CHANNEL QUALITY REPORT PROCESSES, CIRCUITS AND SYSTEMS BACKGROUND

In a wireless network, a base station (designated by Node B or eNB) communicates with user equipment (UE), such as a cell phone, a laptop, or a PDA. Base station eNB transmits reference signals or pilot signals to UE, which generates a channel estimate based on the reference signal, as impacted by interference and noise. The system bandwidth is divided into frequency-domain groups or subbands that encompass resource blocks (RBs) according to group size or subband size. An RB is the smallest allocation unit available in terms of frequency granularity allocated to UE by a base station scheduler module. UE determines a channel quality indicator (CQI) for each RB or for each subband based on the channel estimation. The CQI metric is suitably a signal to interference noise ratio (SINR) after detection, the index to a supportable modulation and coding scheme, the index to a supportable code rate, a channel throughput measure, or other quality measure. UE feeds back the CQI for each subband or RB to eNB. More favorable CQI permits a higher data transfer rate of data streams by eNB to UE. By using multiple transmit and multiple receive antennas with transmit pre-coding in a multi-input multi-output (MIMO) system; improved throughput and/or robustness are obtained. The number of independent data streams (number of spatial codewords) is termed the transmission rank. A high level of operational overhead and uplink bandwidth is believed to have hitherto been involved if each of many UEs is to deliver feedback about many subbands to eNB. This can undesirably increase system processing delays and dissipation of power and energy which is especially problematic in mobile handset forms of UE. Accordingly, further ways of reducing the amount of communications feedback between user equipment and base station are desirable. SUMMARY

A form of the invention involves an electronic device that includes a first circuit operable to generate at least a first and a second channel quality indicator (CQI) vector associated with a plurality of subbands for each of at least first and second spatial codewords respectively and generate a first and a second reference CQI for the first and second spatial codewords, and operable to generate a first and a second differential subbands CQI vector for each spatial codeword and generate a differential between the second reference CQI and the first reference CQI, and further operable to form a CQI report derived from the first and the second differential subbands CQI vector for each

spatial codeword as well as the differential between the second reference CQI and the first reference CQI; and a second circuit operable to initiate transmission of a signal communicating the CQI report. Other forms of inventive electronic circuits, devices, processes and systems are also disclosed and claimed. BRIEF DESCRIPTION OF THE DRAWINGS

FIGS. 1 and 2 each provide a respective system block diagram of a MIMO OFDMA receiver and transmitter for UE or eNB improved as shown in the other Figures.

FIG. 3 is a flow diagram of an improved UE process having different CQI reporting modes. FIG. 4 is a diagram of subbands in the frequency domain for different codewords, showing an improved CQI reporting process.

FIG. 5 is a diagram of subbands in the frequency domain for different codewords, showing another improved CQI reporting process.

FIGS. 6 A and 6B are each a diagram of subbands in the frequency domain for different codewords, showing an improved Best-m Average CQI reporting process.

FIGS. 7 A and 7B are each a diagram of subbands in the frequency domain for different codewords, showing an improved Best-m Individual CQI reporting process.

FIGS. 8 A and 8B are each a diagram of subbands in the frequency domain for different codewords showing another improved Best-m Individual CQI reporting process. FIGS. 9-10 are each a diagram of codewords versus subbands enumerating a process sequence for scanning-based CQI reporting. FIG. 9 CQI reporting is in subband order, codeword by codeword. FIG. 10 CQI reporting is in codeword order, subband by subband.

FIG. 11 is a pair of side-by-side flow diagrams of a user equipment UE and a base station eNB, and shows an improved process for CQI reporting in UE, and an improved process in eNB to reconstruct the CQIs for subbands of codewords from the CQI report.

FIG. 12 is a flow diagram of a process for use in base station eNB to apply a scanning pattern and use a subband vector SV indicating selected individual subbands for Best-m CQI reporting involving multiple codewords in a MIMO system. DETAILED DESCRIPTION OF EXAMPLE EMBODIMENTS

In FIG. 1, a receiver 100 for a MIMO frequency division multiplex system is in a mobile handset 1010 or a fixed device. A receive portion 105 includes an OFDM module 106 having Q OFDM demodulators (Q >= P number of antennas coupled thereto, a

MIMO detector 107, a QAM demodulator plus de-interleaver plus FEC decoding module 108 and a channel estimation module 109. A feedback generation portion 110 includes a precoding matrix selector 111, a channel quality indicator (CQI) computer 112, rank selector 114, and a feedback encoder 113. Receive portion 105 in FIG. 1 receives data from a transmitter 150 of FIG. 2.

