GB2603155A - Improvements to multi-carrier MIMO systems - Google Patents

Abstract

There is herein provided a method of processing a MIMO signal using multi-carrier modulation, such as OFDM. Channel state information (CSI) is measured for a communications channel in relation to a first tone, the first tone being a component tone of the MIMO signal. A second tone is then precoded using the measured CSI of the first tone. The second tone is also a component tone of the MIMO signal. Preferably, a precoding matrix is calculated from CSI for the first tone and the same precoding matrix is applied to both the first and second tones. The first and second tones may be adjacent tones. The re-use of CSI in this way reduces the number of channel measurements and processing computations required.

Description

Improvements to multi-carrier MIMO systems The present invention relates to MIMO signals using multi-carrier modulation, such as Orthogonal Frequency Division Multiplexing (OFDM) or Discrete Multi-Tone (DMT), e.g. in the transmission of wireless signals from a cellular base station to a user's mobile phone.

Channel conditions, especially for high frequency signals, vary over frequency and over time. This means that channel coefficients must be estimated (i.e. measured) over a range of frequencies and this must be done repeatedly. The estimated channel coefficients are then used to optimise transmission parameters, for example to precode MIMO signals prior to transmission in order to mitigate interference.

In a typical OFDM system, the transmitters each transmit 2' tones, each in a respective

BW

approx. -2N-wide sub-channel, where N is an integer number greater than 0 and BW is the system bandwidth. Consider an example system in which N = 11 and hence there are 2048 tones. In this example system, estimating the channel involves constructing 2048 versions of the channel matrix, each for a different one of the tones. The channel matrix contains coefficients which are a measure of the interference caused to each tone by the channel (which includes the signals transmitted from the other antennae). The more antennae the base station has (it may have more than a hundred), the more elements there are in the channel matrix. In addition, a precoder matrix for each tone must be determined from the channel matrix for that tone. The determined precoder matrices must then be multiplied by each tone prior to transmission. A very large number of measurements and computations are therefore required. This is resource-intensive.

It is desirable to find a method of precoding which overcomes and/or substantially mitigates some or all of the above-mentioned and/or other disadvantages of the prior art.

According to a first aspect of the invention there is provided a method of processing a MIMO signal using multi-carrier modulation, the MIMO signal being for transmission over a communications channel, the method comprising: Measuring channel state information for the communications channel in relation to a first tone, the first tone being a component tone of the MIMO signal; Precoding, using the measured channel state information, a second tone, the second tone being a component tone of the MIMO signal, wherein the second tone is a different tone to the first tone.

Embodiments of the invention enable channel state information in relation to a first tone to be used in precoding the first tone and re-used in precoding other tones. This reduces the number of channel measurements and processing computations required. The invention is particularly advantageous if the channel state values are "coherent" over time and/or frequency. The channel state values exhibit coherence if they remain relatively constant for periods of time and/or over sections of the frequency spectrum.

The form of multi-carrier modulation may be orthogonal frequency-division multiplexing. The method may further comprise transmitting the second tone. The second tone may be transmitted, as part of the MIMO signal, to a receiver apparatus that may be a cellular telephone or customer premises equipment.

The method may further comprise receiving the second tone at the receiver apparatus. The method may further comprise determining an error associated with the second tone. This may comprise comparing a received value of the second tone at the receiver apparatus to the transmitted value of the second tone. This may comprise determining a signal to noise ratio associated with the second tone. The method may comprise determining if the determined error is above a predetermined threshold. The method may be repeated if the error is determined to be below the predetermined threshold.

The method may further comprise precoding the first tone using the measured channel state information in relation to the first tone and transmitting the first tone.

The MIMO signal may be a wireless signal and may be transmitted from a transmitter apparatus which may be located in a cellular base station. The transmitter apparatus may have a plurality of transmitters. The transmitter apparatus may have at least 2 transmitters, i.e. antenna elements, and may have more than 50 transmitters and may have more than 100s transmitters.

The step of measuring channel state information may comprise transmitting training signals from the transmitter apparatus into the communications channel and analysing the received training signals.

The method may further comprise using the channel state information measured in relation to the first tone to determine a channel matrix in relation to the first tone and may comprise using the channel matrix to determine a precoder matrix in relation to the first tone. The step of precoding the second tone may comprise multiplying the second tone by the determined precoder matrix.

In some embodiments, the first and second tones are component tones of the same symbol of the MIMO signal. I will refer to these embodiments as frequency-coherence embodiments as they exploit the frequency-coherence of the channel by enabling the channel estimation in relation to a tone to be used to precode that tone and re-used to precode a tone transmitted at a later time. In these embodiments, the second tone may be adjacent in frequency to the first tone. By adjacent in frequency it is meant that the second tone is one carrier frequency interval higher or lower than the first tone. In some embodiments the second tone is more than one carrier frequency higher or lower than the first tone.

