I was thinking about how to do the synchronization test of the
different experiments. Here's something that may or may not be
feasible. Comments, suggestions, experience, ideas are
very welcome.
The best way to test is to have actual light in the all the detectors
at the same time, with relative times known to about 10 microsecs
(although 100 microsecs may be OK).
So, what about an "LED time bomb": you build a bunch of identical
devices that have very stable free-running clocks in them. They are
synched at some time. Then they are rigged to fire a pulse which
drives an LED after a fixed number of clock cycles. So, for example,
suppose the clock is accurate enough to drift less than 10 microsecs
in a week. LVD, MACRO, SNO and Super-K people could all get together
to synchronize their devices, then take their time bombs on a plane
back to their experiments (being very careful what they say to the
gate security agents), put the devices in their detectors and have
them all go off at the same time. (Actually this would be very fun.)
Or, if 10 usec drift in a week is impossible, but 10 usec in a few hours
isn't, then we could all synch to some straightforward, believable GPS
signal outside our respective laboratories, then hurry inside and do
the test.
(Of course, you could have an LED triggered by a distributed GPS
signal inside the lab. But the distribution system for event
time-stamping is part of what we'd be trying to test. In Super-K's
case, for instance, the GPS distribution system is not particularly
simple. There's no reason to believe it doesn't work, but we want to
be *really* sure, using the simplest possible time reference.)
The problem is getting a stable enough clock. It would
have to resist possibly significant temperature changes, and
shocks equivalent to at least a bumpy Land Cruiser ride. I poked
around on the web to see what is possible, and found typical df/f
values:
Cesium atomic clock: 10^-11 to 10^-13
Rubidium atomic clock: 10^-9 to 10^-10
Double oven crystal: 10^-9
Oven crystal: 10^-8
Temperature compensated crystal: 10^-6
Crystal: 10^-5
10 microsecs in a week is 1e-12, so a cesium atomic clock is needed
for the physical meeting scenario; too bad. 10 microsecs in an hour
is 2e-9, so we need at least ppb accuracy for the GPS-synch-and-run
scheme. That looks doable in principle with "double oven crystals"
(OCXO's, Oven Controlled Crystal Oscillators). These are crystal
oscillators which are maintained at some constant temperature
exceeding maximum ambient temperature, to keep them stable.
So, say we build 3-4 time bombs. How much would it cost? The
dominant cost is probably the oscillator. Simple GPS receivers that
output pps signals (good to 50-100 ns) can be had for less than $200;
also experiments may already have GPS receivers. Some additional
components would be required-- chips, cables, case for LED, etc. --
but probably don't add up to much.
I got a quick quote for ballpark costs:
Oscilloquartz sells better than ppb OCXO's for about $1000 apiece.
That's pretty steep-- puts the cost of the whole project at
$4000. It may well be possible get some much cheaper; I've been told that
military surplus may be a good source (also amateur radio groups).
(I found on the web some "low-cost" portable rubidium atomic clocks,
e.g. Temex, but haven't received quotes for these yet. I have no
feeling at all for how much these cost, but I'm guessing more than
$1000. And of course for this case, as Alec pointed out to me, we'd
have to be *really* careful not to mention to gate security agents the
"atomic clock time bomb" in our luggage...)
Some other notes:
-- Is ppb time drift really achievable? It looks to be from my
naive look at the OCXO component specs, but I'm not sure in practice
how various tolerances add up. Anyone with experience in this realm,
please shout.
-- We'd need to make the LED light delivery system compatible with
all experiments. SK and SNO are probably straightforward (just drop
an LED in). But MACRO and LVD have separated compartments and
multi-LED systems (at least MACRO does; don't know about LVD) so it
would be trickier. And I'm not sure about AMANDA. For these cases
there may be some delays at the LED-driving end that would have to get
measured carefully (ideally, the LEDs should get fired by the
time-bomb pulse as directly as possible) but so long there's no jitter
it would be OK.
-- I don't think this will work for OMNIS, which needs neutrons. But
maybe something could be dreamed up.
-- Alternative general idea (although I can't think of a way to
implement it that would work to high accuracy): what about an
*analog* fixed time delay instead of a digital one, a "fuse" of some kind
based on a physical process of known duration (light propagation?
radioactive decay?)
Anyway, just some thoughts...please bring your ideas to the video meeting.
Kate.