Each cable pairs (transmission line) exhibits a characteristic impedance, which is constant, and the same for all DOM pairs between the counting house and the place where the pair splits to serve the two digital optical modules. That splitting imposes a problem. The obvious (naive) solution of connecting all pairs in parallel at the point they come together is attractive in its simplicity, but has pitfalls. An argument can be made that the short lengths of the stubs (lines from the split to their respective DOMs) will not introduce electrical problems. Connecting cables in this way, however, produces a transmission line mismatch which produces 'impedance bumps' and signal reflections distort the signal. One can further argue that the mismatch is small, therefore one can ignore errors in the calibration of the electrical length of the cable from surface to DOM caused by this configuration. These assertions are hard to verify without considerable expenditure of effort and the assembly of a test stand exactly replicating the installed configuration. The passive, resistive 'splitter', and alternate solution, wastes DC power, and attenuates the communications signals. The attenuation of the communications signals reduces the noise margin thereby increasing the bit error rate of the system.
A power splitters/combiners based splitter circuit results in a impedances matching at all three ports, and also features DC power splitting without the IR losses being incurred in a resistive splitter. The power splitter/combiner solution also affords suppression of the signaling sent from each DOM to the other, while maximizing the coupling to the counting house. The signal attenuation is also half that of the passive resistor splitter.
The figures below are screen captures from an oscilloscope attached
to all three ports of the splitter network under various test conditions.
The objective is to demonstrate that signal output from the ports faithfully
reflects the input signal, and the anticipated isolation can be observed.
Furthermore, since the splatters are built around a rather specialized
configuration of inductors, the DC current flow must not account for significant
signal degradation.
Study of a set of pictures for a pair of 50 ohm power splitters with 0.004 to 60 MHz frequency range
Comparing the sets of results leads to the conclusion that the 50 ohm splitters will satisfy the electronics requirements for IceCube.
In IceCube the splitter components reside in the 'Octopus' wiring harness
which exists between the quad cable break-out at the main cable, and the
DOMs. The Octopus cable harness has connectors for four DOMs, including
within it the neighbor coincidence signaling cables for the four DOMs and
connectors on the ends to connect to the next neighbor wiring harness to
provide
neighbor cable continuity from subgroup to subgroup.
The Octopus wiring harnesses have rubber molded over the places where
wires are connected to each other, and to the impedance matching networks.
That means the components either have to be sealed against pressure or
have to withstand the pressure at depth.