Cable Splitter for IceCube

To achieve a better balance of cost and complexity, the collaboration chose to install two digital optical modules on each twisted pair in the deployment cable. In order to deal with this constraint, the DOM firmware becomes more complex.

One DOM per twisted pair would result in a cable too big, and too heavy to transport to the pole in an LC-130, and too large for easy handling. The alternative would be to use two cables per string, since deploying with multiple cables is manageable and has been done in the past. So, flight availability and cost comprise the major limitations.

Two DOMs per pair carries several consequences:

the communications scheme changes from simple half-duplex to multi-drop half-duplex (and the possibility of full duplex is excluded)
the cable carries double the current compared to one.
the current must be split to the two DOMs within a few meters of the DOMs.
the resistive losses in the splitter must be minimized
the impedances for communications signals must be matched at all ports
the splitter must minimize signal loss
the splitter must sustain the operating voltage across the pair.

The firmware and software on both ends of the communications link will be changed to meet requirements. Bandwidth is adequate.
As a compromise, the wire diameter in the deployment cable was increased somewhat compared to previous years cables.

The solution described here addresses the remainder of the list items.

The PSC-2-2 parts pose a problem since they must either be sealed into a pressure resisting housing, or must be made to tolerate the high ambient pressure of deployment and over pressure during the freeze-in.

The manufacturer provided partly unpackaged power splitters, above figure, for evaluation.

The mechanical example of a possible implementation might be this:

Imagine a proper printed circuit board with the right interconnections, instead of prototyping board.

In order to make the above assembly more pressure resistant, imagine that the power splitter components are, perhaps, each cast into epoxy.

Alternately, imagine epoxy extending to the edge of the fiberglass-epoxy PCB. (requires soldering before potting)

The epoxy must fill the gap between the metal base and the white ceramic PCB to equalize stress under pressure and prevent cracking.

Several solutions exist for how to connect wires to the board. Solder, for instance. It's simple but could be problem prone.
For instance, cold solder joints coming are mechanically weak, and can break when flexed. If the break occurs during freeze-in, the optical modules are probably lost.

The green terminal blocks depicted above would be soldered to PC board before potting.
The pair of wires to the surface connect on the right side, and the pairs of wires to each DOM connect on the left.
The terminal blocks are screw type for easy, reliable installation with known, repeatable, mechanical properties

Two of these assemblies, or one double assembly, would be molded into 'octopus cable assemblies' by the contractor for IceCube.

Some preliminary measurements appear here.

G. Przybylski February 15, 2001

IceCube Implementation: Update 9/25/04

The proof of the pudding is the eating... The power splitter band width didn't appear to be adequate, causing sag in communications waveforms. Furthermore, the communications waveform risetimes were rolled off sufficiently, to control EMI/RFI pick-up in the ATWD front end, that a 'bridged-T' configuration produced adequate performance, as long as the stub (located 17 meters from the terminated DOM) is kept in the range of 1 to 2 meters.