19 June 2012
How do you develop a standard that meets the need for thousands – maybe millions – of devices to connect without human intervention? Welcome to the ‘Weightless’ world.
When machine-to-machine (M2M) connectivity problems are discussed there tend to be two responses: one is to note that there have been predictions of inter-machine connectivity for years and the fact that the technology has failed to deliver on them suggests that an inherent impracticality exists that technology cannot overcome. The second, more positive response is to say that cellular transmission systems can handle it all, but that they will need major reinvestment along with backhaul support from other communications infrastructure. The flaw in both viewpoints is that they overlook the requirement of a bottom-up assessment of what ‘machines’ would actually require; these requirements should then inform the baseline specification of an M2M communications standard, and explain if existing technologies are, or aren’t, up to the task.
There are many possible M2M applications. Common basic characteristics tend to be: very low cost, both for the chipset and the annual fee for sending data; ubiquitous coverage – even better than cellular; and – in some cases – battery life of around 10 years. Challenging enough starting points, but there are also some characteristics of M2M traffic that can be exploited in system design to make them more achievable: most messages are very short; delays of a few seconds are rarely problematic; data rates can be low; sleep times in some cases can be long; and seamless handover is not required.
This list supports the argument that there is no existent wireless communication standard that meets these requirements. Cellular provides coverage to almost the level needed, but cannot achieve the cost points and prolonged battery life. Bluetooth and Zigbee do not provide the range and coverage needed. Systems like Paknet do not have the capacity or low device costs.
So a new standard that exploits the simplifying characteristics of machine communications in delivering the cost, coverage and battery life needed is proposed in the shape of ‘Weightless’, now being developed as an open standard in a manner similar to Bluetooth. The Weightless standards body – the Weightless SIG – has just issued version 0.7 of the specification, and expects to publish the first complete and stable specification (version 1.0) in early 2013. It is likely that the specification will then be ratified by a body such as ETSI to turn it into a formally recognised standard.
What defines a ‘machine’?
Mobile phones allow people to talk or send emails and texts, Wi-Fi systems allow people to browse the Internet from a laptop or netbook, and Bluetooth links let people use cordless headsets etc. The M2M model envisages an entirely different class of applications for devices that do not directly have ‘users’, and whose communications are not instigated or mediated by people.
The smart meter is a prime example. It might send meter readings to a database every hour. It has no direct linkage with humans – although indirectly it makes their life better by enabling smart grids and automating the meter reading process.
There are so many different M2M ‘machines’ and applications that, arguably, a clear definition of both is not possible right now; but broadly, these are devices where transmissions occur due to the function of the machine rather than a person.
Applications include automotive engine management updates, healthcare monitoring, smart city sensors and actuators, smart grids, asset tracking, industrial automation, and traffic control. They send information not to another person, but typically to a database within the network from where it can be processed by other machines. Of course, sooner or later someone benefits from the service provided, but typically not from the radio transmission itself.
Here’s where the term ‘M2M’ starts to become redundant and confusing, as it might be taken to imply one remote device talking to another remote device (one smart meter sending data to another); in practice, the communications are typically machine to network, or machine to control node. In the context of Weightless, the term ‘machine communications’ is used in preference to M2M (although for the time being M2M has established itself as the de facto ‘label’ that describes the technological context within which Weightless would operate).
Defining distance issues
Machine communications fall into two key types – short or long range – as regards the use of wireless communications. Short-range applications are typically those that occur within a building – home or office. This might include wireless control of in-room lighting from a central controller on the premises. The range of such applications is typically around 100m; this can be extended using mesh-architectures or repeaters within the building.
This type of solution is sometimes called a home area network (HAN). There is a range of technologies already available for HANs including Zigbee, Bluetooth, Wi-Fi, as well as some proprietary offerings.
Long-range applications are those that need to communicate wherever they are and sometimes as they move around. Automotive is a good example of such an application. They often require near-ubiquitous coverage of a country or even global reach. As a result, they require the deployment of a cellular-like network and many similar arrangements as for cellular communications, including billing and roaming.
This begs the question: why not use cellular networks for long-range machine communications? Indeed, cellular networks are already used for some machine applications such as monitoring vending machines. They bring the benefits of good coverage and availability, plus widely-available components. But cellular is far from ideal for machine communications. Coverage is imperfect, especially within buildings. Terminals cannot run off batteries for extended periods. Cellular networks are not well adapted to short messages, and so are very inefficient for most machine applications. Treating each terminal as an individual ‘subscriber’ adds costs (including SIM cards), expanded billing systems, and more. And cellular networks are moving towards higher data rates, away from the functionality profile required for machine communications.
