There is a huge range of free OpenSource software and hardware designs for many applications. We looked at whether there’s enough available to build your own LTE small cell and found it wanting. At the moment, you’d be better off buying a professional product unless for lab and experimental purposes. Meanwhile several well-funded projects with substantial support from Facebook and several large operators are planning to change that, ultimately aiming for commercial deployment within major mobile networks.
The mix of truly open source vs commercial product
Many open source projects will have a commercial element to them. Typically you get the core software for free and then either pay extra for commercial add-on modules or for professional support. A typical example is the popular Wordpress website software – the excellent core software together with many basic themes and plug-ins are free, but the more sophisticated, specialist or robust modules cost extra. For those running a commercial website, professional 24x7 support including bug-fixing is available for a fee.
Small cell open source software is also available in a similar guise but not with the same level of maturity. There are several projects that allow you to download and operate a simple system suitable for lab testing and experimentation, but you’ll have to look to professional equipment vendors for products you can use in the field.
Choices for GSM
We’ve looked at GSM before, and the OpenBTS or Osmocom projects still seem to remain the most popular places to find code for that. There are several RF hardware boards (Software Downloadable Radios) that can plug into a PC or laptop to provide a basic platform. Vendors such as Fairwaves, NuRAN and others ship functional equipment with RF power up to 10 Watts at a price.
Generally speaking, you would require permission to use licenced spectrum to operate this equipment. Low power test licences can be arranged in many countries for a small fee.
Choice of hardware
Some people will tell you that commodity general purpose processors are now powerful enough to support a standalone small cell. But could it really be true that the huge efforts put in by dedicated silicon vendors over the years (Intel, Qualcomm, Broadcom, Cavium etc.) was for nought? The reality is that to process and decode the full stream of raw radio data in real time takes a huge amount of processing power with some specialist firmware and optimisation.
One option I’ve written about in the past has been to connect a relatively dumb RF board to laptop or PC and download open source code. This is quite feasible for basic GSM voice and text service, using the OpenBTS project but don’t expect too much from the GPRS data features.
LTE gaining traction
If you want LTE, then you can find code available through the Open Air Interface (OAI) project complete with instructions. I spoke to independent consultant Phil Claridge who spent some long evenings figuring out exactly how to configure and tune the system to provide a standard LTE end-to-end data connection.
Choice of front end RF hardware boards include the Ettus USRP 210 (approx. $2000) and the Lime SDR-mini ($139). For the latter, you’ll need as good a PC platform as you can find, with an i5 Intel processor really the minimum. Turn off all the power-saving features and expect it to run hot, because it’s going to be performing continuous fast-fourier transform calculations on the data stream regardless of whether there is any user data or not.
We came to the conclusion that while such open source code is freely available and may be worth playing with in the lab, it’s not yet at a point where you could download and run your own LTE small cell as part of a working network. You’ll continue to need assistance and probably some specialist hardware designs from elsewhere.
This is what some influential companies are trying to change.
Facebook, through their OpenCellular Project, is encouraging a more fully functional open hardware platform design. They partnered with NuRAN to build the first commercial versions which are now becoming to fruition.
NuRAN announced a 2G OpenCellular basestation (OC-2G) in June, which uses their own proprietary software stack. They also have their own separate LTE product.
The OpenCellular project is developing a flexible open-source hardware platform for a 4G basestation (OC-SDR) based on Cavium chipset. Fairwaves, another opensource basestation vendor, expects to be trialling units in the field in Africa soon.
Facebook has stated they will contribute all intellectual property and designs into the Telecom Infra Project (TIP) which aims to develop a new approach to building and deploying telecom infrastructure. It’s hard to establish how much open source vs proprietary software will be involved, or whether the hardware design will be limited to the Cavium chipset.
The Open Networking Foundation (ONF) has evolved from purely focussing on network operating systems into a few high profile projects. The awkwardly named M-CORD project (Mobile Network - Central Office Re-Architected as a Datacentre) includes the full end-to-end scope of a mobile network, from cellsite to IMS and VoLTE core. The organisation has around 30 full time employees and aims to stimulate open source contributions from hundreds of others. It aims to release a sister project R-CORD (for residential wireline broadband) by next year, with a timeline of 2-3 years for M-CORD aligned with its goal of 5G.
It will support the hardware platform of TIP (above) for radio nodes as well as other choices.
There’s a heavy influence of new technical architecture throughout, looking to introduce all the latest and popular methods of cloud computing, virtualisation, network slicing, mobile edge computing and programmable RAN. Goals include setting an architectural foundation for 5G.
What’s slightly different with their approach from pure opensource projects is that their framework architecture supports a choice of both free-of-charge opensource modules sitting alongside chargeable binary downloads. You can choose which to use for your application.
Some of the truly opensource free modules may be less functional or based on earlier standards releases. For example, I understand that Radisys has donated source code for a small cell stack, but only with up to Release 9 functionality and not their latest version. Sprint and Intel have donated a “deployment grade” opensource EPC stack. Netsia, Accelleran and Quortus are just a few of those offering their professional software as binary downloads for a licence fee.
It seems to me that this choice of paid vs free is a good concept and at some point we might be able to download small cell software that runs on low cost SDR modules to build our own private LTE network. Perhaps CBRS will enable certain configurations to run in lightly licenced spectrum. In the future, I could see this potentially being attractive for MulteFire running in unlicenced spectrum.
At the moment, I’d see this primarily for use in experimental lab testing by academics. It’s still going to be some time before you can download and deploy your own opensource cellular network. When these new frameworks and modules become commercially deployed by live networks, you’ll know they are mature enough for more general use.
I suspect there will be a big gap in functionality and features between the free opensource code and the proprietary versions for quite some time. For more basic applications, perhaps the free software will make cellular service more accessible in the most undevelopment or commercially unattractive locations.
The M-CORD framework may help bring more startups and newer vendors to market, especially if it encourages interoperability and ease of end-to-end system management.