The method may further comprise measuring channel state information in relation to a plurality of first tones in the symbol and precoding a plurality of second tones in the symbol using the channel state information measured in relation to the plurality of first tones.

If the error rate is determined to be below the predetermined threshold, then the method may be repeated in relation to one or more subsequent symbols. In other words, the method may further comprise, in relation to a subsequent symbol of the MIMO signal, measuring channel state information in relation to a tone having the same frequency as the first tone and precoding, using the measured channel state information, a tone having the same frequency as the second tone.

In some embodiments, the second tone is a component tone of a second symbol of the MIMO signal, where the second symbol is transmitted subsequently to the first symbol.

I will refer to these embodiments as time-coherence embodiments as they exploit the time-coherence of the channel by enabling the channel estimation in relation to a tone to be used to precode that tone and re-used to precode a tone transmitted at a later time. The second tone may be a component tone of the symbol immediately following the symbol comprising the first tone. The second tone may have the same frequency as the first tone. In these embodiments, the method may be performed in relation to some or all of the tones in the symbol. In other words, the method may further comprise measuring channel state information for the communications channel in relation to a plurality of first tones, the plurality of first tones being component tones of a symbol of the MIMO signal and precoding, using the measured channel state information, a plurality of second tones, wherein the plurality of second tones are component tones of a subsequent symbol of the MIMO signal.

If the error rate is determined to be below the predetermined threshold, then the method may further comprise precoding, using the channel state information measured in relation to the first tone of the first symbol, a third tone having the same frequency as the first tone, where the third symbol is transmitted subsequently to the second tone.

According to a second aspect of the invention there is provided a system for processing a MIMO signal using multi-carrier modulation, the MIMO signal being for transmission over a communications channel, the system comprising: Measurement apparatus adapted to measure channel state information for the communications channel in relation to a first tone, the first tone being a component tone of the MIMO signal; A precoder adapted to precode, using the measured channel state information, a second tone, the second tone being a component tone of the MIMO signal, wherein the second tone is a different tone to the first tone.

An embodiment of the invention will now be described in detail, for illustration only, and with reference to the accompanying drawings, in which: Fig. 1 is a schematic view of the components of a transmitter apparatus in accordance with the invention; Fig 2 is a flow chart summarising the method steps in accordance with a frequency-coherence embodiment of the invention; Fig 3 is a flow chart summarising the method steps in accordance with a time-coherence embodiment of the invention.

Fig 1 is a schematic view of a transmitter apparatus 10 in accordance with the invention for the transmission of wireless MIMO signals at a cellular base station to a receiver. The transmitter apparatus 10 has a transmitter array 4 having three transmitters 1, 2 and 3 (although, in a real-life base station there may be more than 100 transmitters). Prior to transmission the signal is precoded. This involves multiplying the signal by a precoder matrix. The purpose of this is to remove interference caused by the signals transmitted by the other two transmitters and by the channel itself. The precoder matrix is constructed using the channel matrix which contains channel state information. This channel state information is obtained using channel estimation. This involves transmitting training signals from the transmitters 1,2,3 and analysing the effect that the channel has on the training signals.

The process of producing the training signals will now be described. Within the processor of the transmitter apparatus 10 there are a plurality of data encoder and mapper components, shown generally in Fig 1 as 5. These provide 2048 outputs (referred to as tones) each having a unique amplitude and phase. The tones are precoded as described above by precoder component 6. The tones are then input into an Inverse Discrete Fourier Transform (IFFT) 7 and then an analogue front end (AFE) 8. The AFE 8 multiplies the tones by a high frequency signal so that the tones are in an appropriate frequency band for wireless transmission. This is known as up-conversion and carries the baseband into a carrier frequency in the GHz range. In this example that is approx.

2.4GHz.

The AFE 8 then transmits the 2N = 2048 tones from transmitter 1. Each tone has a different frequency, such that the tones are spaced apart in frequency (by approximately SW\ This process is repeated for further data sets such that a further 2048 tones are transmitted from each of transmitters 2 and 3.

A receiver apparatus (not shown) detects the tones using multiple receivers (i.e. antennae) and the detected values are compared to the transmitted values in order to determine, for each tone, a channel matrix. The channel matrix comprises multiple elements: [h11 h12 h13 h21 h22 h23 h31 h32 h33 where e.g. h11 is the channel coefficient for transmitter 1 to receiver 1, hi2 is the channel coefficient for transmitter 1 to receiver 2 (i.e. a measure of the extent to which the signal transmitted from transmitter 1 interferes with the signal transmitted from transmitter 2).

