Black Lattice has developed the world's first high performance, low cost wireless MIMO detector using machine learning techniques. This groundbreaking new technology solves one of the hardest parts of a modern baseband modem chip, the multi-antenna processing. This approach will help alleviate the wireless congestion created by the spectrum crisis and provide better performing and higher reliability devices to consumers.
The Wireless Spectrum Crisis
The wireless industry has a dirty secret. We all rely on the free flow of data on the wireless networks to supply our connected lives, but there's a problem. The frequency bands, or spectrum, that our data travels through over the air to our devices is a scarce resource and in many areas has been exhausted, creating an invisible traffic jam all around us.
Each frequency band is like a lane on a road and each packet of data is like a car. The major LTE and WiFi improvements over the years have managed to pack more data onto the spectrum more efficiently, just like packing more cars closer and closer together on the highway, but sooner or later there's no more room between the data before collisions occur.
The newest 5G standard has a few technologies that are supposed to help including massive-MIMO (hundreds of antennas at base station) and millimeter wave (super high frequency bands that behave more like light than radio waves), but these are expensive and limited options. Millimeter wave in particular can't go around corners, pass through walls, and even struggles to pass through a single tree leaf. Both of these options will help in special usage cases, but they clearly aren't the panacea they're sometimes claimed to be.
Spatial-multiplexing multi-input multi-output (MIMO) antennas are able to dramatically increase data speeds and spectrum efficiency by sending a new stream of data on each additional antenna. Using the car and road analogy, using MIMO is like stacking cars on top of each other. The trick is that it's extremely hard to do without consuming a lot of power and damaging the data in the process.
MIMO technology has been in the LTE and WiFi protocols since 2008, but the industry has struggled to perfect the technology, only getting a fraction of the possible gain with high cost and battery consumption.
The huge computational complexity of MIMO is a perfect fit for modern machine learning techniques, but is extremely hard to get right. The trick is to forget everything we think they know about the problem and start from scratch. This approach has allowed our patent pending techniques to achieve unprecedented maximum MIMO performance without sacrificing chip cost or battery life.
Since our techniques are purely digital and improve on existing features of the wireless protocols, we can easily integrate our design into new smartphone modem chips, upgrade existing cellular basestations, and enable new 5G smallcells.