Channel estimation module 109 employs previously transmitted channel estimation pilot signals. Precoding matrix information can be obtained via an additional downlink signaling embedded in the downlink control channel or in a reference signal, or from a previously selected precoding matrix. Precoding matrix selector 111 determines a precoding matrix selection for the data transmission based on the channel/noise/ interference estimates from block 109, in tandem with rank selector 114 determination of a preferred rank R for number of spatial code words to be accommodated. CQI is calculated based on the selected precoding matrix or its index (PMI) in a PMI codebook. The precoding matrix selection and CQI are computed for the next time the user equipment UE of FIG. 1 is scheduled by the transmitter (e.g., a base station FIG. 2) to receive data. Feedback encoder 113 then encodes the precoding matrix selection, CQI information and rank R and feeds them back to transmitter 150 before the data is transmitted.

FIG. 2 illustrates a system diagram of a transmitter 150 such as for a base station eNB 1050 in an OFDM communication system. The transmitter 150 includes a transmit portion 155 and a feedback decoding portion 160. The transmit portion 155 includes a modulation and coding scheme (MCS) module 156, a pre-coder module 157 and an OFDM module 158 having multiple OFDM modulators that feed corresponding transmit antennas. Feedback decoding portion 160 includes a receiver module 166 and a decoder module 167. MCS module 156 maps the codeword(s) to the R layers or spatial streams. Each codeword consists of FEC-encoded, interleaved, and modulated information bits. A selected modulation and coding rate for each codeword are derived from the CQI. Pre- coder module 157 employs a precoding matrix selection indexed by a precoding matrix index (PMI), corresponding to receiver 100 grouping of RBs, from feedback decoder module 167. Precoder 157 linearly cross-combines the R spatial stream(s) into P output data streams.

In FIG. 3, UE proceeds from BEGIN 2105 to select a hybrid CQI feedback configuration mode at a decision step 2110 which monitors configuration transmissions

from eNB. If a mode called UE Configuration Mode here is selected, a step 2114 activates UE Configuration Mode unless eNB mandates a mode called eNB Configuration Mode here and a branch to a step 2118 activates the eNB Configuration Mode. UE Configuration Mode and eNB Configuration Mode, respectively establish parameters and controls over hybrid CQI feedback and can also define a scanning pattern or sequential feedback order, see e.g. FIG. 9-10. A step 2128 provides information for step 2165 on configured CQI feedback mode.

For various forms of Best-m CQI feedback, i.e. involving selected and unselected subbands, step 2122 responds to configuration for CQI Relative Mode or CQI Absolute Mode or CQI Directed Mode, step 2124 for a number m of selected subbands, step 2126 specifying total number M of subbands and a width granularity number L for m selected subbands, and step 2128 specifies a Feedback Process Code for Best-m. If step 2122 establishes CQI Absolute mode, operations go to a loop having steps 2130, 2135, 2138 that select all CQI(J) such as SINR(J) in subband j that exceeds a predetermined threshold. (CQI can also index to highest supportable MCS, or code rate. The loop uses the parameters M and L that were established in step 2126 to search all M subbands using width L. Step 2130 detects whether the threshold is exceeded by the SINR in a given sub-band j. If Yes in step 2130, then step 2135 records a one (1) at a position j (current value of index j) in a subband vector SV(j) and then goes to step 2138 to increment the index j and/or codeword index r. If No in step 2130, then operations instead record a zero (0) at position j in SV(J) and proceed to step 2138 to increment the index j. If incremented index j exceeds the number M of subbands, the loop is Done and step 2130 goes to a step 2145. If step 2122 establishes CQI Relative mode, a step 2140 selects a number m of subbands j of width L having, e.g., highest CQI. Step 2140 uses the parameter m established in step 2124 and M and L from step 2126, and searches all M subbands, whence step 2140 goes to step 2145. If step 2122 establishes a CQI Directed mode under eNB or UE Configuration Mode, operations go to step 2145 and directly load a subband vector SV(J) (or SV(rj)) with a particular series of ones and zeros responsive to, and/or as directed and/or specified by base station eNB or UE. In this way, particular subband(s) can be selected for UE to report respective CQI values for the subband(s). At step 2145, a subband vector SV(J) or SV(rj) is now constituted and has M elements forming a series of ones and zeros that represent whether each subband is selected or not. Next, a step 2150 counts the number of ones in subband

vector SV to establish the resulting number m or m(r) of selected subbands resulting when the CQI Absolute Mode has been executed. Note that defining the subband vector (SV) as a IxM vector containing Is and Os is one exemplary method to indicate the position of the selected subbands. Alternatively, the position of the selected subbands is reported using a compressed label or codebook index (e.g., bits joint quantized into fewer bits by UE codebook lookup using for example Iog ₂ (c^ ) bits, where C^ denotes number of combinations of M elements taken m selected subbands at a time.