The channel matrix is frequency-dependent, i.e. it varies from tone to tone. Therefore a total of 2048 channel matrices are determined, one for each tone. Precoding matrices P are then determined, one for each tone, using the respective one of the channel matrices H1-2048. This is done using the following formula: P1-2048= (H1-2048)14((11-2048) (H12048)' y1 The precoder matrix is then used to precode the real data signals. Specifically, each of the 2048 outputs of the data encoder and mapper is multiplied by its corresponding precoder matrix. For example, the data encoder and mapper output corresponding to the lowest frequency tone (xi) is multiplied by the precoder matrix H that corresponds to tone xl. (P, was derived from the channel matrix HI corresponding to tone xi). Similarly, the second lowest frequency tone (x2) is multiplied by the precoder matrix P2 corresponding to tone x2. (P2 was derived from the channel matrix H2 corresponding to tone x2). This continues until all 2048 tones have been precoded. The precoded tones are passed through the IFFT and AFT and transmitted.

This process involves a very large number of computations. The present invention reduces this number of computations. The channel matrix displays "coherence" over certain sections of the spectrum. In other words, two tones of similar frequencies can have similar channel matrix matrices. In the present invention, instead of precoding adjacent tones (eg xl and x2) using two different precoder matrices (P1 and P2), both xi and x2 are precoded using Pl. If the channel matrix is similar for xl and x2, ie if it remains fairly constant over the frequencies of tones xl and x2, such precoding will be reasonably effective (in other words the interference suffered by tone x2 will be effectively mitigated as a result of the precoding). This is because precoder matrix Pi is derived from channel matrix Hi which is similar to channel matrix Hz. This simplifies the training stage, as channel matrix H2 does not need to be estimated. Moreover, P2 does not need to be determined. This method therefore greatly reduces the number of calculations required.

This embodiment of the invention can be summarised as the re-use of a tone's precoder matrix in the precoding of an adjacent tone. This method is summarised in Fig 2.

The next step is to check the effectiveness of this method. The receiving apparatus receives each of the transmitted tones and detects their values Y. The receiver apparatus then transmits these values Y back to the base station. At the base station, the received values Y of the tones are compared to the original values X of those tones and the error rate is determined. If the determined error rate is below an acceptability threshold, then, in the transmission of the next symbol, PI is again used to precode both xl and x2. If the determined error rate is above an acceptability threshold, the subsequent transmission will require the determination of P2. Therefore the training stage will require the estimation of both H1 and Hz.

One way assess whether the error is above an acceptability threshold is to use the following formula: BW/Af maxal log2 (1 + In the above formula, si is the amount of power applied to the ith tone. yi is the signal to noise ratio for the ith tone which is determined by comparing the value of the ith tone as received by the receiver to the original value of the ith tone. 1-1 is the required Bit Error ratio for the i'h tone. BW is the signal bandwidth (e.g.100MHz) and A is the frequency interval between adjacent tones (approx;2). If the result of the above formula is above an acceptability threshold, the method is determined to be effective, i.e. coherence has been detected. If not, the method is determined not to be effective, i.e. coherence has not been detected.

As well as varying with respect to frequency, the channel matrix also varies with respect to time. For this reason the training stage needs to be performed before every transmission. Systems of this nature transmit signals at time intervals. As described above, in prior art systems a training stage is performed which estimates the 2048 precoder matrices. This occurs at a particular point in time. These precoder matrices are then used to precode the tones, which are then transmitted. One time interval after the first training stage is performed, a second training stage is performed. This estimates the channel matrix at this subsequent time. These measured channel matrices are used in the precoding of the next signal.

As with frequency, the channel matrix can display coherence over time. In other words, the channel matrix for a given tone at T=0 can be similar to the channel matrix for that tone one time interval later (ie at T= At). The present invention exploits this to reduce the number of measurements and calculations necessary. This is done as follows.

At time T=0, a first training stage is performed. This estimates a channel matrix H1 for (among other tones), the lowest frequency tone xl. This channel matrix is then used to determine a precoder matrix Pi for tone xi. Tone x1 is then multiplied by P1 and transmitted (via the IFFT and AFE). At time T= At a second training stage is performed.

This second training stage does not include estimating a channel matrix in respect of tone xl at time T= At. Instead, tone xl (T= At) is precoded by multiplying it by the previously-determined Pi, ie PI at time 1=0. The thus-precoded tone xi (T=At) is then transmitted (via the IFFT and AFE). As this method does not require a channel estimation or precoder matrix determination in respect of tone xl (T=At), this method is less resource-intensive that the prior art. This method is summarised in Fig 3.

