[time-nuts] how good an oscillator do you need for a GPS simulator
hmurray at megapathdsl.net
Sat Dec 17 23:20:51 EST 2011
jimlux at earthlink.net said:
> But the real question isn't how to generate the signals (that's straight
> forward).. it's "how good does the oscillator have to be" to effectively
> test the receiver, in the sense of measuring it's timing performance.
My 2 cents, which could be way off...
One of the things the receiver has to solve for is the actual frequency of
the local oscillator.
In the process of doing that, the receiver can't tell the difference between
an offset in the local crystal and a coordinated offset that's the same in
all the transmitter frequencies.
So the actual accuracy on a fake transmitter only has to be good enough to
fall within the band the receiver thinks is OK. That's a software vs
manufacturing/testing issue. If the software will take X ppm but the
manufacturing guys are buying crystals good for X/2, the transmitter can be
off by X/2. If the software will take X but the manufacturing guys are
buying junk crystals, it may not even work with a transmitter that is right
One thing I don't understand about this area. Is the receiver clock offset
an independent unknown?
The usual reasoning goes that a receiver needs 4 satellites to solve for 4
unknowns: X, Y, Z, and T. If you are willing to assume you are on the
surface of the earth, you can get away with only 3 satellites. That's a
simple application of N independent unknowns needs N equations. (That's
assuming the receiver isn't moving.)
Is the receiver frequency another unknown? Do I need another satellite? Or
does it drop out somehow?
It might drop out. For example, take the 1D case with two satellites. (I
know where they are and where they are going.) After I correct for the
satellite Doppler, the velocity of the satellites should be zero. I guess
that's another equation, so it drops out.
These are my opinions, not necessarily my employer's. I hate spam.
More information about the time-nuts