There was a time in my life when I used to work on rotation. That meant that I would work overseas, away from home, sometimes for two months. I would have one month off in between assignments. Sometimes the two months would be extended to three and once, it became six. That six month rotation was followed by redundancy when another corporation bought my employer. The meant I needed to reinvent myself and I emerged months later as a self-employed geophysical consultant.
Typically, overseas work rotations consisted of 12 hour days, seven days a week. Some projects were 94 hour weeks, some weeks just 70 hours. One hideous project required almost 120 hours for five weeks and the sleep deprivation nearly broke me.
I was office based for the most part. The office might have been beside a presidential palace on a main city street or in a converted 20 foot container on wheels in the middle of nowhere. A quirk of the rotations was that if the work was on a ship, my time offshore would be more likely four or five weeks with an equal time off. When I started into the five year stretch that my consultancy lasted, I tried to ensure my rotations were never more than four weeks long.
I’m lucky to be a good sleeper and normally sleep for eight hours every day. On rotation, I would aim for an extra hour every night. The additional dividend of that extra hour over 90 nights would pay out as a five day relief. It was an excellent economy of scale. If only I could have slept two extra hours!
The opposite side of that was a three month search for just two milliseconds. But first, a short digression to try explain why two milliseconds mattered.
Time was the principle commodity throughout my professional life. The geophysics for geoscience applications is denominated in time. Projects rarely have enough time yet the data are sampled in time. You might have assumed it would be sampled in depth but the collection of seismic echoes is a time based pursuit. Did you ever count the time between lightning strike and thunder clap? Or consider the journey of an echo from a cliff face in a quarry. You’d measure the time elapsed from the clap of your hands to hearing the echo. You could estimate the distance to the lightning strike or cliff as the product of the time and the velocity of sound in air. In round numbers, a one second quarry echo would place you some 340 m from the cliff face.
My professional interest would start when sound entered rather than reflected from the cliff face. Except the cliff is vertical and my main interest is below the ground beneath our feet. So let’s rotate the analogy. I would be interested in the very weak energy reflected back from layers behind the earth’s surface. Suddenly there are many challenges. How much energy went into the earth? What is the velocity of sound in the rock? And each of the layers of rock gets denser with depth of burial. It just got too complicated for a journal like this. The point is that echoes are measured in time from which we estimate depth later.
The search for two milliseconds started in an office in London. Our pictures had random ripples in them. Like a photograph not quite in focus, the images had random blurred patches. A previous experience caused me to ask about the clocks. As the saying goes: “Someone with a watch knows what time it is. Someone with two watches is never sure.”
And there can be no computer that does not have a clock. And there are many computers and clocks and chronometers on a ship. The question becomes which clock is being used.
Three months elapsed before the eureka moment. The problem was the tides. Our data was collected from a ship. Ships bob up and down on the tides so we must correct for tidal variations that happen twice daily in projects that take months to record. This is standardly and routinely done using predicted tidal heights from Admiralty Tide Tables or their regional equivalents.
It turned out that the first versions of our data were corrected for tides in universal time (UT) which is an industry convention and equivalent to GMT. However, a procedural glitch timed the ‘infill’ data as local to the project area which was the Persian Gulf. The tidal tables used were out by four hours. The tidal variations in the Gulf are small nonetheless the introduction of a four hour tidal error manifested as a blur in some areas of our images.
We finally became sure that it two milliseconds and it was only so because there were two clocks.
Bonus: Imagine what an error of two milliseconds might have done to the Rosetta mission to the comet 67P? It took ten years of journey time after decades of preparation. If you didn’t know that we landed a probe on a comet in 2015, read here.