Over 20 new femtocell vendors have appeared on the scene in the last year, but none (to my knowledge) is yet in commercial deployment. How has the landscape changed and how can existing femtocell pioneers retain their position? We spoke to several leading industry players to find out.
For the purpose of this article, we’ll restrict ourselves to mainstream 3G UMTS domestic femtocells, which are expected to ship in their millions in the near future.
(We’ll cover the CDMA market in a separate article.)
Today’s Supply Chain
To put the femtocell vendor in context, we first need to describe today’s femtocell supply chain:
- Network Operators, who buy, own and operate the full system
- System Integrators, who handle all of the back-office integration, deployment and field testing of the solution. It only makes sense to have a single System Integrator per operator. These include the likes of Alcatel-Lucent, NEC, Cisco, NSN and Huawei. They’ll assemble all the network equipment, including femtocell gateway and security gateway, which would typically be one vendor for the network.
- Femtocell Pioneers, who provide the standalone femtocell units used today.
- Original Device Manufacturers (ODMs), who can provide very low hardware cost but today have relatively little experience with femtocells or cellular/mobile technology.
- Component vendors, such as silicon, crystal and software stack.
Frankly, a femtocell on its own isn’t much use today. David Swift of Alcatel-Lucent says “For an operator to deploy it is more about an end-to-end solution that includes ‘Self-everything’ to handle the installation and on going management to ensure heterogeneous networks are created. This includes, but is not limited to,
- End-to-end network capacity optimization (xSON, WNG, PCRF)
- Access capacity maximization
- Integrated management (unify across macro, small cells, IP networks, Motive)
- Data overload prevention”
What the industry has been working towards (heavily driven by the network operators) is a more open supply chain where many femtocell vendors can actively compete to help drive costs down and increase innovation. The Iu-h standard interface should enable this, but there are strong feelings from some in the industry that compliance with the interface itself isn’t enough.
What we are seeing are several different ways being used to address the industry requirements for low cost/high volume/multiple sources of supply. These need to be balanced with the minimum expectation of end-to-end system operation in real world conditions that are essential during the early launch stages in particular.
The alternatives to designing a femtocell
There are different approaches to building and selling a femtocell:
- Develop your own from scratch (Custom).
- Base it on a reference design and standard software stack (Chip and Stack).
- Cost Engineer an existing design.
- Separate the hardware platform from the application software
Developing from Scratch (Custom)
Femtocell pioneers, such as Ubiquisys and ip.access, have taken the hard route of developing their own design and extensive software. This has been debugged for femtocell operation through years of intensive field trials and lab testing. NEC, the system integrator working with Ubiquisys, told us they had conducted interoperability testing with well over 100 different handset devices. In many cases, some workarounds are needed to cater for particular issues with individual products – with so many handsets in the field today, it’s not practical to replace or fix them.
There are perhaps four femtocell vendors in commercial deployment today: Alcatel-Lucent, Ubiquisys, ip.access and Huawei. All have reused software from other projects or brought in software expertise from partners and/or software houses to get to market more quickly.
Some of these vendors have further developed significant enterprise femtocell features, such as their grid femtocell approach, which can negotiate and handover calls between femtocells to match the coverage and capacity requirements of overlapping areas.
Chip and Stack
A rapidly growing contingent of Original Device Manufacturers (ODMs) and Original Equipment Manufacturers (OEMs) has appeared recently. Over 20 are known to be using a reference hardware design with software stack to create their own products.
Software stacks, such as those from Continuous Computing, have a long pedigree. Their Trillium stack was originally designed some 7 or 8 years ago and has been field hardened in many macro, micro and picocells before being adapted for use in femtocells. CCPU haven’t written everything themselves, partnering with TeamF1 for the TR.069 remote management and with Airhop for self-organising network functionality. To date, Continuous Computing has announced 21 femtocell customer wins and so dominate this market. Other software stack vendors, such as Node-H and Aricent, also compete for this type of business.
Many ODMs were present at the recent Femtocell Plugfest, proving interoperability with femtocell gateways and adherence to the Iu-h interface. This will help understand how mature the Iu-h standard is (whether there are too many ambiguities, uncertainties or options which were not implemented identically by different vendors).
The outcome was very positive, although participants are currently awaiting the final report summarizing total test cases and pass rates.
So what’s the difference between Chip and Stack versus mature femtocell design?, asks Analyst Peter Jarich. Andy Tiller, of ip.access claims several million lines of code need to be baked in to make it complete. Manish Singh, of Continuous Computing, counters that its extensive macro cell software library has been deployed for two decades and is therefore fully baked.
I asked Andy Tiller to be more explicit about the difference - he estimates some $10 million and at least 18 months timeframe based on his experience - specifically claiming that:
- The Radio Resource Management algorithms and packet scheduler would need to be completely rewritten
- Operators each want their own customized features and extensions
- Integration with the end-to-end femtocell solution is still required even for fully compliant Iu-h standard based products
- Field testing on the ground is required to diagnose the 1001 different radio issues that crop up in the real world but not in the lab (and then you need to tweak the algorithms and parameters to fix them)
(As you’d expect, Continuous Computing would strongly disagree).
Cost Engineer an existing design
Another scenario is for a mature femtocell design to be given to an ODM to be adapted for mass production. The ODM brings its manufacturing design expertise to bear, radically reducing the unit cost and increasing production capacity.
