Is MIMO just another fad waiting to be placed on the much-hyped ‘nearly’ shelf and forgotten about, or will it provide the extra wireless capacity boost needed by the industry. Viktor Bale, consultant at PA Consulting Group, examines the technology’s potential
Anyone who takes at least a mild interest in wireless communications will have heard the phrase multiple-input multiple-output (MIMO) systems being banded around for at least a couple of years in industry. And he or she will have possibly raised a question regarding the pronunciation of this most-popular of acronyms.
In the past “mee-moe” seems to have been the pronunciation of choice, although the alternative “my-moe” is being heard increasingly often. A possible reason for this may be as the former sounds as if it should be a character in a Disney movie rather than an advanced communications structure. But certainly the more important debate relates to the questions: what is MIMO, is there a “true MIMO”, does it deserve all its hype, and, let’s face it, will it make anyone any money?
MIMO is a acronym that increasingly seems to be thrown around rather recklessly, slapped on almost anything that can be passed off as having at least superficially MIMO properties, as people cotton on to the marketing potential of this new buzz word. But is it just that — a fad, here today, written off to the same place in the sky that ERMES, the pan-European radio messaging system, TFTS (terrestrial flight telephone system), to some extent TETRA (digital PMR trunked radio) have gone tomorrow?
So what exactly is MIMO? From one perspective is could mean any system that has more than one input and more than one output. This could of course mean anything to anyone, so let’s constrain the meaning to be purely technical. One kind of MIMO system is found in the area of acoustics. Stereophonic echo cancellation has long involved an element of MIMO systems, as has 3D sound reproduction. Here loudspeakers are the inputs (to the channel), your ears the outputs and the channel is the air in between. This system potentially means that each of your ears could hear something completely different. Indeed much of the MIMO research performed for acoustic applications is directly applicable to wireless communications, on which this article now concentrates. In this application, the inputs are the multiple co-located radio transmitters, the outputs are the multiple receivers and in between there is a channel with some response.
Even in this form the MIMO concept has existed for decades in some form, e.g. by using two transmitters with perpendicular polarisation to send two independent signals and trying to recover these signals at two or more receivers. However, the recent explosion of interest started about ten years ago with the publication of some key information-theoretical work.
For most of this time, interest has mainly been confined to academia, but is now slowly filtering through to industry as it recognises there is some truth in the apparently wild claims made about MIMO systems. These are that, in fact, wireless systems are no longer limited by the Shannon capacity (the holy grail of communications — theoretically the most data you can reliably transmit per second in time and per hertz in frequency). By correctly exploiting the spatial characteristic of the channels between each pair of transmitters and receivers, the capacity, and potentially the achievable data throughput, scales almost linearly with the number of transmitters (at least in theory). Each transmitter can then send a data stream that is completely independent of any other. If we then have at least as many receivers as transmitters we can recover the data streams through sophisticated signal processing techniques at the receiver. In this way we use the extra capacity to achieve so-called multiplexing gain. Critically this can be done with no need for additional bandwidth, time allocation, codes (in the case of CDMA) or transmit power. You could argue then that it is extra capacity for nothing — although in reality it is paid for by the need for extra processing power at the receiver to implement the required advanced signal processing.
An alternative use for the capacity is to implement stronger coding. If we use the same logical end-to-end data rate as previously then we may use the extra capacity to implement space-time codes, which take advantage of the spatial diversity introduced by the channel to implement lower error rates than before. The explosion of interest in space-time coding started around the same time as the MIMO system, and similarly has until recently been mainly confined to academia. We may also use a combination of this coding gain and multiplexing gain but this, of course, adds an extra layer of complexity in development.
Hijacked by marketing
There seems to be some question about what MIMO is and what it is not. Since this acronym is currently receiving so much attention, it is inevitable that some people are marketing their products as being “MIMO enabled” when in absence of this popularity they would probably not be considered so. As stated previously, from one point of view, a MIMO system is anything that has multiple-inputs and multiple-outputs, and by this definition the implementation of beam-forming at the transmitter and receiver diversity at the receiver would superficially appear to form a MIMO system. However, this meaning was not the generally accepted one when MIMO systems were being conceived and researched in academia, and it should be argued that the meaning should not be bent by industry to use it as a marketing tool.
In wireless communications it is generally accepted that a MIMO system is one which benefits from additional capacity over that stated by the Shannon capacity. Beam-forming and receive diversity does not appear to fall into this category since this consist of two largely independent techniques that have been, for decades, glued together. The beam-former electronically steers the energy of the signal in the direction of the receivers and the receiver diversity is more effective at capturing more of this energy than a single receiver — but the system is still limited by the Shannon capacity.
This does not mean to belittle this alternative method — indeed we shall see later that there are times when this will perform better than a “true” MIMO system. And so we have introduced the term “true” MIMO, which in this article at least is one were the multiple transmitters and receivers are essential in working together to create extra capacity.
There are attempts to incorporate MIMO techniques into emerging standards but these seems to be plagued with confusion as no-one can decide on what form of MIMO to implement. This may be caused by the confusion surrounding the definition of MIMO and what constitutes “true” MIMO. Hopefully this article will serve to clarify some of the issues.
