This invention relates generally to novel relay designs to increase coherence bandwidth of wireless channels with Finite Impulse Response (FIR) filters in wireless systems.
With the proliferation of mobile applications, there is an increasing demand for higher throughput of wireless systems at a staggering pace. Given the fact that the limited spectrum under 6-GHz is already crowded, millimeter Wave (mmWave) has emerged as a promising technology of future Fifth-Generation (5G) wireless systems [1].
Properly designed repeaters can play an important role in wireless systems. In Wireless Fidelity (WiFi) or Long Term Evolution (LTE) systems, repeaters are used to extend the coverage range. For mmWave, the role of repeaters is fundamentally different. Given the strong radio propagation directivity and large reflection loss of mmWave signals, repeaters are essential for seamless coverage [2]. Note that repeaters can be divided into two categories: Amplify-and-Forward Repeater (AFR) and Decode-and-Forward Repeater (DFR). Since the DFR introduces considerable propagation delay especially for multi-hop repeater scenario, AFR enhanced wireless systems offer advantages over DFRs. The energy efficiency of repeaters was studied in [3]. The problem of minimizing the number of repeaters and maximizing network utilities was studied in [4]. In [5], an iterative algorithm is developed for jointly designing the Receive/Transmit (Rx/Tx) Radio Frequency (RF)/baseband processors. It was demonstrated that multi-hop repeaters can greatly improve the connectivity versus single hop mmWave transmission in [6].
There have been previous inventions on utilizing repeaters in wireless systems. However, little attention has been paid to the impact of repeaters on wireless channel coherence bandwidth. Coherence bandwidth means all the sub-carriers within it share the similar channel characteristics, so that channel estimation only needs to be performed once for all the sub-carriers. In [7], it is demonstrated that multi-hop repeaters will make the channel coherence bandwidth narrower, but no methods were proposed to combat this issue of narrowed coherence bandwidth. Narrower coherence bandwidth means more resources (e.g., pilot spectrum and computation) have to be spent on channel estimation.
One embodiment of this invention is an innovative repeater design with an FIR-based channel equalizer to increase the coherence bandwidth of wireless systems. Based on the fact that the channels between repeaters are slow varying due to the repeaters being static (i.e., not moving), and thus have long coherence time, each repeater can adaptively equalize the channel using recently obtained channel estimates. With the equalized channels by this novel design, the energy sensitive User Equipments (UEs) can spend much less resources on channel estimation. For example, in an mmWave system with 4 repeaters and 100 UEs, all UEs can save half the resources in channel estimation, if the 4 repeaters are equipped with the FIR filters proposed in this invention.
Another embodiment of this invention is to equalize channels with beamforming with multiple antennas at transmitter or receiver. If the number of transmitter antennas is Nt and the number of receiver antennas is Nr, the repeater would need Nt Nr FIR filters. To reduce the complexity especially for systems with large numbers of antennas, this invention describes a method to first perform transmitter or receiver beamforming and then equalize the channels at each receiver antennas. Then, only Nr FIR filters are needed for the repeater with Nr antennas.
Reference may now be made to the drawings wherein like numerals refer to like parts throughout. Exemplary embodiments of the invention may now be described. The exemplary embodiments are provided to illustrate aspects of the invention and should not be construed as limiting the scope of the invention. When the exemplary embodiments are described with reference to block diagrams or flowcharts, each block may represent a method step or an apparatus element for performing the method step. Depending upon the implementation, the corresponding apparatus element may be configured in hardware, software, firmware or combinations thereof.
The system model of the repeaters is shown in
Motivation of Equalization to Increasing Coherence Bandwidth: Channel frequency selectivity can be characterized by coherence bandwidth. Since sub-carriers within the coherence bandwidth have the similar channel, the UE or the BS only needs to estimate the channel once for all the sub-carriers within the coherence bandwidth. If the coherence bandwidth is wider, the system could spend less resource (e.g., pilot and computation) in channel estimation. As an example of LTE systems, there are totally 1200 sub-carriers. If the coherence bandwidth is 48 sub-carriers, the 1200 sub-carriers are divided into 25 groups, and each group only needs to conduct channel estimation once.
Assume that the transmitted signal is x(t), and the received signal y(t) through the wireless channel denoted by h(t) is then denoted as
y(t)=x(t)*h(t)=∫τ∞x(t−τ)h(τ)dτ, (1)
where * denotes convolution.
Let hk(t) denote the impulse response of the kth, k=1, 2, . . . K, hop, where the first hop begins from the BS, and the last hop ends at the UE. If there is no repeater during signal propagation, then K=1. Hence, the received signal through K hop of repeaters is given as
y(t)=x(t)*h1(t)*h2(t)* . . . *hK(t)=x(t)*hr(t), (2)
where hr(t)=Πk=1Khk(t) denotes the overall channel impulse response through K−1 repeaters.
One embodiment of this invention is the repeater shown in
Assume that the impulse response of FIR filter on the ith repeater is wi(t), i=1, . . . , K−1, then the final received signal can be written as a discrete time form
y(m)=x(m)*Πi=1K−1[hi(m)*wi(m)]*hK(m), (3)
Where y(m)=y(mTs) with Ts being the sampling rate. Note that the destination of the last hop hK(m) is the receiver, so there is no corresponding equalizer.
