Qualcomm makes radio baseband chips that small cell developers have been integrating in their products for years. They are, in essence, slightly scaled up versions of the modems and RF units it puts in its phones.
But in October 2020, Qualcomm took a further step into the RAN market for 5G. It said it would be expanding its chip presence into three new platforms that sit within the Open RAN architecture.
One of these is a chip platform designed for a Distributed Unit (DU). It is also creating a platform that can go in a Radio Unit (RU). The third product area Qualcomm is supporting is a combined DU-RU (DRU). This option puts the DU and the RU in the same “box”, with an interface back to the CU.
One thing that Qualcomm says it that it has products that can support macro as well as micro RUs. That means it is developing silicon that can support macro power levels and Massive MIMO configurations. So how will Qualcomm look to play in the RAN market – what is its intended role?
In discussion with operators it was a very clear message that there is no way that the current solutions out there that enable vRAN operation can compete in terms of power and performance with the vertical solutions
Gerardo Giaretta, Senior Director Product Management, spoke to TMN to explain the company’s ambitions and how it wants to play in the Open RAN ecosystem.
“The overall silicon ecosystem right now for base station is almost bifurcated to very, let’s say performant, very ad-hoc type of solutions that Ericsson and Huawei may have, versus something that is more commercial off the shelf and related to vRAN and Open RAN, like the Intel platform and so forth.
“So we think that the vRAN and Open RAN system, because of the flexibility of the system, is the way that the network should eventually be deployed. But in discussion with operators it was a very clear message that there is no way that the current solutions out there that enable vRAN operation can compete in terms of power and performance with the vertical solutions. So that’s the gap that we want to address.
“And that’s why the massive MIMO, high performance types of use cases are the ones we are really focussing on. We really want to bring vRAN and Open RAN, and it will take a few years, to the same level of performance that an integrated solution will have. That’s why we are entering, or re-entering I should say, the business of macro.
“We are not going to do vRAN software, we are not going to compete head to head with Intel on x86 server class platforms. The idea is to focus on the PHY layer, Layer 1, accelerators and things that are more to do with the RF – the areas where we think we can really make a difference.”
“Also, we are not going to develop radios, RUs ourselves. We will develop SoCs that our customers can use to develop radio units. And those customers could be new entrants or incumbents. It’s really a silicon SoC solution that will enable [units] up to those use cases (64T/R MIMO).”
“Take the DU platform, it’s not a full alternative. The idea is not that we are going to have a complete solution – the DU is a complex system depending on which O-RAN split you are looking at. So the question is how far can you drive an x86 platform to do the higher PHY layer functionality for many carrier, high capacity type use cases? And right now the answer is “Oh we are going to offload all the coding, the FEC (Forward Error Correction), right? If you really want to control three sectors of massive MIMO it simply doesn’t scale as well.
That doesn’t mean that our solution completely replaces an Intel solution that’s not the plan. It’s almost like this is a solution where Intel and Qualcomm can work better in the long run
“There are two models of accelerator. There’s the look-aside accelerator – which is what Intel has been working on, offloading only the FEC, and then there’s the concept of the in-line accelerator, where you are going to offload most of the user plane functionality on the PHY layer. And that’s what we think is the right approach long term, and that’s what we are pushing with our product.
“That doesn’t mean that our solution completely replaces an Intel solution that’s not the plan. It’s almost like this is a solution where Intel and Qualcomm can work better in the long run, because Intel can focus on what they are doing best – server class architecture and so on – and we can focus on what we do best, which is more the real time sub-millisecond processing for multiple layers that you have to do on the PHY layer.”
“The DRU platform that Qualcomm is proposing is more in support of a small cell split known as Option 6 type product, but in this case targeting to a macrocell where a developer wants to have a more compact design and lower cost. So again it’s all about how much we can integrate into the SoC, integrating the full baseband into the SoC.
“In general the reason why we are announcing we are developing flexible platforms is that nobody clearly knows how these Open RAN networks will be deployed. There is a lot of focus right now on 3GPP Split Option 7.2, that’s the where the standard is and the starting point, but at the same time we are hearing a lot of interest in going beyond that depending on your backhaul capacity, how much do you have to optimise that and so how much processing do you end up packing in the RU versus the DU. So there are a lot of unknowns there too, and that’s where we believe the flexibility of the platform will provide value.”
Qualcomm has said it will have its 5G RAN platforms available from 2022.
* This article is extracted from TMN’s Open RAN Market Report 2020, which is available for free download here >>>