[time-nuts] Improving on basic L1 timing

[Angus]

>Make sure you have good skyview (aka good geometry) and very little
>multipath. Both effects can affect the precision of your solution
>much more than the ionospheric delay variation.

Main one is on the peak of a slate roof. Nothing else much higher
around except a large sycamore tree. There's another antenna at the
edge of the roof about 1.5m below the peak, close to one end of the
house, and although it does give slightly poorer stability, it did not
look to be by a huge amount when I checked it.

The reason that I was particularly looking at the ionospheric delay
was that the diurnal variation was quite pronounced. The rubidium and
gps were both in temperature controlled enclosures. That does still
leave the antenna, cable and splitter, along with the counter and PPS
buffering.

>If you live in continental Europe, then you will inevitabely have an
>IGS station nearby. They are everywhere! :-)

I think it's a few hundred km for me!

>The only other thing I can think of is using multiple Rb's as reference,
>measure their temperature, air pressure and drift against GPS. 

That actually is the plan. Most would also be temperature controlled.

>Use all
>this data to build a clock model (aka Kalman filter) that compensates
>for temp/pressure/aging and measure the Rb under test against this ensemble.
>But that's not something that already exists and is ready to use.

That's just the sort of thing that was always the end goal for this
project - control what can be, understand and compensate for what
can't. 

The rubidiums are what I'm primarily interested in - the gps was
originally there for a basic reference, as well as for long term drift
correction. 
Thing is, since I don't have access to a H-maser or good cesium, I'm
also trying to see how well I can get gps to do as a medium to long
term reference. 

It may well be best just to run some ordinary timing receivers for a
while, and use the rubidiums for the rest. By then I'd have a better
idea what performance would be needed to be useful, and if anything
available to me would do that.

Another complication is that a lot of the old data I have was done
with an M12+T without sawtooth correction. With filtering or averaging
it's irrelevant medium to long term which is why I didn't bother, but
it messes up Adev plots making comparisons with a lot of the data out
there more difficult. Some tests with other receivers and corrected or
sawtooth-free data would be a good start.

Angus.