Another embodiment of this invention is the method to calculate the values of wi (m) shown as follows. At the ith receiver, it estimates the channel hi, then calculates wi to equalize it. Let xiP(m) denotes the training pilot used to estimate hi The reason that the channel hi is frequency selective is that the received signal at the ith repeater y′i(m)=xip(m)*hi(m) consists some delayed replica of previous data xip(m−1), xip(m−2), . . . . If y(m) is not corrupted by previous data, then y′i(m)=xip(m)hi(m) and the channel is flat. Therefore, the problem is essentially designing a filter wi so that the output {circumflex over (x)}ip(m)=yi′(m)*wi(m) is close to xip(m). Without loss of generality, assume that the pilot signal xip(m) is the same for all the repeaters. For simplicity, let {circumflex over (x)}(m) and x(m) denote {circumflex over (x)}ip(m) and xip(m), respectively. Then, the goal of equalization is to choose wi to minimize E(|{circumflex over (x)}(m)×(m)|2)
Assume that wi has L taps, then,
{circumflex over (x)}(m)=
where
i(m)=[y′i(m−L+1),y′i(m−L+2), . . . , y′i(m)]T (5)
and
i(m)=[w*i(L−1),w*i(L−2), . . . , w*L(0)]T. (6)
To estimate the channel, the repeater does not need to decode the signals, but need to do analog-to-digital sampling. Like other RF components on the repeater (such as bandpass filter and amplifier), the FIR filter will introduce additional delay, but it is fixed and the maximum delay is the length of taps and can be designed to stay within the delay tolerance of the total channel, e.g., keeping the cyclic prefix under a maximum value. Therefore, the problem can be defined as the following:
minE(|{circumflex over (x)}(m)−x(m)|2)=min
Noticing the above is essentially a Minimum Mean Square Estimation (MMSE) problem, and the optimum solution satisfies that the estimation error
E{y′
i(m−1)[
with l=0, L−1. Then, the optimum solution is
i(m)=Cov[
where Cov denotes covariance.
Optionally, to guarantee that the input and output (of the FIR filter) signals have the same power,
With the knowledge of x(m), the repeaters can compute the optimum wi(m) based on the received signal y′i(m). Since the channel of each hop hi(m) is equalized, the overall channel of repeater hops h1(m)*h2(m)* . . . *hK−1(m) is equalized. Note that the repeaters can be trained based on the existing downlink/uplink pilot signals. Since the repeaters are static, the channel hi has a long coherence time and this training can be done much less frequently than the downlink/uplink channel estimation. In addition, the repeaters are always less energy-sensitive than UEs. With this novel design, the energy-sensitive UEs can spend much less resources for channel estimation.
Note that it is not required that every repeater has an FIR filter for equalization. For example, the system can use the kth repeater to equalize the channel from transmitter to it through the k−1 repeaters.
The following simulation results show impact of the FIR equalizer on coherence bandwidth and BER. In the simulation, the delay values of the N taps of channel responses are generated based on the Poisson distribution. As verified by the mmWave campaign that “The distribution of power among the path clusters is well modeled via a 3GPP model” [2], the power levels of the N taps of channel responses are generated based on the exponential distribution [8]. Assume that the sampling rate at receiver/repeaters is 3.072 GHz, and the channel bandwidth is 2 GHz divided into 1200 subcarriers. Note that the values are chosen to achieve the same ratio of sampling rate over bandwidth as LTE [8]. The maximum tap of an FIR filter is L=50, i.e., 16.3 ns.
Coherence Bandwidth:
BER:
If there are multiple transmitter or receiver antennas, each antenna on a repeater 4 will receive signals from multiple antennas on the transmitter 15, which could be a BS, a UE, or anther repeater, as shown in
If the number of transmitter antennas is larger than the number of receiver antennas (repeaters in the first layer, i.e., the BS or the UE, has more antennas than repeaters), the transmitter side beamforming is needed. If the number of receiver antennas is equal or larger than the number of transmitter antennas, the receiver side beamforming is required to separate the data streams. If the transmitter has more antennas than the receiver, the transmitter needs to know the channel which can be obtained through uplink channel estimation (based on channel reciprocity) or channel estimation feedback from receivers to transmitters. Otherwise, only the receivers need to know the channel to separate data stream, and the channel can he estimated by downlink pilot transmission.
As shown in
The precoding matrix is defined as x=Ps, where x=[x1, x2, . . . , xN
One procedure to compute the precoding matrix is described as follows. In the Time-Division Duplex (TDD) scenario, the Nr receiver antennas send pilot signals to the Nt transmitter antennas, then the transmitter estimates the downlink channels based on channel reciprocity. In the Frequency-Division Duplex (FDD) version, the Nt transmitter antennas send pilot signals, and the Nr receiver antennas estimate the channels and feed hack the channel estimates to the transmitter. Based on the channel estimation feedback, the optimum beamforming matrix can be computed, e.g., using ZF, MMSE, or other methods.