Weightless is a standard for long-range machine communications. Determining whether an application is short or long range is not always simple. For example, a smart meter in a home could communicate via the HAN and then the home broadband connection into the network. However, this is complex to set up, and leaves the energy supply companies vulnerable to the home user changing their HAN, or even just the password on their Wi-Fi router, and disabling the smart meter. For that reason, long-range communications are preferred for many devices even within the home. Weightless networks would need to be deployed by network operators and a service provided to companies interested in machine applications.
One of the key design aims of Weightless is to be application-agnostic – that is to provide a platform on which as many machine applications as possible can be based. One shared platform across multiple applications is clearly much more economic than separate networks for each major application. Hence, Weightless has not been designed with any one specific application in mind.
Immediate design implications
One of the most important Weightless requirements is the need for ubiquitous coverage. This implies a cellular architecture. Along with this comes the need for a network, roaming, authentication, billing, and many other aspects of cellular technology.
It implies that at a high level the system architecture will look very similar to that of conventional cellular systems. However, as will be discussed below, the scale of the various network components can be much reduced compared to cellular.
An extra option has emerged for spectrum access: the use of the ‘white space’ spectrum – the unused portions of the spectrum band in and around TV transmissions. White space spectrum meets all of the requirements for machine communications.
It is unlicensed – access to it is free. It is plentiful with estimates of around 150MHz of spectrum available in most locations – more than the whole 3G cellular frequency band. It has the potential to be globally harmonised, as the same band is used for TV transmissions around the world. And it is in the perfect low frequency band which enables excellent propagation without needing inconveniently large antenna in the devices. So access to white space provides the key input needed to make the deployment of a wide-area machine network economically feasible.
Achieving coverage even deep indoors has a further implication. Cellular systems have relatively poor indoor coverage and white space transmitters will typically be restricted to lower power levels than cellular base stations. One solution would be smaller cells; but the result of this would be a costly network deployment. Instead, Weightless needs to achieve better coverage than cellular with fewer base stations – and less transmit power.
This can be achieved with spreading. Direct sequence spread spectrum (DSSS) multiplies each transmitted symbol by a codeword resulting in either a high-transmitted data rate or longer effective bit duration. This enables range to be extended at the cost of data rate. It is a technique employed in GPS transmissions to allow the weak satellite signal to be received with a handheld device at ground level. Spreading can achieve a 30dB gain in link budget or more – sufficient to achieve the objectives set out above. It has other design ramifications, however.
The need for devices to work from batteries for years, and the regulatory restrictions that result in lower power for the portable devices, cause further challenges. With more powerful base stations than terminals there is a risk of an unbalanced link budget where the terminals can hear the base station, but not vice versa. In Weightless it is normal for the base station to be transmitting at 4W EIRP (36dBm), but the terminal to only transmit at 40mW EIRP (16dBm) resulting in a 20dB difference in the link budget.
This can be accommodated by using narrower bandwidth channels on the uplink resulting in a lower noise floor at the base station receiver, and enabling the SNR targets to be achieved. Using uplink channels of 1/64th of the bandwidth of the downlink provides a noise floor 18dB lower which approximately balances the budget. This approach works well with UK regulations, but US regulations measure transmit power in narrower bandwidth which mitigates against using narrow channels.
White space flexibility
Another implication of the use of white space is to adopt time division duplex (TDD): the availability of two appropriately spaced white space channels as needed for frequency division duplex (FDD) cannot be guaranteed. TDD also provides flexibility in that at the time of design it was far from clear what the balance of downlink versus uplink traffic would be on a machine network. A further implication of white space for the initial design was a need to be able to avoid random interference from other unlicensed white space users.
The classic approach to this, used by systems such as Bluetooth, is frequency hopping. Hopping also brings other benefits like averaging of self-interference, good neighbourly behaviour to other white space users, and mitigation against being stuck in a fade.
White space operation also strongly biases designs towards structured synchronous solutions where there are frames, frame headers, and devices are provided with allocations rather than transmitting randomly. This is because base stations must communicate information to terminals such as the frequency hopping pattern that is in use and in some cases restrictions on transmit power.
With such a structure in place it then makes sense to schedule traffic rather than allow devices to transmit whenever they wish since scheduling gives much high efficiency of loading by avoiding random transmissions colliding. This does require a more complex system and terminal design, but still one significantly simpler than even 2G cellular systems.
