Cohere Technologies has been an often-heard name throughout the stages of defining 5G standards and use cases, most recently because of new trials with Deutsche Telekom.
Last month, the RAN start-up announced new functionality for its key technologies – Massive MIMO scheduler and Delay-Doppler channel software. It also said it had received new funding (without specifying how much), and said its offering would be launched commercially later this year.
CEO Ray Dolan, who previously headed up mobile broadband chip start-up Flarion as well as Sonus Networks, may be hoping he can repeat Flarion’s achievement of attracting a high-priced acquisition from Qualcomm – now that the mobile chip giant is moving back into network infrastructure chips. Flarion contributed highly valuable patents to the Qualcomm arsenal (at a time when the larger firm was trying to transition from CDMA to OFDM), and so didn’t need to bring large-scale revenues to the table. Cohere, also, has large numbers of patents and expertise in core 5G technologies, but Dolan is seeking to build up the commercial success quickly anyway.
“I knew this was going to take a while,” he told LightReading. “The question is: Can we get commercial traction?” He said that, by the end of this year, the answer will be clear, as Cohere expects to announce several large operator engagements.
It will be celebrating its tenth birthday by then, and has had several setbacks en route, including an attempt to have its modulation technology accepted into 5G Release 15 standards, and then a pitch to the fixed wireless market (see below). Dolan was appointed in late 2018 and set about refocusing the company and, like many start-ups, jumping on the Open RAN bandwagon to try to attract operator and partner attention and get involved in a more open environment for trials and sales.
It has also defocused on its previous hardware-based products and has built its new offering around its Spectrum Multiplier software platform, which is claims can double overall spectral efficiency and so support high network speeds in conjunction with 5G-centric technologies like Massive MIMO.
Cohere’s software approach is central to an important trial that DT has been running with the start-up, plus Intel, VMware and Mavenir, for over a year. The German telco is
exploring one way to address the higher costs of general purpose processors (GPP), compared to dedicated ASIC chips, in a high performance RAN. While the ability to run RAN functions as software on COTS hardware is central to the vision of a fully open, virtualized network, it does run up against the high cost of the chips – processors and accelerators – needed to support the very demanding tasks involved in running a RAN, especially those related to low latency and constant updating.
While acknowledging the issue, Alex Choi, SVP of strategy and technology innovation at DT, believes one solution is to offset the high cost of the processors by using the hardware far more efficiently. This is the focus of the project, and the resulting system, based on Intel’s FlexRAN stack, completely separates the radio resource management (RRM) from the underlying hardware and runs it as a discrete application on top of the VMware virtualization platform.
The efficiencies were increased by implementing an advanced RRM algorithm from Cohere. The result, said Choi, was that “in total the overall cost can be lower than in the case of using a traditional platform” because of the improvement in radio performance. Choi said throughput doubled on the test network, leading to a reduction in total costs.
Its next step is to open up the platform and invite others to develop RRM software so that this approach can become more mainstream and competitive.
This has been a welcome endorsement for Cohere and a chance to put its algorithms through their paces.
As mentioned above, the company had high hopes of getting its OTFS (orthogonal time frequency and space) air interface technology into 5G standards when it was supported by AT&T, China Mobile, Deutsche Telekom, Telefónica and Telstra as an alternative air interface to OFDMA (OFTS actually sits on top of OFDM, ensuring backwards compatibility, but claims to be more reliable). In the end, the 3GPP RAN 1 group played it safe and stuck with extensions of OFDMA, which led then-VP of business development, Anton Monk, to describe Release 15 as “LTE with Massive MIMO and beamforming — nothing really new except for including Huawei’s polar codes in the control channel and using LDPC (low density parity check) everywhere else”.
Having failed at the 3GPP, Cohere focused on building equipment that incorporated OTFS, particularly for fixed wireless access (which has the advantage of not requiring standardized devices like mobile cellular). Its technology has been trialled by Telefónica, Telstra, Altice, Sprint, DT and Charter but made little commercial headway.
Under Dolan’s watch, the OTFS technology – complete with Cohere’s own Delay Doppler-based channel detection, estimation and prediction technology for beamforming – was made available as a software module that can be integrated into existing RANs or placed in an x86-based server and connected via interfaces such as O-RAN. In a vRAN architecture, the code can run as an application on the RAN Intelligent Controller (RIC), which is becoming an important 5G element because of its near-real time performance for low latency environments.
“Currently, 5G offers only about 20% more capacity than 4G,” Dolan said last year. “But the radio is not where the magic is – that’s in the beamforming and you can’t do that efficiently with TDD [time division duplex] because currently it’s all done by the likes of Ericsson, Huawei, Nokia and Samsung at the cell site. Cohere’s technology can be deployed in the cloud and enables the spatial replication of radio. We believe we can double network capacity initially – significantly more as more antennas are added.”