Meeting requirements for high-performance mesh networks in compact designs
WiFi mesh networks are gaining momentum in the consumer market. The reason? Users increasingly demand whole-home coverage. There are a plethora of vendors and products to choose from. Big names like Google are offering solutions for the retail market. Internet service providers are taking notice and evaluating their own offerings.
For a WiFi mesh to be effective it has to pack at least two 5GHz radios in each node: one radio to connect clients distributed throughout the house or the room; and another 5GHz radio to connect back to the main unit at the cable or DSL access point. Some vendors are offering low-performance solutions that use a single 5GHz radio for both access and backhaul, but this comes with a severe throughput penalty, since the radio can only provide access or backhaul at any given time, the user throughput is degraded by 50%.
In addition to counting radios, it is also important to count antennas. A 4x4 MIMO radio (4 receivers and 4 transmitters) would typically outperform a 2x2 MIMO radio. Both solutions exist in the marketplace. A 4x4 MIMO radio requires four antennas and a 2x2 MIMO radio requires only two antennas. For the best performance, a WiFi mesh node would require eight 5GHz antennas. In addition, there are chip vendors who are promoting new 8x8 designs that would require 16 antennas for two radios!
Since mesh units are distributed throughout the home -- as opposed to the old-style router that was tucked away in the already cluttered office -- they must be aesthetically appealing. They can’t be enormous black boxes – they must be small and elegantly designed. But the size of the device impacts performance. Placing the antennas too close together makes it hard to provide the independent spatial streams that are the key benefit of MIMO (when the antennas are for the same radio) and increases interference between the radios (when the antennas are for separate radios).
Actually, it is impossible to place the antennas for two separate radios operating in the same frequency band inside a single access point without the radios negatively affecting each other’s performance. To achieve the additional isolation between the 5GHz radios necessary for them to operate side-by-side, RF filters are required. One radio is configured to operate in the lower part of the UNII 5GHz band with filters on each of its antennas to reject noise coming from the upper part of the band. And the other 5GHz radio is configured to operate in the upper part of the band with filters on each of its antennas to reject noise coming from the lower part of the band. Eight antennas require eight filters – four low-pass filters and four high-pass filters.
These issues create a dilemma for providers of whole-home Wi-Fi solutions. A large and unattractive product is unacceptable to consumers. And small, aesthetically appealing, but low-performing WiFi mesh nodes cannot deliver the necessary throughput and capacity for the modern, connected home.
What’s needed is a way to share one set of antennas between two 5GHz Wi-Fi radios. This eliminates half of the 5GHz antennas otherwise required for maximum performance. But in order to share antennas between two separate radios a special-purpose filter called a diplexer is required. A diplexer is essentially a combination of high-pass and low-pass filters. Now, let’s assume the number of filters required is the same as when using dedicated antennas for each radio (no sharing). When the antennas are physically separate, the standalone filters only need to reject about 35dB of noise, whereas the filters in the diplexer need to reject about 70dB of noise. That’s a big difference. Cascading two of the 35dB filters would double the filters’ already high insertion loss to an unacceptable level. A filter with 70dB rejection and low insertion-loss to make sure that energy is not wasted in the filtering process would be physically large, heavy, and expensive. These are the kind of filters traditionally used in higher power 4G/LTE mobile base stations. As such they are impractical for consumer WiFi access points.
This is where Self-Interference Cancellation (SIC) technology can assist. SIC was originally developed to allow a radio to transmit and receive at the exact same time on the exact same frequency. This is accomplished by cancelling out its own transmit signal as it appears at the receiver’s input. Unlike radio filters, which are inserted in the signal path to block unwanted frequencies (inevitably creating losses to wanted signals), SIC creates a new signal -- a cancellation signal -- that is added at the receiver’s input to surgically remove just the unwanted signal. A very high-resolution cancellation signal can be produced by directly sampling the transmitter’s output and continuously monitoring and making adjustments for reflections from the environment. For many applications, the additional power required to produce the cancellation signal is worth the extra cost, if any.
It turns out, however, that SIC has other uses, such as enabling two 5GHz radios in the same box, sharing the same antennas, to operate simultaneously without degrading each other’s performance.
The spectrum analyzer trace in Figure 1 illustrates the effectiveness of self-interference cancellation technology. To operate a second radio on a nearby frequency, it’s necessary to reduce in power (at the input of the receiver) both the transmit signal on the transmit frequency (to the point labeled “3”) and the noise produced by the transmitter on adjacent frequencies (to the point labeled “4”). Notice that after cancellation the noise from the interfering transmitter has been reduced to the noise floor.
Figure 1 - SIC technology can be employed to significantly reduce adjacent channel interference (Source: Kumu Networks)