Then, the jth receiver antenna receives data through the equivalent channel
One embodiment of this invention is the transmitter or receiver beamforming algorithms to separate the data streams, so that the equalizer filter coefficients can be calculated for each data stream. One embodiment of this invention is that if the number of receiver antennas is equal or larger than the number of transmitter antennas, the receiver can separate the data streams through data processing such as ZF, MMSE, or other methods. In
One difference from the single pair of antennas system is that the jth receiver antenna might receive interference (transmitted data other than si). If the precoding matrix is not perfectly calculated, then, the estimation of
This section describes the procedure of the channel equalization with repeaters in wireless systems, which includes the FIR parameter estimation, channel feedback, and UE channel estimation.
As described in previous sections, it is important to estimate the channels between repeater antennas and transmitter antennas, so that the optimum weighting of FIR filters can be calculated. As there might be many repeaters on the same hop, we define the repeaters receiving signals yi (m) as the repeaters on the ith layer. Note that the channel estimation can be obtained either by direct downlink channel estimation (the repeater on the ith layer equalizes channels between the (i−1)th layer and the ith layer) or through channel feedback (the repeater on the ith layer equalizes channels between the ith layer and the (i+1)th layer). Note that the 0th layer is the BS for the downlink and the UE for the uplink.
One embodiment of this invention is that the repeaters on the same layer use orthogonal codes (such as m-sequence) or spatial division to avoid interferences to the repeaters in the next layer. Another embodiment of this invention is that these pilots are transmitted in the system Guard Period (GP) for a TDD LTE system. In a LTE system, there are some dedicated OFDM symbols reserved for pilot transmission which can be used for filter coefficients calculations. The upper layer controls the signal propagation process, and then each repeater knows the previous hop sources, and their pilot signals. With orthogonal pilot sequence, each repeater in the ith layer only receives the signal from the desired transmitter in the (i−1)th layer. In addition to the code division, spatial division can also be used to avoid interference. The transmitters on some layer that are sufficiently separated in distance can be scheduled to use the same pilots, e.g., using in the same frequency and/or code at the same time to avoid interference. The transmitters on the same layer can also use high-directional antennas (common for mmWave systems) to send signals to different receivers with sufficient angular separations to avoid interference.
If the direct channel equalization is used, the repeaters in the first layer equalize the channels between them and the BS in the downlink scenario. However, in the uplink scenario, the first layer repeaters equalize the channels between them and the UEs. The channels between repeaters and UEs always have less coherence time than the channels between repeaters. Therefore, it is desired to use the equalization through feedback in the uplink scenario, so that the repeaters in the first layer equalize the channels between the repeaters in the first layer and repeaters in the second layer, and the repeaters in the last layer equalize the channel between them and the BS. This method guarantees that all the repeaters equalize the channels with long coherence time, to reduce system resource on equalization. After equalization, the total channel between the BS and the UE are still not perfectly flat, because the equalization at repeaters is not perfect and the channels between the UE and repeaters are not equalized. One embodiment of this invention is that the BS or the UE uses OFDM or other methods to estimate channels. For example, the system has a bandwidth of 2 GHz, and the channel is not flat over the 2 GHz bandwidth. However, if the 2 GHz bandwidth is divided into W subcarriers based on the OFDM technique, then the channel can be considered to be flat for every w subcarriers. Hence, the UE or the BS can estimate the channel for each group of w subcarriers. In this way, the UE or the BS has a good channel estimation in the overall 2 GHz channel.
The repeaters can achieve equalization using either one of the following two embodiments: (1) Direct amplify-and-forward mode: The FIR filter is constructed with tap delay lines, and each tap has one or more adjustable attenuators and/or phase shifters with values set to match the values of
When all repeaters have obtained the optimum FIR settings, the communication between the BS and the UE is the same as without repeaters, since the repeaters operate in the amplify-and-forward mode. However, since the repeaters' inside paths might be asymmetric, special attention should be paid if the channel reciprocity is used to the downlink channel estimation. In summary, if the uplink and downlink channels of repeaters' inside symmetric, the FIR filters for the uplink and downlink have the same setting, then the overall channel from the BS to the UE is symmetric. If the uplink and downlink channels of repeaters' inside AFR paths are asymmetric, the channel estimation from the BS to the UE can be obtained through feedback.
Although the foregoing descriptions of the preferred embodiments of the present inventions have shown, described, or illustrated the fundamental novel features or principles of the inventions, it is understood that various omissions, substitutions, and changes in the form of the detail of the methods, elements or apparatuses as illustrated, as well as the uses thereof, may be made by those skilled in the art without departing from the spirit of the present inventions. Hence, the scope of the present inventions should not be limited to the foregoing descriptions. Rather, the principles of the inventions may be applied to a wide range of methods, systems, and apparatuses, to achieve the advantages described herein and to achieve other advantages or to satisfy other objectives as well.
This application claims the benefit of U.S. Provisional Application No. 62/157,471, filed on May 6, 2015.
Number | Date | Country | |
---|---|---|---|
62157471 | May 2015 | US |
Number | Date | Country | |
---|---|---|---|
Parent | 15567471 | Oct 2017 | US |
Child | 17305865 | US |