Spreading and heading
These design decisions have subsequent impacts. One of the most far-reaching is the use of spreading. Spreading extends the duration of messages. Any frame header information transmitted in a cell must be at the highest spreading factor supported in the network to ensure that all terminals are able to receive it.
Although every attempt has been made to minimise header information, it cannot be removed completely. The minimum size of the header information times the symbol rate times the maximum spreading factor dictates the time spent at the start of each frame transmitting header information. This works out at around 100ms. In order to keep the overhead of the header information to below 10 per cent this implies that the frame duration should be of the order 1s or more.
In Weightless the duration can be set as a variable within the network but a length of 1-2s is recommended. This is much longer than the frame duration in most wireless systems, hence Weightless can be considered to have a long frame duration.
Frame length factors
The long frame duration has implications. One is that the minimum round-trip delay is about the frame length – of the order 2s – at best case and twice this at worst case. This would be disastrous for voice calls (or even for Internet browsing), but is typically not a problem for machines.
The second is that it allows a different base station implementation where most of the processing is removed to the core network. There is ample time for the core to prepare a complete frame, and send it to the base station for conversion to RF and transmission.
This enables low-cost base stations, a simple upgrade path, and more intelligent scheduling decisions across the network. Power management follows on from this (see boxout, ‘The art of napping terminals’).
A set of implications flows from the requirements for a long battery life. This implies terminals that want to save energy are able to enter into a sleep mode. However, too long a sleep mode would compromise the ability to contact them unexpectedly (e.g., with an alert message) or increase the probability that network such as updated frequency assignments would take place while asleep. For Weightless, calculations suggest that a sleep time of around 15 minutes would result in a battery drain sufficiently small that battery life is constrained more by the shelf life of the battery than the current consumption.
Weightless has therefore been designed with the idea of a super-frame that repeats at around 15 minute intervals. The start of a super-frame is a point where all terminals are expected to wake up and listen and hence it can be used to alert them to network changes and send other relevant control information.
Low battery drain is only achieved if terminals listen for the minimum amount of time, then revert to sleep mode. To achieve this all the information needed by a terminal is contained within the header of each frame. Hence, any terminal need only listen for about 100ms and if there is no information destined for it, can then return to sleep.
This requires careful header design to avoid the terminal having to listen to subsequent frames to obtain a complete set of information. For example, it implies that the hopping sequence cannot be communicated by listing in the header the next frequency to be used and requiring the terminal to listen to sequential frames until the pattern repeats. Instead, the entire pattern, albeit efficiently encoded, must be transmitted in each frame.
The need to be able to handle sudden peaks in traffic due to some event such as a power failure stimulating multiple devices requires careful control of the uplink resource. Mechanisms to forestall devices sending alerts once the error condition has been noted by the network are also needed.
Mobility support requires terminals to be able to move from cell to cell. In cellular systems the network controls handover based on measurement reports provided by terminals. However, this generates substantial network traffic in terms of measurements and imposes a heavy battery load on the terminals when monitoring adjacent cells.
Because machines do not need seamless handover a much simpler approach is adopted in Weightless. Handover is almost entirely driven by terminals (there are exceptions to this).
Once a terminal detects it has moved out of coverage of a cell it re-starts its acquisition process and attaches to a new cell providing coverage. This means there is little need for any signalling traffic either from the terminal or the network which dramatically improves network efficiency; hence, handover is terminal-driven.
The need to achieve stringent adjacent channel emissions has an impact on the modulation approach used. Tightly filtering orthogonal frequency division multiplexing (OFDM) transmissions tends to distort the waveform more than the same degree of filtering on single carrier modulation due to the higher peak-to-average power ratio requirements of OFDM. Hence single carrier modulation is preferred for white space operation. Weightless uses single carrier modulation but benefits from the frequency domain equalisation possible in OFDM by using single carrier frequency domain equalisation (FDE) where a cyclic prefix is inserted as in OFDM and then used to determine the channel frequency response.
Lastly, the need to handle a very large number of devices requires considerable intelligence in the network to schedule communications and adapt network parameters according to load. The loading problem is exacerbated by the varying nature of the frequency resource available with white space channel availability changing and interference potentially occurring randomly from other white space users. This, then, sets the key parameters of Weightless as a TDD system with single carrier modulation, direct sequence spreading, broadband downlink and narrowband uplink, long frame duration, frequency hopping at the frame rate and 15 minute sleep cycle capability.
All of this adds up to a wireless solution custom-designed for machine communications in the emerging white-space radio spectrum. Maybe this will be the last piece in the puzzle needed to realise the visions of billions of connected devices making our environment a better place to be.