MIMO communication, the use of multiple antennas transmitting or receiving on the same frequency, is one of the most surprising technologies in my lifetime.  I first encountered it in MATLAB simulations for my master's degree program at USF.  The amazing part came five years ago when I experimented with my first inexpensive 802.11(n) and got it, at least under ideal conditions, to transmit and receive separate streams on the same frequency at the same time.


In a paper in the latest issue of IEEE Communications, researchers show how MIMO may be particularly beneficial on bands above 10 GHz because on these bands there is more need for antenna gain and antennas are smaller.


The International Telecommuncations Union (ITU) allocated the the E-band, 71 – 76 GHz and 81 – 86 GHz, over 30 years ago, but it has not be used until recently.  A band at these high frequencies opens some exciting possibilities.  It's 10 GHz of spectrum, more than 50 times the current mobile phone spectrum. 


The downsides of such high frequencies are increased path loss and attenuation from objects, even rain.    On top of this, phase noise is a bigger problem.  As a result, current E-band systems use low-order modules schemes, i.e. modulations that send one or two bits per symbol.  High-order modulation schemes send several bytes per symbol but require a higher signal-to-noise ratio (SNR).  So although the E-band represents a huge amount of spectrum, the data rate per MHz of spectrum is lower than it is on lower frequency bands. 


The short wavelengths, however, offer the advantage of small antenna size.  Many high-gain antennas can be mounted in a small enclosure and used for MIMO with a large number of streams.  The trouble with this is multi-stream MIMO depends on reflected signals, while the E-Band lends itself to direct line of sight (LOS) links.  So while the short wavelengths allow many antennas in a small space, you cannot get as many streams as you would on conventional microwave bands. 


EbandMIMO.jpgResearchers propose using large arrays of antennas but limiting the number of streams to four (4) and using the MIMO array for beamforming.  Beamforming provides gain by directing signals similar to a directional antenna.  They show that using the array for beamforming and multiple streams can theoretically make up for the drawbacks of the higher frequency. 


I asked one of the researchers, Dr. Sayeed, if they've done any of this testing with actual hardware.  They have a 10 GHz prototype and plans to build another prototype for 30 GHz or higher. 


The researchers suggest this technology can be used for backhaul to connect mobile base stations.  My first thought was base station operators could accept the larger antennas associated with MIMO arrays on lower frequencies.  Dr. Sayeed says small antenna size is always welcome, especially in pico cell bases. 


The most surprising comment from Dr. Sayeed was his opinion that eventually handheld mobile devices will use arrays of MIMO antenna, antennas that could only fit in a handheld device at these high frequencies. 



Since I first started working with digital communications, there has always been a tradeoff between SNR and data rate.  If you have a strong signal, you can use higher order modulation schemes that send multiple bits per symbol.  As SNR decreases, a system needs to fallback to lower datarates, which work at lower SNRs and are more tolerant of non-linearity in transmit amplifiers.  Starting five years ago, when MIMO became available on inexpensive WiFi cards, number of streams became another parameter for a fallback algorithm to work out. 


Dr. Sayeed suggests that higher frequencies will make MIMO practical (because of small antenna size) and necessary (because of the need for a high SNR) on handheld devices operating over 10 GHz.  This is hard to imagine in a device moving around in someone's hand. 


Using arrays for backhaul in fixed base stations is something researchers are already prototyping.  If it follows the same path as 802.11(n), pico-cell base stations with 70 GHz MIMO backhaul links will be cheaply available in ten year's time.