Hi
> On Apr 8, 2019, at 12:54 PM, jimlux <[email protected]> wrote: > > On 4/7/19 10:37 PM, Peter Monta wrote: >>> >>> Does NIST publish the transmitter bandwidth? I've never seen it, but I >>> haven't done a serious search. >>> >>> Maybe somebody near enough to get a clean signal could measure it. What >>> does a spectrogram look like? >>> >> Some of the KiwiSDR receivers are close enough to get an excellent signal. >> Coincidentally I gave this very experiment a try some weeks back; give me a >> few days to find the recordings and analysis scripts. But the bottom line >> is that 100 microseconds seems possible. The antenna bandwidth is some >> hundreds of Hz, if I remember right, but the high resolution would come >> from tracking a specific point on the amplitude trailing edge (say, the 90% >> point). > > > The precision of measurement isn't so much related to only the bandwidth of > the signal as to the combination of bandwidth, integration time, and SNR. > > THe trick is knowing what the "decorrelation time" of the channel is, because > that sets an upper bound on integration time. And, of course, the phase > noise of the transmitter and receiver. > > In any case, extracting microsecond timing from a 1 kHz BW signal is > straightforward (assuming sufficient SNR, etc.) > > As an extreme example, we measure timing of the round trip radio signal to > Jupiter and back with an accuracy of 1E-15 (5 picoseconds). The signal is > very narrow band (<<1 Hz), but we do integrate for 1000 seconds. And we use > other means to disambiguate things like "which cycle" are we on. One would *assume* that there is a fixed relation between the code and the carrier “as transmitted”. By the time it gets to you. it may not be something nice like zero degrees, but there will be a (over some period of time) stable relation. If you can get the code demodulation to the point that you can start looking at carrier phase, you then have a *lot* more edges to integrate against. Since the vast majority of the code bits can be determined ahead of time, you are not looking for a needle in a haystack. You are looking for at least a pallet full of needles. :) Can you get to the 10 us range off of the code? It’s certainly worth trying. My guess is that indeed you can with enough samples. Once you are looking at carrier phase all is not perfect. Around sunrise and sunset, you will have a tough time with WWVB. You also have the basic issue that propagation *does* swing your local signal by more than a cycle, even at fairly short distances. Without some sort of propagation “aiding” your GPSDO or time source will be wandering a bit over each and every day. The sunrise / sunset part is ….. errr …. predictable as long as you know where you are. The first order part of the propagation (day vs night) can be roughly estimated. Do you get two eight hour long “windows” of data each day or is it more? If your local standard goes into holdover for 8 hours, is that better or worse than tracking a couple cycles (at 60 KHz) of swing? If you have the sort of OCXO that is commonly used in a CDMA GPSDO, it should be capable of 10 us for 24 hours. That should put it in the “couple of us for 8 hours” range. Compared to (many) 10’s of microseconds due to propagation, that sounds like a good choice. Bob > > _______________________________________________ > time-nuts mailing list -- [email protected] > To unsubscribe, go to > http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com > and follow the instructions there. _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com and follow the instructions there.
