[time-nuts] Compensating phase differnces in dual frequency GPS receviers?

[Jim Lux]

On 12/2/11 6:28 AM, Attila Kinali wrote:
> On Fri, 02 Dec 2011 05:57:35 -0800
> Jim Lux<jimlux at earthlink.net>  wrote:
>
>> Combs are used all the time for this kind of thing (e.g. calibrating
>> Deep Space Network).  There's an old paper about calibrating a
>> interferometer radio telescope at Stanford using this kind of thing (by
>> Bracewell, as I recall)
>
> My google skills fail me on locating this paper. Could you give
> me a bit more information? Like the title or the journal it was published in?
>
>
I'll see if I can find it. It's in my files at work

there's a Bracewell paper out there called "the Stanford Five Element 
Radio Telescope" from  Sep 1973 IEEE Proceedings  but that doesn't have 
the description I'm thinking of (it was an earlier paper.. I'm thinking 
late 50s or early 60s)

"The local oscillators at the antennas are phase locked to a reference 
signal which is distributed from a centrally located oscillator. A 
system for monitoring variations in the electrical lengths of the 
reference-signal cables is incorporated using modulated reflectors at 
the five antennas. "


There are a fair number of papers out there describing how we do 
calibrations at deep space network, and they're not behind a paywall. 
I'll look for those links too.


>> It's easy to distinguish the comb from the GPS signal... the GPS is PN
>> coded, the comb is not.  If you pick your levels right, depending on
>> your digitizing strategy, it might not even jam the GPS, so you could
>> leave it on.
>
> Hmm.. right...
>
>> However, even that's not a panacea, because generating and distributing
>> that idealized comb is non trivial without destroying the phase
>> relationship between the comb "teeth".  I guess it really depends on how
>> nutty you want to be.  1 nS is pretty easy, I would think. 1pS is a lot
>> harder. 1 fS is very hard.
>
> I don't think that sub 100ps is even necessary, as the rest of the
> system will be hardly able to get to that resolution (The available ADCs
> have<100Msps and too long integration times do not work well in hardware).
> But<1ns should be feasible and desirable.
>
> But, what's the problem in distribution of the signal? I would have
> (probably naively) feed that into the input LNA.

dispersion in the components after the comb generator.  You may generate 
them all in phase at the diode, but by the time they've propagated 
through the buffer amplifier, filter, coax, they're no longer aligned. 
Again, if you're just looking for nanoseconds, that's like whole cycles 
at 1GHz, so these factors are irrelevant.  It's when you're looking at 
fractions of a wavelength at 30 GHz it gets tough.

What people who obsess about such things do is look at the reflected 
signal coming back from where you inject it into the receiver, and 
calculate the round trip time.

There's some papers about calibrating the VLA that also describe the 
techniques.  Radio Astronomers have been doing comb calibrations for 
decades, so it's pretty well known (and all the myriad sources of error 
have been identified, as well)