[time-nuts] Question about the PLL of Trimble Thunderbold

[Tom Van Baak]

Hi Ferran,

Here's another idea for your multi-OCXO synchronization project.

Normally when we think about synchronized oscillators we imagine two of them in side-by-side or perhaps separated by a few meters of cable. Through some PLL magic they remain in perfect phase (and frequency), either between themselves or also referenced to some third oscillator, like a GPSDO.

The question is what are these two OCXO being used for.

In the special case that the OCXO are being used as part of some signal measurement or data acquisition system, then consider this idea. Maybe you don't actually need the OCXO to be synchronized. Instead of requiring synchronization, why not just record what their phase relationship is at all times. You still compare the OCXO but you don't bother to steer them.

If your application is going to do signal processing based on the OCXO it seems to me that you don't need the OCXO to be synchronized. What you need to know is what their relative phase is. And that offset is just fed into any existing math that you're already doing.

This is not unlike how the clocks in GPS are "synchronized". They are not physically sync'd in phase or frequency. Instead each clock is free-running and several correction numbers are sent down as part of the data stream. The a0 and a1 numbers give the phase and frequency offset so that the user can construct virtual clocks that are synchronous.

So if your application doesn't actually need physical synchronization but instead would work with virtual synchronization then you don't need to hack a GPSDO. Instead all you need are good OCXO's and high-performance phase comparators.

The TAPR TICC is good to about 60 ps and works at 1 PPS. There is another device you may want to look at, the PicoPak [1]. It is good to 6 ps and works at 10 MHz. You could use one or more of these to monitor each OCXO in your ensemble and then use the comparator data stream to construct a virtual clock. Any measurements made from each OCXO could be adjusted in s/w relative to the virtual clock. So you could achieve ps-level virtual coherence in your data analysis without requiring ps-level physical coherence in each OCXO.

I'll give you one example. If you wanted to build a ToA (time of arrival) system based on two independent receivers some distance apart you might need them to be highly synchronized. My point is, skip the synchronization, which might difficult at the ps level, and just measure the clocks continuously and precisely at the ps level. Then when you do your ToA math you just factor in the known clock offsets. This method would also help deal with cable propagation issues (e.g., temperature, stress), which you will want to measure (two-way) and factor into your synchronization.

/tvb

[1] http://www.wriley.com/PicoPak%20App%20Notes%20Links.htm
There are 40+ papers in this folder by Bill Riley, who is one of the best T&F guys out there.
Since you're new to time-nuts, you may also enjoy reading everything else written by Mr Riley. See: http://www.wriley.com/