Transmit Beamforming with Implicit FeedbackBy CWNP On 09/19/2007 - 8 Comments
- Introduction -
Transmit Beamforming (TxBF) is a method of using a set of separate antennas as a virtual array to form high-gain beams focused at client stations. It is important that each transmitter (called the beamformer) understand the characteristics of the MIMO channel between itself and the receiver (called the beamformee) since transmissions will be tightly focused on a specific area.
- Channel Sounding -
There are several cases where it is desirable to obtain as full a characterization of the MIMO channel as is possible, thus requiring the transmission of a sufficient number of High Throughput Long Training Fields (HT-LTFs) to sound the full dimensionality of the channel. These cases of MIMO channel measurement are referred to as MIMO channel sounding. A sounding frame may be used to sound available channel dimensions. A sounding request is like saying, "where are you?" while a sounding response (PPDU) is like saying, "over here!" By listening to sounding PPDUs, a receiver can understand the MIMO channel characteristics, calculate a steering matrix, and is able to "steer" the next frame toward the receiving station in the optimum manner. A sounding PPDU is identified by setting the "Not Sounding" field in the High Throughput Signal field of the PHY header (HT-SIG) to zero. A sounding PPDU has a variable number of HT-LTFs.
In both HT mixed format and HT greenfield format frames, there are two types of HT-LTFs: Data HT-LTFs (DLTFs) and Extension HT-LTFs (ELTFs). DLTFs are always included in HT frames to provide the necessary reference for the receiver to form a channel estimate that allows it to demodulate the data portion of the frame. The number of DLTFs may be either 1, 2, or 4, and is determined by the number of space time streams being transmitted in the frame. ELTFs provide additional reference in sounding PPDUs so that the receiver can form an estimate of additional dimensions of the channel beyond those that are used by the data portion of the frame. The number of ELTFs may be either 0, 1, 2, or 4. PLCP preambles in which DLTFs are followed by ELTFs are referred to as staggered preambles.
A station may also sound the channel using a Null Data Packet (NDP) (which is indicated by zero in the Length field in the HT-SIG) with the "Not Sounding" field set to 0. The last HT-LTF of an NDP is not followed by a Data field, and it is optional for a station to process an NDP.
- An Analogy -
To help you understand the concept of channel sounding, first picture two people (Jim and Sarah) at opposite ends of a large room that is filled with obstacles. The room is completely dark, but Jim and Sarah each have an omni-directional flashlight. Jim and Sarah both want to throw a rubber ball to the other person without it hitting obstacles, but if it does hit obstacles it should be reflected directly at the other person. In order to successfully get the ball to the other person, there are a couple of strategies we can take. We can have Jim pour a glowing chemical (we'll call it glow-chem) on the ball he throws (sounding frames carrying HT-LTFs and/or NDPs) to illuminate the room as the ball travels. This ball might have problems, but in the worst case it illuminates the room a bit. Jim could pour lots of glow-chem on the ball to brighten up the room further (extension HT-LTFs). Jim could turn on his flashlight for a short period of time and indicate that Sarah should do the same in a back-and-forth fashion (calibration). Jim could can also indicate to Sarah that she should pour glow-chem (a normal amount or really soak it) on her ball and throw it. Jim and Sarah could pour glow-chem on multiple cheap balls and throw them all over the place to light up the room well enough to see, and then throw their really nice rubber ball directly at the other person while they can still see (NDPs).
All of these mechanisms can be used in an 802.11n environment where TxBF and Channel Sounding are used.
- Feedback -
The beamformer and beamformee can work together to educate each other on the characteristics of the MIMO channel. To do this, they use either implicit or explicit feedback. In this article, we're only going to address Implicit feedback. Implicit feedback can be both unidirectional or bidirectional. Bidirectional implicit TxBF is where both nodes are playing the roles of beamformee and beamformer. Transmissions between the two nodes use long training symbols (in the PHY layer header) to help the receiving side to understand the RF environment enough to make an estimate suitable for computing a transmit steering matrix (a set of antenna parameters).
Stations must announce that they are Implicit Feedback TxBF Beamformee and/or Beamformer capable in the Tx Beamforming Capabilities field of the HT Capabilities information element.
- Definitions -
Beamformee: A receiver of transmit beamformed PPDUs. Must transmit sounding PPDUs as a response to TRQ=1 (Training Request - also called a Sounding Request).
Beamformer: A transmitter of beamformed PPDUs. Can receive sounding PPDUs and can compute steering matrices from MIMO channel estimates obtained from long training symbols in sounding PPDUs received from the beamformee.
- Calibration -
The 802.11n TxBF calibration procedure provides the means for calculating a set of correction matrices that can be applied at the transmit side of a station to correct the amplitude and phase differences between the transmit and receive chains in the station. If this is done in both the stations that are communicating with each other, reciprocity is restored in the baseband-to-baseband response of the forward and reverse channels. Differences in the amplitude and phase characteristics of the transmit and receive chains associated with individual antennas degrade the reciprocity of the channel and cause performance degradation of implicit beamforming techniques.
If the channel is reciprocal, the beamformer can use the training symbols that it receives from the beamformee
to make a channel estimate suitable for computing the transmit steering matrix. Generally, reciprocity requires
calibrated radios in MIMO systems, hence the need for a calibration procedure.
A responder in a calibration exchange: Can receive and transmit sounding PPDUs. Can respond with a MIMO Channel State Information (CSI) Matrices frame that contains channel measurement information obtained during reception of a sounding PPDU.
An initiator in a calibration exchange: Can receive and transmit sounding PPDUs. Can receive a MIMO CSI Matrices frame sent by a calibration responder.
A station that sets the Implicit TxBF Capable subfield of the TxBF Capability field to 1 must support calibration
and must set the Calibration subfield of the TxBF Capability field to 3 (indicating full support of calibration)
in all HT Capabilities elements that it transmits. A station is only allowed to initiate a calibration training frame exchange sequence with another station that supports calibration.
The 802.11n TxBF calibration procedure consists of 4 frames.
Calibration Start frame / Position 1 / sent by calibration initiator / TRQ=1: This Sounding PPDU (A QoS Null Data frame) initiates the calibration procedure and is followed by an ACK frame.
Calibration Sounding Response frame / Position 2 / sent by calibration responder: This Sounding PPDU (an ACK+HTC frame) is sent a SIFS after receipt of the Calibration Start frame. This frame is used by the calibration initiator to estimate the MIMO channel.
Calibration Sounding Complete frame / Position 3 / sent by calibration initiator: This Sounding PPDU is sent a SIFS after receipt of the Calibration Sounding Response frame. This frame is a QoS Null+HTC frame, has the CSI/Steering subfield of the HT Control field set to 1, is used by the calibration responder to estimate the MIMO channel, and will be followed by a normal ACK frame (the 4th frame in the calibration sequence).
Each calibration sequence has a unique identifier found in the Calibration Sequence subfield in the HT Control field. This identifier remains the same throughout each calibration sequence and is incremented each time a new calibration procedure is started.