WiBro for example is an emerging Korean standard, in many ways similar to WiMAX, being standardised by their Telecommunications Technology Association. But short of stating an intention to incorporate some form of MIMO, information in this area seems to be sparse (or written in Korean!). Further attempts are being made to incorporate MIMO systems into the next WiFi standard (802.11n) by the two groups involved — TGn Sync and WWiSE — although each has chosen to concentrate of a different “flavour” of MIMO.
WWiSE has chosen space-time block coding (STBC) which is a subset of space-time coding and therefore aims to maximise the reliability of transmission in terms of error rate. However, it is currently only using up to a four-by-one antenna configuration (meaning four transmitters and four receivers), technically making it a multiple-input single-output system.
TGn Sync on the other hand is developing the beam-forming approach which steers the energy in the direction of the relevant user, and hence delivers to it more power, however as mentioned previously it would not appear that this approach benefits from any additional capacity over the Shannon capacity. Finally, WiMAX (based on 802.16d- 256FFT OFDM) has also specified a basic STBC option, though only using a two-by-one configuration.
Potential of “true” MIMO
One of the great advantages of MIMO is that it works at its best in environments that have been traditionally thought of as the most challenging for wireless communication. They rely on a rich scattering environment, possibly with significant multipath components to work to their full potential, which is fortunate since the high dates rates that they can enable are most likely to be needed in dense urban and in-building environments, i.e. those that cause the required channel conditions.
Conversely, in environments that are generally considered to be easiest to transmit data reliably — for example in a rural location with a strong line-of-sight component — MIMO systems to do not work well at all. In fact, in this case, the system would perform little better than a single-transmitter single-receiver system. Hence we would be better off using the beam-former approach, as introduced earlier, although this would not achieve a the capacity of a “true” MIMO system. We are therefore presented with the interesting possibility that a mobile multi-antenna system could automatically re-configure itself to a transmission technique which most suits its current environment.
Although the “true” MIMO system works best in a rich-scattering environment, this is in fact a double-edged sword, since it means that complex and highly processor intensive signal processing will be required at the receiver to equalise (or undo the distortive effect of) the channel. As the convolutive effect of the channel increases, the processing power required to recover the transmitted signals also increases usually at a greater rate than the channel response lengthens. This can cause significant problems as powerful signal processors would be needed which of course increases cost and decreases battery life in a mobile device. Fortunately some techniques do exist to lower processing power required.
For extremely distortive channels — e.g. in-building system operating over a very wideband single carrier — a technique known as subband processing has shown some potential. Subband systems can trace their origin to acoustics, were sound channels with extremely long impulse responses (relative to radio channels) are common. Another promising more well-known technique is the combination of a OFDM and MIMO approach. OFDM is a multi-carrier technique whereby the data is used to form the coefficients of the carrier frequencies which are then transformed into the time-domain through the use of an inverse fast Fourier transform (FFT). Hence a fade at any particular frequency will only affect the data symbol corresponding to that frequency. In practice this means that since the data is transmitted over a number of orthogonal narrowband frequency carriers, frequency-selective channel fading no longer causes a great problem and there is no need for an equaliser at the receiver.
When OFDM is combined with MIMO to create MIMO-OFDM the extraction of the transmitted signal becomes quite a simple task; we simply invert a number of narrowband MIMO channels equal to the number of carriers used in OFDM. In practice this is done by simple inversion of scalar-valued matrices. For example, a four-by-four MIMO system using OFDM with 256 carriers simply needs to invert 256 four-by-four matrices, which is not a particularly taxing computational task.
In a static channel this need be done only once, but in a fading channel the inversion should be performed at a rate determined by the fading rate. In reality once the error rate has degraded to above a certain threshold the channel should be re-estimated and re-inverted.
How much then?
A key question is that of cost; great as the potential benefits of MIMO technology are, if it costs too much to implement it is as good as dead in the water. The cost involved in additional computational requirements has already been mentioned but how big is this and what other costs could one expect to incur when implementing a MIMO system on a large scale? Unfortunately there is no simple answer to this question as it depends on which form of MIMO you are talking about.
One cost common to all forms is the need for multiple RF front-ends at both transmitter and receiver. Going any further down the chain is where things start to vary. If implementing a MIMO system for multiplexing gain using only a single carrier, nothing more is needed at the transmitters — the signal needs only be demultiplexed across the multiple transmitters. The receiver however will need a reasonably capable DSP to perform the signal processing tasks involved in recovering the transmitted signals and multiplexing them back into a single data stream. Multi-carrier MIMO-ODFM will also require the FFT stage core to OFDM to be multiplied up, both at the transmitter and receiver, to benefit from the simplicity of inverting multiple narrowband MIMO channels. Implementing space-time coding will require still further the multiplying-up of components in the signal chains.
However all these changes are confined to the PHY stage, hence for example, to implement a two-by-two MIMO system the cost increase from the PHY stage will at most double, but everything higher up the communications stack could remain largely unchanged.
So its seems there is a great deal of uncertainty at present as just what MIMO means as companies vie to make it into what suits them, and the definition of MIMO (as well as the pronunciation) will vary depending on who you ask. But there is one thing that is certain — with all the current interest in and promised benefits of MIMO systems it would seem that any forward-thinking communications company or department should invest at least some of its resources into “finding MIMO” (pronounced “my-moe”).