The same software runs on these cheaper boxes, ensuring interoperability, full range of features and simpler management.
Alcatel-Lucent have taken this route in recent months, expanding the range of femtocell hardware products which all share the same Alcatel-Lucent software release. David Swift, Small Cell Marketing Manager, explained to us how this increases choice, ensures continuity of supply, drives prices down while at the same time reducing risk and maintaining feature capability.
Split the Hardware and Software Components
Ubiquisys have gone a step further with an innovative business model called the Femto-Engine. Any ODM can obtain and build to this reference design, using their expertise to cost optimize for volume manufacture and add their design flair to the physical appearance and format. These boxes are pre-loaded with a small software boot program only.
Operators can source their femtocell hardware from any ODM willing to build to this propriertary standard. They then license the full femtocell software directly from Ubiquisys which can include some customizations or localization as required.
When the femtocell is first installed, it will connect to Ubiquisys servers using the boot loader program. This registers the femtocell, downloads the appropriate software and licenses it. The femtocell can then connect to the operators network, downloading local configuration and provisioning parameters. In this way, the latest software is always used in new femtocells, it’s customized or configured to each operator’s requirements.
SerComm have done this already, leading Ubiquisys to claim the sub $100 cost barrier had been broken with an order for 100,000 units.
The vendor landscape is going to change and it could get ugly
Some vendors believe that the advantages and experience of the early femtocell pioneers can’t last forever. They point to the DSL modem and mobile handset industries, which are now very mature and mass market with many suppliers.
“There will be casualties and dislocations in the market” says Manish Singh of Continuous Computing. “Today you absolutely need a System Integrator for end-to-end delivery of a new femtocell solution. Trying to build a femto network today without an SI is a very risky proposal – everyone today is using one. But it’s possible for operators to go direct to the ODMs in the future. There just isn’t room for all these players in the supply chain”.
“Look at parallel markets, for example DSL, which is now at a stage of maturity that operators go direct to ODMs and other box provider to build to their requirements. Eventually every successful market matures and we see high volume/low cost/low prices.”
How it might work out
Systems integrators hold the key to expanding the range of femtocell vendors they support. ALU’s approach should see additional hardware platforms running their software in use before the end of 2010. Ubiquisys Femto-Engine may find itself on other platforms which are quickly accepted as part of NEC’s solution too.
Chip-and-stack ODM’s will have to undergo a period of trials before being fully accepted into the eco-system. As the market matures and these are proven capable and complete, these ODM’s could gain direct access to the operator. Where femtocells are combined with other domestic products (set-top boxes, routers etc.), these companies can reuse their dominant market position and expertise. There is a major disagreement within the industry on whether this might take anything from a few months to 10 years to be realized.
We are at an exciting stage of the femtocell market today. Early technical hurdles have been overcome, many network operators are due to launch commercial service, a wider range of femtocell formats and opportunities are appearing. The emergence of these four difference femtocell vendor business models adds greater choice for operators seeking to expand the range and reduce the cost of these mass market devices.
The timeframe for the adoption of second and third vendor choices for network operators will depend on just how much extra software is required, how quickly the market accepts ODM designs and how important the price point is seen to be.
I think you need to separate out slightly. You have "Develop your own from scratch" but this should really be put into two different segments.
There is "develop it from scratch". Modem & stack. If you mean develop it *all* from scratch then that is a big task. Only a few people have done that: for HSPA Airvana & Huawei are the only ones I can think of.
Then there is "Chip and in-house stack".
Use a commercial reference design for the PHY (modem, NodeB) and develop your own stacks for the femto specific tasks and RNC.
The PHY is what picoChip supplies today to 20+ customers, and what Percello & Qualcomm also supply. That is a complicated system (100 million transitors, many lines of code, etc) and has a lot of its own protocol complexities. The economics of chips (fierce economies of scale) drive towards this.
This is what ALU, IPaccess, Ubiquisys use, and then develop their own code on top of, using the "known good" PHY from someone else.
Then, the "chip & stack" as the next stage is predicated on commercially available, carrier-class modem chips.
So far miniaturisation has had all the attention: how to make a tiny base station that is smaller, cheaper, faster, more energy efficient. Inevitably the discussion centres on key foundation components like chips and protocol stacks.
But a femtocell is not just a miniature base station. One reason why no off-the-shelf femto reference design has ever made it to deployment is that these components alone do not make a commercial femtocell.
What’s missing is intelligence. For example, commercial femtocells continuously sense their environment and make autonomous decisions about their configuration, about communicating with operator systems and interacting with a range of applications. Plug-and-play, shared spectrum deployment, femto networks – all rely on femtocell intelligence.
This intelligence represents the bulk of femtocell complexity. It’s a very large, very complex software system. And it’s the reason why, despite all the hype, only four companies have deployed commercial femtocells today. And yes, Ubiquisys is one of them.
@Keith: I think you are reinforcing the views in the article expressed by Andy Tiller of ip.access and David Swift. The total cost of a femtocell isn't just the hardware/softwa re price, but the cost of end-to-end service delivery - factoring in the cost of additional customer support because a femto doesn't self-configure in some cases or the benefit of additional macro network capacity freed up because a femto is smarter/better at reducing interference are likely to be taken in to account when evaluating and comparing different femto products. These could make a big difference in the total solution cost/benefit equation, especially for early/smaller deployments.