset so 
tight that we ran into trigger-problems, but offsetting the clock by 
about 8 ns of coax cable we had a about 4 ns in average and 6 ns 
peak-to-peak. The PPS thus jumped between two distinct offsets with 
their respective gaussian distribution around them. Not all that neat, 
and the RAPCO was the quietest in this shoot-out.

Over and out,
Magnus and Björn
--->8

Once you know how your favorite GPSDO behaves, you should know if you 
can use the PPS directly as a quiet source or if you only should use it 
as an external ARM to trigger measurements and then use the clock for 
Start (Channel A) and DUT for Stop signal (Channel B).

Going for the PPS-arming directly avoids the issue somewhat.
When comparing a different frequency than 10 MHz I insert a generator 
which locks to the 10 MHz and produces the nominal frequency onto which 
I will do my measurements.

Don't forget to optimize trigger jitter and verify it separately for the 
Start and Stop signals.

Don't forget to let your DUT warmup.

Cheers,
Magnus