5G

5G, the “fifth generation” cell phone standard, has been in the news a lot. There are people suspecting 5G of causing the coronavirus, and even reports of some of them burning cell phone towers in the UK. But this bizarre suspicion was by no means the start of the bullshit about 5G. Indeed, it’s a natural reaction to the previous over-hyping.

5G, according to the hype, is to provide much higher bandwidth and much lower latency. It will enable self-driving cars; it will enable the Internet of Things. It will enable remote medicine; it will enable artificial intelligence; it will enable quantum computing. All the nations need to race to move to 5G, and the winner will dominate the world. And China is the first to field 5G!

The reality of 5G, as it’s currently being fielded, is that it’s about the same as 4G. It uses similar radio frequencies: any of the frequency bands it’s using could just as well be assigned to more 4G. It uses the same way of encoding digital data (OFDM); in previous generation changes there were major improvements in coding efficiency, but not this time. 5G uses MIMO to boost spectral efficiency, but so does 4G. (There are different levels of MIMO, and top-of-the-line 5G phones will presumably take it farther than has been done before, but this is normal evolutionary improvement, not some quantum leap.) As regards latency, 4G signals already move at the speed of light. And 5G uses about the same power levels as 4G, so health concerns are unchanged. (Not that there seems to be any basis for concern in the first place; it’s hard for microwave radiation to have any consequence besides simple heating.)

This raises a social question. Do any of the people hyping 5G remember back three decades ago, when there was a big fuss about “fifth generation computing” and the Japanese threat? The Japanese, it was said, were not just going to eat our lunch in building DRAMs (which they did), but to leapfrog us in computing in general, going straight to “fifth generation” computers – something to do with artificial intelligence – and dominating the computer industry. I don’t know what the Japanese actually ended up doing in that regard, but I do know it never made it to the market; instead it quietly sunk without a trace. Now it’s China’s turn to be the great menace. What is it about Orientals and the number five that is so conducive to fearmongering?

Anyway, enough amusement; time to unpack the reality of today’s “5G”, which isn’t going to just quietly sink without leaving a trace; it’s a solid, serious incremental improvement, not a vain leap into the unknown. For starters, 5G has nothing to do with quantum computing. It also has very little to do with artificial intelligence, which might benefit slightly from faster communications but is predominantly a computational problem. (Incidentally, today’s “artificial intelligence”, which started out being called “neural networks”, is entirely different from the “artificial intelligence” that the Japanese were going to conquer the computer industry with; that was more of a rules-based logic manipulation exercise – a binary approach rather than a numerical one. As usual, the numerical one works much better.) As for remote medicine, if you’re going to drag a surgical robot and its support staff into some remote place, you can take a surgeon along with it, rather than trying to rely on a high-bandwidth, low-latency link that enables the surgeon to operate remotely. Given the risks of wireless communications being lost mid-operation with the patient open and bloody, you probably should do that anyway. Other sorts of remote medicine require only today’s levels of bandwidth and latency.

When it comes to improving bandwidth, in the long run 5G is indeed promising. But the improved bandwidth is mostly to be gotten by moving to higher frequencies, where unused bandwidth is more plentiful. This is not a quick move: engineering circuits for higher frequencies such that they can be built at low cost and with high performance is not a trivial task. It’s not a showstopper, and we can expect it to be done successfully, but it’s not a wild land race either. Even when it is eventually accomplished, it will have a downside: parasitic energy losses are generally greater at higher frequencies.

Another aspect of operating at high frequencies is that radio waves there are more directional. This already is something that people experience with WiFi, where the 2.4 GHz band is more crowded but is still sometimes easier to use than the 5 GHz band since the signals bend around corners better. With cell phone service, higher frequencies mean that towers have to be more closely spaced: the signals don’t curve around the Earth’s surface as well. But there’s an upside too: antennas are more easily made directional. It’s common these days for a single tower to have antennas pointing in different directions, thereby serving several cells from the same tower. With antennas that are more directional, the number of cells could increase. Even the phones themselves could start using directional antennas, so that their uplinks to the tower would not interfere with each other. These antennas would be of the phased-array sort, made of multiple small emitting antennas to electronically steer the beam, and with it being steered so as to always point at the tower regardless of how the phone was oriented (and despite the interference from the user’s hand). Such a phone, with different software, could be used as a radar. But it would require more hardware (many antennas, plus driving circuitry for each), and might never be economically viable.

All these factors (more unused bandwidth, towers being more closely spaced, and more-directional antennas, even if only on the tower) multiply together to make for a lot more usable bandwidth to be had at higher frequencies… but only eventually, and at a cost.

The promises of hugely decreased latency from 5G are more of a mystery. For those unfamiliar with the term, bandwidth is the number of megabytes per second that you get; latency is the time it takes to get the first byte back. As mentioned, they aren’t going to get 5G radio waves to go any faster than the speed of light. The backhauls (the fiber optic links from the cell tower to the rest of the world) aren’t any different in 5G, either. So whence the promises of drastically lowered latency? Nothing has really come my way via my usual sources to explain this, and doing some web searches has yielded little in the way of an answer.

Perhaps what they’re talking about is that 5G allows the network to set aside a portion of the bandwidth to be used for things other than communication from cell phone to tower. It could instead be two self-driving cars next to each other on the road talking to each other directly to coordinate their actions. Since the transmission would be direct car to car, the radio round-trip time would be miniscule, with no tower overhead. This is also presumably where the talk of 5G enabling self-driving cars comes from. But in order for this to actually happen,

1. the cell system operator has to actually set aside that bandwidth,

2. there have to be standards promulgated for inter-car communications, and

3. car makers have to manufacture cars built to that standard.

All this is a lot of work, not something that immediately follows from adoption of 5G; and given the number of different entities that would have to cooperate on it and the time that would be needed to get everything built (it takes years to get new cars through design and into production), my guess is that this part of the 5G standard will remain unused. Cell system operators are huge companies which are quite reluctant to give away any of their precious bandwidth. Self-driving cars will have to cooperate with human-driven cars, so they can’t rely on car-to-car communications to solve any of their problems. Since they’ll have the sensors and systems to deal with uncommunicative cars, they might as well use them for communicative ones too. There are also security risks in communications: the other participant in the wireless conversation might be a hostile hacker rather than a normal car. In general (not just for cars), this setting-aside of bandwidth seems likely to become one of those parts of the standard that nobody actually ever uses. If it does get used, that’ll only be after much additional work to flesh out the details.

In sum, 5G is really nothing that any ordinary person needs to care about, except when shopping for a new cell phone and worrying about compatibility. It’s evolution, not revolution. But instead we have this mixture of crazy hype and crazy fear. The latter should be a lesson to the people behind the former: something that’s promised as doing everything can be feared as doing everything.