These are great points, and thank you very much for the links and part recommendations. I've already got a list of several brands of two-twisted-pair CAT-5 that I was going to try out, so I'll add the Hyperline to it as well. I may also get the Ethernet to Fiber converters to play around with, although they would have to be stripped down a bit to leave any room in the electronics tube.
The train of thought you guys have is right on. Just as the decision tree shows, there are many technical routes to go, but I think that for now, focusing on one and doing it well is the best approach (for me at least). My hope, of course, is that other people will try different branches of the tree and whichever ends up being the best negotiation of cost, capability, and practicality will prevail.
I wanted to share the philosophy I've had about tethers with you guys and hear your thoughts:
I think that the best architecture for tether is one where the ROV (or possibly also electronics on the top side) have sufficient capability to communicate through an extremely low quality/ bad characteristic wire.
There are many options for getting good performance out of a higher quality line- for instance, we could purchase custom made, neutrally buoyant ROV tether that has two twisted pairs in it, or we could use very thin coax such as Belden RG-179DT which would be more easy to get the range we want with, but both of these would cost somewhere on the order of $200 per 100m length. We already know that it is possible to communicate over bad wires- for instance, using Ethernet Over Powerline adapters, one could achieve on the order of 100mbps through thin, non-category wire, but the devises are large, expensive, and can be dangerous if used improperly.
I think that with the ingenuity of people in our community, we'll find a way of doing it with devises that are small, inexpensive, and very safe. Being able to use low quality wire will also make it easier for people to get tether materials locally, create make-shift replacements in the field, have multiple tethers for different deployment scenarios. For these reasons, I believe that leaving less to the wire is a good path for the long run.
For those of you who want to help tackle this problem, you rock. Let me arm you with some of the thing's you'll need to be aware of:
1. Impedance. Impedance defines how voltage and current impulses in a modulated line relate to each other in a given circuit based on that circuit's effective inductive, capacitive, and resistive properties. The characteristic impedance of a wire is effected by many things, including the positioning of one conductor relative to another (e.g. twist rate and coiling) and the dielectric between and surrounding those conductors (e.g. the insulation material being used, and the substance the wires and insulation are immersed in like air, water, and salt water). In our experience a 100 ohm impedance wire can drop to something like 95 ohms when placed in water. Also, it's a reason that a stripped CAT-5 wire with out the PVC jacket may not work as well. If the impedance of an electrical circuit (like our Ethernet-to-twisted-pair adapters) is not the same as the impedance of a wire, then some amount of signal will be rejected and losses will occur. Not all wires have the same characteristic impedance, so that's important to look that up before trying a wire out or you'll have bad results. If a wire is non-category wire, its impedance may very greatly even over sections of a given length.
2. Cross talk. Cross talk is a type of signal interference that happens as a result of EMF (Electro-Magnetic Frequency) coming from near-by wires that are also carrying a signal. A single tether going through water is unlikely to experience cross talk, but the portion of a tether that is coiled up on the surface is very likely to experience it. Ethernet uses twisted pairs (one twisted pair for transmit, and one twisted pair for receive) in order to reduce cross talk. TX and RX are done over two wires each so that 0's and 1's can be sent deferentially (see this link to understand differential signaling), and the pairs are twisted so that each of the two wires gets the same dose of any interference (making it possible to remove that interference deferentially). Twist rates in each of the four pairs in CAT-5 are often different to assure that no one wire in a set gets a greater dose then its counterpart by always coming close to the same wire in a different set.
3. Resistance. The thinner a wire is, the more resistance it will have. We generally want to use the thinnest wire possible to reduce weight and drag, but it should be noted that that means there will be more attenuation down the line. A 5v DC supply going through 100m of 18AWG copper wire with a 100mA load will experience a drop of around 8 or 9% (resulting in 4.5v at the other end), but the same scenario using 28AWG copper wire will experience a drop of nearly 90% (resulting in only about 500mV on the other end). Also, poor quality stranded wires often don't have continuous strands (strands may break internally) so effective resistance may be even worse then what is typical for that type of wire.
It seems to me that many of these issues can be dealt with using brawns rather then brains- that is, we can just create an amplifier that will boost the signal enough to counteract attenuation, but at some point, it might be good to optimize even further by introducing a tune-able impedance matching circuit (I picture a small set screw in the side of a little black box that can be adjusted to match whatever random tether is being used), or by moving away from the standard Ethernet signal all together and going to something like the IEEE 1901 standard that Homeplug devices use. Of course, going this rout may once again create issues of cost.
Whatever method is used, I think we'll know we've been successful when we can get a satisfying amount of bandwidth (on the order of 100mbps or more) and little lag (<10ms) on thin (around 28AWG), low quality, non-category wire (that would cost on the order of $0.25USD/m or less), with a system that is small, cheap, and safe.
There is clearly a lot of work to do here, and we're only on step one, but I think we'll be able to climb pretty high pretty fast. Let's start by getting these ROVs in the water!