As a performance enhancement technology for a physical layer, MIMO has been adopted in IEEE 802.11n, and the standard requires at least two antennas for MIMO operations. This, however, is equivalent to have at least two radio chains, which are in fact the most power-consuming component in a wireless device.
To avoid such an energy issue, an awkward approach is being taken: IEEE 802.11n with single antenna. That will have full MAC layer performance enhancement and part of the physical layer specified by IEEE 802.11n such as 40 MHz channel and higher coding rate, only lacking MIMO part. Broadcom announced this approach at the end of last year:
Excerpted from NETWORKWORLD
...
The 65-nanometer chip supports 11n in both the 2.4GHz and 5GHz bands, and can run simultaneously in both. Given the constraints of handset size and power, the chip supports only one 11n data stream and one antenna, compared with notebooks or access points that typically support two or possibly three streams, each with a corresponding antenna. The chip's Wi-Fi performance maxes out at 50Mbps of wireless throughput, according to the vendor. That's still nearly twice the throughput of 11a or 11g Wi-Fi.
...
Similarly, Marvell, whose has the only 802.11n products with three spatial streams by three antennas, also announced its new chip-sets that do the same as Broadcom in Jan, 2009:
Excerpted from PCMAG.COM
..
Until now, mobile phones have been stuck at slower 802.11g speeds because 802.11n required chips and antennas that are too large to fit into a mobile-phone form factor. The 1x1 single-stream 802.11n solution can take 802.11n down to a chip as small as 50 mm square, which could easily fit into a phone. That could boost cell-phone Wi-Fi speeds by several times. The 1x1 802.11n chips are available to manufacturers today, he said.
...
This is quite against the communication theory that tells us that even if the total transmit power is fixed, MIMO can give you multiple times of throughput performance. It sounds to me that manufacturing IEEE 802.11n chipsets somehow imposes artificial constraints to the design itself. And this point is where the protocol research should be worked.
