Matt Matt Ettus wrote: > On Mon, Jan 5, 2009 at 3:02 PM, Bruce Griffiths > <bruce.griffiths@xtra.co.nz> wrote: > >> Matt >> >> Matt Ettus wrote: >> >>> On Mon, Jan 5, 2009 at 1:54 PM, Bruce Griffiths >>> <bruce.griffiths@xtra.co.nz> wrote: >>> >>> >>> >>>> If the application is somewhat analogous to VLBI, then the maximum >>>> (uncorrectable ie random) allowable carrier frequency phase errors >>>> between receivers depends on the integration time. >>>> Maximum integration times for VLBI are typically 10,000 sec or less, >>>> with corresponding relative ADEV (after correction for drift etc) on the >>>> order of 1E-14 (or better) @ 10,000 sec for a carrier frequency of a few >>>> GHz. >>>> >>>> What is the integration time in your application? >>>> >>>> >>> We are trying to receive data which is modulated at rates which vary, >>> but are in the 100 Hz to 100 kHz range, so the integration times are >>> very short, all below 10ms. >>> >>> >>> >>>> What is the carrier frequency? >>>> >>>> >>> 30 MHz and below for now, but we'd like to be able to go as high as >>> 500 MHz or so. >>> >>> >>> >>> >> What's your tolerance for phase variations between a pair of LOs? >> Its on the order of 50 degrees or so for VLBI. >> At 500MHz this corresponds to about 140ps. >> The corresponding relative ADEV is around 1E-8 @ tau = 10ms. >> Almost any good crystal oscillator can achieve this. >> > > As you said, our ADEV requirements are pretty loose and could be met > by any decent crystal. > > But Said just told us that he sees about 25ns difference between > units, which is about 180 times worse than 140 ps. > > That estimate should have been 280ps with a 500MHz LO, however that still doesnt help very much. This is more a function of the timing receiver noise than the characteristics of the OCXOs themselves. You cant expect too much better than that using code phase timing receivers. However at 30MHz a sawtooth corrected M12+T or M12M will get close to the 50 degrees of phase error ( 4.76ns). Using carrier phase techniques you can easily achieve sub 100ps resolution at least for integration times of up to a few minutes. Longer term cycle slips can cause problems. > In VLBI I think that they can deal with an unknown phase error between > the two LOs as long as that phase error remains relatively constant > over the period of integration. That is why they care about ADEV. > They just correlate the two signals, and subtract out that constant > phase error. My problem is that we can't just do that in this > application. We need to know that phase error a priori. > > Matt > > > For the carrier itself absolute phase measurements cant be done. Such phase measurements only have meaning when modulation is present. Bruce

> For the carrier itself absolute phase measurements cant be done. > Such phase measurements only have meaning when modulation is present. Alternating modulation frequency may do it for instance. Choosing a set of modulation frequencies being relative prime to each other alows for several readings of alternating cycle ambiguity to be resolved using the Chinese Reminder theorem. Similar set of frequencies can be used, for instance multiple of 10 alows the phase of the lower frequency to determine which of the 10 cycles the next frequency is. However, this is not easilly extendable to carrier frequency cycles, since it would imply a very wide modulation, so for that a chinese reminder approach is of interest. A PRN sequence is just a multiple of these frequencies at the same time. Alternating PRN length could be used for similar purposes of course. Multiple simultaneous PRN sequences can be added upon each other. Using multiple carrier frequencies may also be an interesting option. Consider for instance COFDM which creates a dense set of carriers using IFFT modulation and FFT demodulation. Cheers, Magnus