Keith Dyer, Editor of The Mobile Network, talks to Hossein Moiin, CTO, Nokia, at 5G World Congress. Here Moiin lays out what it is that makes 5G actually 5G, talks timescales and pre-standards rollouts, and identifies the most difficult element of the architectural change that 5G will require.
Can we start with the radio? There’s a lot of talk about right now of Pre-5G and 5G-ready and the like. Nokia’s messaging here at 5G World has included the claim to be demonstrating end-to-end 5G. So can you tell me what makes a radio a 5G radio?
“In order to be a ‘G’ you need a different radio and there are three elements of a new radio that come to my mind. The first is maybe we’ll use different waveforms than what’s currently used. Then we will use a flexible frame, a way to change the frame of the packets that you send so that it can meet extreme requirements. So for highly latency-sensitive real time apps we’ll use a different frame format than we would use for broadband, and a different frame for massive connectivity that we would use for either of those two.
“And the third thing is we are on a journey to cloud-based radio access networks. We have seen it is possible to decouple the lower real time layers from non-real time layers so some functionality can be centralised in a baseband hotel type of thing and the real time stuff remains at the [radio]. I think we will see that accelerating and maybe with better backhaul and fronthaul we will be able to move more and more to the centralised location. This has the advantages that you can be right next to the content, have better physical security, and overall optimise the cost per bit more efficiently than you would with existing macrocellular deployment.
In order to be a G you need a new waveform and we don’t have it yet. You need a flexible frame and it hasn’t been standardised yet.
So at the 5G World show Nokia’s messaging was that it was running the first end to end 5G on a commercial AirScale radio platform. [Here’s a picture from Moiin’s own presentation stating “5G radio access”]
But how can you have an end to end 5G call when by your definition we don’t yet have a 5G radio?
“AirScale is 5G-ready. In order to be a G you need a new waveform and we don’t have it yet. You need a flexible frame and it hasn’t been standardised yet. So in fact yes AirScale has characteristics such as Cloud RAN. It has the ability, with modification on the software side, to run different protocols and flexible frames. We’re showing 2×2 MIMO here but it has the ability to run 8×8, and further down the line 64×64 MIMO. That’s why it is 5G-Ready.”
On the new waveform, are you thinking that the new 5G waveform will be some version of filtered OFDM? [Editor’s note – Ericsson and Huawei both told me at this event that F-OFDM was the likely 5G waveform candidate]
“I think OFDM will be there for lower bands but at higher bands I am not certain and my personal thinking is probably I will not use OFDM, I will use pencil beams which will be non-OFDM. But in reality in lower bands I will probably use OFDM. But if I have maybe a GHz of available spectrum, I’m not so sure. At the moment we are in the research phase, but I believe we may have a different waveform in the mmWave. F-OFDM is what we are using at the moment here but I am not sure that will be the case – maybe things will change.”
If we can’t get it this year we need to get it next year because that is when at least one of our customers wants to deploy 5G pre-standards
And on that flexible frame for the new air interface?
We are hoping that we can get the flexible timeframes standardised in the next months – we want to get it by November but there is some work to be done to make sure it will have enough flexibility to meet all three key targets. So having said that if we can’t get it this year we need to get it next year because that is when at least one of our customers wants to deploy 5G pre-standards. But if it is so pre-standardised that there is no agreement on the basic frame, that will be an investment that is not forward compatible.”
To meet those aggressive timescales, do you think we will see a phased approach, with a Standalone 5G coming after a non-standalone version, where the 5G radio is connected to an 4G packet core (EPC)?
Standalone 5G is the ultimate target and that will mean standardisation on the core and radio, but in order to meet aggressive timescales we have agreed to a phased approach where Phase 1 will standardise much of the radio but connected to an EPC type of core. And I am hoping that flexible frames will be part of that – network slicing so much depends on that particular feature. So I am hoping that by then end of 2017 we will have Phase I completed.”
Can you explain how Networks Slicing will work end to end including the radio?
The basic idea is you have a number of assets – spectrum, cell sites with radio, baseband processing, core, applications. The concept of network slicing, my vision, is that it is about creating a virtual network with specific characteristics that have a number of requirements that it needs to meet. And in order to meet them you assemble this network from existing assets. And we have done this in the IT world, in the core – but now it is coming to transport and radio it’s far more challenging.
So here you will be using maybe some of the spectrum assets in very low bands to address the needs of IoT, then got to mmWave bands to address extremely high broadband, and to whatever spectrum for real time communications, and providing that policy is something the network OS will do. Enforcing that policy is something that’s done at the edge of the network – cloud based would be a good way to do it. Eventually different behaviours from elements – what wave, power, frequency you send out and receive – these to me can be programmed, it’s just that we haven’t done it so far. And this is in a roundabout way coming back to the concept, which is not a new concept, of Software Defined Radio, that will make it real.”
In your talk you said that one of the necessary conditions for commercialisation of 5G will be the creation of a new network architecture, and you named seven things that have to change there. Which of these will be the most challenging?
I think they all have their own challenges, the challenge of creating distributed edge cloud should not be underestimated because that means using the central office like mini data centres. Creating a core for mobile and fixed access, using AI to automate network functions service provisioning, we have ideas but we haven’t done it yet.
I think the most challenging however is on digital security, and security of networks is becoming imperative in order to make sure that they can be used in sensitive domains. So we talk about a common fabric and that common fabric means trusting other users of the fabric that they will not intrude on your network and functionalities and we need to make sure that this can be automatically prevented and minimised. To me security looks like the biggest challenge of the seven.”