At the DAC theater, Cadence's Ian Dennison talked about 5G Intelligent System Design. He repeated his presentation internally at Cadence a couple of days later. Plus, I sat next to him at both CDNLive EMEA and DAC, while I signed books and he talked about 5G. So I'm going to steal the first half of his presentation, which was more of an introduction to 5G. Today and tomorrow's posts will give an overview of what goes into 5G. I'll cover the second half of his presentation, on how you can use Cadence's portfolio of tools and IP to accomplish a 5G design in a separate post later in the summer.

This is not the first time I've talked about 5G. For example, see:

• CES: 5G, All Hat and No Cattle from CES earlier this year.
• CDNLive: Ericsson Paving the Way for 5G the Ericsson view of the world of 5G.
• 5G is Coming. Qualcomm Keynote at CES the Qualcomm view of the world of 5G.
• 5G, Coming Soon to a Phone Near You from 2016, a snapshot of how the future looked like three years ago.

What Is 5G Going to Do for Me?

The diagram above shows the opportunity of 5G:

• 10X peak data rate (20 Gbps)
• 10X lower latency (1 ms)
• 100X connected devices in a given area

Those are the headline attributes for users. Other attributes, more of a target for implementation, are:

• 90% reduction in network energy usage
• 10-year battery life for IoT devices
• 100% coverage and 99.999% reliability
• 1000X bandwidth available in a given area

I like to say that there is only one reason that new generations like 5G come along, and that is to take advantage of more powerful silicon to make better use of the radio spectrum, which is the real limited resource in mobile. 5G is no exception, going to steerable beams, and more complex encoding schemes with 16 bits per symbol. The rest is all marketing! Well, not quite. Read on.

New Architecture

4G is built using those towers that you often see at the side of the freeway or sometimes on hilltops trying to pretend they are trees. At the top of the tower is the mast radiohead with the antennas and some minimal electronics. From there the radiohead is linked to the per-mast baseband, usually at the bottom of the tower, using coax. It seems overkill to call this the fronthaul when it just runs down the tower. The basestation baseband is typically a large rack of electronics and is connected to the internet backbone through regular telephone network technology. It might be fiber, it might be copper, it might be microwave. The basestations are required every 2-5 km in most areas, although in rural areas they can be more widely spaced.

5G consists of two network technologies with very different characteristics. There are actually three bands, known as low frequency, medium frequency, and mmWave. But the low and medium frequencies are usually grouped as sub6Ghz or <6GHz. The low frequency consists of the old spectrum from 4G and 3G repurposed to 5G as the migration takes place. The medium frequency is new spectrum. The maximum frequency is 6GHz. The lower frequencies are actually more valuable since they transmit further and better, but the same basic radio technology is required for the whole frequency range. The sub6GHz can be handled using the same approach as 4G, reusing the towers and buildings, but with new electronics.

However mmWave is not just a little higher frequency, it is over 24GHz. There are neither 4G nor 5G between 6GHz and 24GHz. The military and other things have that. If you want to see just how complex the spectrum map is, then see my post from a couple of years ago GOMAC: A Conference that Starts with the National Anthem.

But mmWave, at its very high frequencies, has two major attributes, one good, one bad. The good one is that there is a huge amount of spectrum up there. For example, the band for 4G data is 1.7 to 2.1GHz, so 0.4GHz bandwidth, whereas the 5G mmWave goes from 24.5GHz to 28.35GHz, so nearly 4GHz bandwidth, 10 times as much capacity.

The bad attribute is that almost everything attenuates mmWave. In particular, buildings and hands. This means that mmWave cannot be used outside a building to provide service inside a building. Handset design also needs multiple antennas just to take account of how you hold the phone. But wait, it gets worse. Air (actually oxygen) attenuates mmWave so severely that the range is limited to 2-300m (about 700-1000 ft).

So basestations need to be small cells every 200m or so. In rural environments, there simply won't be any mmWave provision. In urban environments, there will need to be cells on every streetlight or building. The 4G architecture of having the baseband at the bottom of each radiohead is clearly not going to work. Instead, the implementation technology wiil be C-RAN, which stands for Cloud Radio Access Network (or Central, if you prefer). I first wrote about C-RAN three years ago in Infrastructure: Connecting Mobile to the Cloud. The basic idea of C-RAN is to do as little processing as possible at the radioheads (since there are so many of them) and do as much as possible in the shared baseband unit so that it doesn't need to be duplicated at the top of every streetlight. It is very similar to the idea behind cloud datacenters for regular computing. Each small cell is linked to the baseband by fiber. The baseband, which is shared among many cells, provides ultra-low latency and, probably, some level of edge compute (that some people call fog, since it's neither the edge nor the cloud). That is connected to the internet backbone in the same ways as 4G, although being more modern it's more likely to be fiber.

So-called "fixed wireless" can be used to replace last-mile copper to homes, by putting one of those small cells near a group of buildings with the equivalent of a cellphone on the outside of the building (remember, mmWave doesn't penetrate buildings). The technology is actually simpler than a cellphone since it doesn't move, or need to switch between basestations (and it certainly doesn't have to play Fortnite).

In large buildings, such as offices or conference centers, the network service providers expect to use large numbers of femtocells, which are even smaller cells (that can probably self-configure so installation is trivial—maybe they go in light fittings). But this is not without controversy, a topic I'll touch on tomorrow.

Another big difference between 5G and 4G, especially in the mmWave bands, is that both handsets and basestations need to steer beams and direct all the radio energy in the correct direction rather than simply broadcasting in all directions and accepting that most of the energy is wasted going in the opposite direction from what is required. This requires going beyond the multiple antennas of 4G, known as MIMO, and going to steerable phased-array antennas. Beamforming is more complex than it sounds since it also wants to take advantage of reflections, so even a handset may be tracking multiple beams. Of course, the small cell basestations (on the lamp posts) may be tracking hundreds of users simultaneously, each with multiple radio paths.

This is another example of my contention that generations are about making use of lots of computer power to enable more efficient use of power and bandwidth.

Putting that together you get a 5G network that looks like this:

• Existing masts provide 4G and 5G service (gradually moving the mix towards 5G as the handset market transitions)
• There will need to be many new femtocell radioheads inside buildings (mmWave doesn't penetrate walls and <6GHz won't have enough capacity)
• Radio energy has to be concentrated into beams.
• Beams enabled by large arrays of small antennas.
• New mmWave radioheads much more deeply embedded in urban environments (every 200m, so every lamp post)
• Fast download speeds of up to 20Gbps, upload may be slower.
• Optical fronthaul for low-latency between baseband and radiohead.
• C-RAN shares baseband across many radioheads for performance.
• Edge compute for massive AI and lower latency than going all the way to a cloud datacenter.

For More Details

See the Cadence 5G page.

Tomorrow

Part 2 of this introduction to what 5G is, and what to watch out for when people start trying to sell it.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.