The next step is to check the effectiveness of this method. As described above, tone xi (T=At) is precoded using tone P1 (T=0) and transmitted (along with all the other tones). The receiving apparatus receives the transmitted tone and detects its value Y1. The receiver apparatus then transmits value Y1 back to the base station. At the base station, the received value Y1 is compared to the original value xl of the tone and the error is determined. If the determined error is below an acceptability threshold, then the method of re-using Pi (1=0) to precode tone xi (T=At) has been found to be acceptable, indicating a time coherence in the vicinity of tone xl. In the subsequent transmission, ie at T=2At, PI (1=0) may again be used to precode xi (T=2At).

One way assess whether the error is above an acceptability threshold is to use the following formula: BW/4 maxar log2 (1 + 1.

In the above formula, si is the amount of power applied to the ith tone. yi is the signal to noise ratio for the ilh tone which is determined by comparing the value of the i'h tone as received by the receiver to the original value of the ill' tone. IT is the required Bit Error ratio for the ill' tone. BW is the signal bandwidth (e.g.100MHz) and A, is the frequency interval between adjacent tones (approx = 48.9 kHz). If the result of the above formula is above an acceptability threshold, the method is determined to be effective, i.e. coherence has been detected. If not, the method is determined not to be effective, i.e. coherence has not been detected.

This embodiment of the invention can be summarised as the re-use of a given tone's precoder matrix for precoding the subsequent transmission at that tone. The skilled person will understand that this can be done for other tones besides the lowest frequency tone. Indeed, the precoder matrices in respect of all the tones can be re-used to precode the subsequent transmission at each of those tones.

The optimal modelling for coherence detection, for both the frequency-coherence and time-coherence embodiments can be formulated in more detail as follows: BW/Ay max Ai log2 (1 + (1) s.t.

0 < BW, BW/4f Af Si PT, Ernin Ei Emax, 0 HBW a.) er,/3)",,, implicitly satisfied Where an OFDM system with: - a fixed bandwidth (BW) - total power budget is PT -Quality of Service (QoS) demands are in the form of Bit Error Ratio (BER) and captured in Fi - carrier to noise ratio is assumed to be perfectly known (y) - carrier spacing (al) is to be optimised to ensure satisfaction of QoS - assumptions to be made on the DSP operations (WE, MIMO), computational complexity of each technique ft, p)(.01, denote time and bandwidth coherence To solve problem one whilst many algorithms can be devised by those who are skilled in the art, we opted for the simplest solution using bi-section search which goes as follows: Initialisation and loop: 2" with N E (8w/a1) M E t10, 9, ...,1) (Modulation) Compute: Water-filling (convex form) Objective check: - Modulate (QAM) Solution found Received symbol equalisation - Determine BER If BER higher than target value then: - Decrease M & repeat compute If 1 fails, decrease and reset M and Compute - Repeat until solution is found -EXIT

Claims (12)

1. Claims 1.A method of processing a MIMO signal using multi-carrier modulation, the MIMO signal being for transmission over a communications channel, the method comprising: Measuring channel state information for the communications channel in relation to a first tone, the first tone being a component tone of the MIMO signal; Precoding, using the measured channel state information, a second tone, the second tone being a component tone of the MIMO signal, wherein the second tone is a different tone to the first tone.
2. 2. A method as claimed in claim 1, the method further comprising transmitting the second tone.
3. 3. A method as claimed in claim 1 or claim 2, the method further comprising receiving the second tone at a receiver apparatus.
4. 4. A method as claimed in any preceding claim the method further comprising determining, based on the measured channel state information, that the communications channel is coherent for the first and second tones.
5. 5. A method as claimed in claim 4, the method further comprising determining a signal to noise ratio associated with the second tone and using the signal to noise ratio to determine whether an error associate with the second tone is above a predetermined threshold.
6. 6. A method as claimed in any preceding claim, wherein the first and second tones are component tones of the same symbol of the MIMO signal.
7. 7. A method as claimed in claim 6, wherein the second tone is adjacent in frequency to the first tone.
8. 8. A method as claimed in any of claims 1 to 4, wherein the second tone is a component tone of a subsequent symbol of the MIMO signal to the first tone.
9. 9. A method as claimed in claim 8, wherein the second tone is a component tone of the symbol immediately following the MIMO symbol comprising the first tone.
10. 10. A method as claimed in claim 9, wherein the second tone has the same frequency as the first tone.
11. 11. A system for processing a MIMO signal using multi-carrier modulation, the MIMO signal being for transmission over a communications channel, the system comprising: Measurement apparatus adapted to measure channel state informafion for the communications channel in relation to a first tone, the first tone being a component tone of the MIMO signal; A precoder adapted to precode, using the measured channel state information, a second tone, the second tone being a component tone of the MIMO signal, wherein the second tone is a different tone to the first tone.
12. 12. A computer-readable medium programmed to carry out the method as claimed in any of claims 1 to 10.