Project Sunshine: Part 1

Interlude: Ethernet Primer

Before I go through what I actually did here (which isn’t a lot so don’t get too excited), I’d like to run through the highlights of what Max-from-a-week-ago would have considered more than anyone outside of professional networking needed to know about ethernet cables.

To start with the very basics, ethernet cable is cable over which data is transferred. Typically this is for networking, but ethernet cable is used for all sorts of data transfer, including access control, telephone cables, video, or any other signal. Ethernet cable always seems to be a set of twisted pairs, and comes in shielded or unshielded varieties.

Shielding? Twisted pairs? Transmitting anything over a physical medium involves introducing some amount of interference, or noise. This is caused by resistance in the medium, as well as external electromagnetic waves that hit the cable. Shielded cables include a foil around them that attenuates incoming electromagnetic waves to increase signal fidelity. This protects from interference from things like electrical current, fluorescent lights, air conditioners, and almost anything else that does stuff with wires. Shielding has its downsides: it’s stiffer and thicker, making it more difficult to work with, and it costs more. The cable I have in my walls, and in my patch cables, is unshielded, which is apparently the standard. I have seen shielded cabling like this on site at work, though.

Twisted pairs are another technique to reduce interference along a cable run. This is an Alexander Graham Bell invention, and the Wikipedia page goes into a lot more detail than I can here and is genuinely fascinating reading (It also discusses standards for shielding in some detail, which I also recommend reading. Then you too can know the difference between F/UTP and S/FTP.) The gist of it is that twisting a pair of wires equalises the average distance of each component from a source of interference, and sending the same signal down each component allows the error to be isolated at the end. It’s something like this:

\begin{aligned} W_{1} &= s + e_{1} \\ W_{2} &= -s + e_{2} \\ W_{1} - W_{2} &= (s + e) - (-s + e) \\ &= 2s + e_{1} - e_{2} \\ &\approx 2s &\text{ if } e_{1} \approx e_{2} \end{aligned}

We aren’t even close to done with the complexity inherent in ethernet cabling. Next up is the actual materials used in the wires. Ethernet can be either solid core or stranded core. That is to say that each component wire can be made of a single solid piece of material, or a lot of strands of a material. In ethernet cables, the material in question is always copper. Solid core cables experience less signal loss, but are also harder to bend, harder to stuff into RJ45 plugs, and easier to break. Stranded cables are the opposite side of those compromises. Generally speaking, solid core cables are the ones that are in wires, and get punched (technical term) into keystone jacks, which is the thing you plug an ethernet cable into. And the ethernet cables that you plug in (I think they are called patch cables) are usually stranded.

Ethernet cables are categorised into numbers, each representing different standards. These standards specify things like the materials, the number of pairs, the frequency of the twists, and then the throughput capacities that are supported, and the distances over which those capacities are supported. For example, Category (hereafter Cat) 5e (e for enhanced) was standardised in 2001, and guarantees at least gigabit speeds, and over short enough runs, is good enough for 2.5Gbps. Cat 6 is backwards compatible with Cat 5e, but it has stricter standards, and cannot run for as long. In return, you get support for 10Gbps. Again, each category supports both solid and stranded core, and offers different options for shielding.

Interlude: Network Switches

Oh boy. I thought cables were complex. So you can’t just splice ethernet cable. The transmission requirements and the razor thin tolerances in the signal strength prohibit that. A switch is a powered piece of hardware that receives something on one of its ports, and retransmits it to one or more of its other ports.

Switches come in two variants: managed, and unmanaged.

A managed switch is the easier one to understand. A managed switch is a computer that has rules about what is allowed to be retransmitted, and the patterns in which that retransmission can happen. They can prioritise traffic, or indeed do pretty much anything else you can imagine doing with signal processing and a computer.

Unmanaged switches do something simpler, yet also less obvious, at least to me. My understanding is that it maintains an index of MAC addresses mapped to the port on which they reside, or at least the port that points towards them. If the switch doesn’t know how to route a particular request, it broadcasts it, and records the port on which it receives a response.

MAC addresses are the address structures at the link layer, as opposed to IP addresses, which are the addresses at the internet layer. They are lower level, less structured, and less abstract. In practice, the routing that the switch needs to do to find out where a particular device is located happens early during the networking relationship. Suppose a device wants to connect to another device on the same network: typically, it will send a request to an IP address. But to get to its destination, that IP address needs to be resolved into a MAC address. That happens with an ARP (Address Resolution Protocol) request. This is what the switch intercepts.

I am only running a home network. Nothing is exposed to the public internet, and all devices on it are implicitly trusted. Unmanaged switches are fine by me.

Switches also come in a variety of speeds. I didn’t want to bottleneck anything, but I’m also on some kind of budget. I chose 2.5Gbps, because 1Gbps was insufficient, and 10Gbps was (a) overkill, and (b) costly. Our two PCs only have 2.5Gbps ethernet, my Caldigit TS3+ dock has 1Gbps ethernet, and there’s only Cat 5e in the walls. This way I can keep my prices reasonable, and not bottleneck my internet.

I went with the TP-Link TL-SG105S-M2. While it doesn’t have a snappy name, it makes up for that by being a reasonably-priced, reasonably-sized five-port 2.5Gbps capable unmanaged switch. To be on the safe side, all my patch cables are Cat 6, and I think I’m going to plug my future home server into my fibre modem.

Taking a look in the network cabinet, there was a mixture of good news and bad news. The good news is that I could see two Cat 5e cables coming down. The bad news was that they weren’t really connected to anything I’d ever seen before.

With some reverse image searching, I learned that this is apparently a 66 punch-down block. These are designed for telephone cables (as pictured, how quaint), but builders of the era sometimes parked ethernet cables in them when there was no immediate use for them, but wanted the wires in the walls. Works out well for me.

I also learned a lot of other things. In no particular order:

• I can’t reasonably use that panel.
• Ethernet cable that goes in walls is solid core and not very flexible, whereas the patch cables we use for devices are stranded. Stranded patch cables tend to be crimped into RJ45 connectors, which is what I have always called an ethernet port. Solid core internal cables do not. Solid core ethernet cables are instead more commonly terminated into the “port” version of that connector, which is called a keystone jack. (I refuse to call them “female”)
• These aren’t just normal push-through or screw-in terminal blocks. You need a special tool called a punch-down tool.

I was a little nervous about playing in this space because:

1. I rent, and this is not my infrastructure. However, I’m going to do an excellent job, and I actually unwrapped this 66 block, indicating I’m the first occupant of this apartment ever to use it. Furthermore, I’m providing a legitimately useful and obvious improvement here.
2. I’ve never done any of this before, and I’m worried I’ll do it wrong. I’ve solved this by reading a lot and watching a lot of YouTube videos.

So I went onto the internet and I purchased a well-reviewed punch-down tool, some 1ft Cat 6 patch cables, and a dual-port keystone jack (I actually purchased five of these, because they came in packs of five. This stuff is really not designed for hobbyists).

Look out world, here I come.

The bottom one pictured is actually the second one I did. I was trying to match the length of the second one for its placement in the panel, and I didn’t quite get it right. But once the cover is on, you can’t really tell. I considered re-terminating it, but I wanted to test it first.