Step 2155 generates one or more CQI values (ki in number) for the number m of selected subbands. A step 2160 generates one or more CQI values (k ₂ in number) to either individually or collectively describe the un-selected subbands that are M-m in number. A total number k = ki + k ₂ of CQI value(s) are generated to describe the M subbands, where 1<= k <= M. (M is number of subbands.) The CQI value(s) generated in step 2160 for the un-selected subbands have a precision or accuracy that is less than the precision or accuracy of the CQI values for the number m of selected subbands generated in step 2155. This feature efficiently reduces bits used to communicate the CQI values for all M subbands. Operations go from Best-m last step 2160 to a succeeding step 2165. For CQI feedback other than Best-m, operations go from step 2128 to step 2165 and bypass steps 2122 and 2130-2160. Step 2165 assembles the CQI values into a CQI vector, to which is associated a Feedback Process Code from step 2128, a scanning code identifying a scanning pattern (e.g., from FIG. 9-10), mode specifiers from step 2122, an identification UE_ID of the UE, CQI, rank and any other relevant configuration information or representing-information not already communicated by UE or already stored at base station eNB. Then depending on the Feedback Process Code, a decision step 2170 flows operations sequenced according to the scanning code to any one (or more) of several MIMO CQI feedback processes for rank R >=1 described elsewhere herein, such as Down Sample 2175, Predistortion 2180, Frequency Differential 2185, Transform/Wavelet 2190, Mean and Delta 2195 and/or Parameterized 2198. CQI feedback output is transmitted from each UE to base station eNB on uplink UL as a service request, or with a service request code. Quantization of the CQIs can involve an absolute value of at least one CQI or an average CQI of the selected sub-bands. Averaging uses arithmetic mean, geometric mean or exponential averaging, etc., as suitable. Also, a difference ) is used and implemented any of a variety of ways. For example, given two multibit values a and b,

f _dt ff(a,b) = RND(a-b) where rounding RND is any appropriate rounding function such as one that rounds to the nearest value expressed in the fewer number of bits to be supported. Another difference function performs a table lookup that outputs a table value to which the input values a and b are mapped by the table. CQI is expressed by any of decibels dB, ratio of signal power/noise power, index to the supportable modulation and coding schemes (MCS), index to the supportable spectral efficiency, and the difference function can be chosen based on an exponential difference log[exp(a)-exp(b)] or an arithmetic difference a-b or some other difference mapping.

It is believed that hitherto, CQI feedback methodology has responded to system constraints by applying a differential wideband CQI feedback approach as an alternative to subband-specific differential CQI feedback. The wideband alternative provides a limited amount of CQI feedback and thus gives an appearance of feedback bit-efficiency and sufficiency by itself. Analogously, subband-specific differential CQI feedback provides a more granular amount of CQI feedback and thus gives an appearance of wider feedback coverage sufficiency by itself. However, such alternative-process methodologies each insufficiently respond to the need for CQI feedback that is both highly efficient in terms of CQI feedback bits per subband per codeword and provides more extensive CQI feedback coverage of subbands and codewords. Some of the embodiments herein solve this problem by recognizing that these hitherto alternatively-treated CQI feedback processes should not be treated as alternatives but instead as complementary parts of a comprehensive CQI reporting solution as taught herein. Accordingly, subband-specific CQI reporting in FIG. 4 delivers such a comprehensive CQI reporting solution combining complementary parts wherein subband-specific CQIs for a given codeword are encoded differentially relative to the wideband CQI or other reference CQI (mean, median, etc.) for that codeword, and the wideband CQI (or other reference CQI) for all but one of the codewords are encoded differentially relative to the wideband CQI (or other reference CQI) for the remaining codeword. In this way, both the bit-efficient feedback of differentially encoded wideband CQIs for the codewords and the bit-efficient feedback of differentially encoded subband-specific CQIs for the codewords complement one another in the FIG. 4 embodiment and other analogous embodiments herein to provide both enhanced bit-efficiency and more extensive CQI coverage of subbands and codewords for MIMO.

In FIG. 4, the wideband CQI of codeword 2 is encoded differentially relative to the wideband CQI of codeword 1 (CWl). In this embodiment, instead of feeding back the absolute value of wideband CQI ^{2 wi} , the process feeds back its spatial differential ^ ² -"* ⁼ A ^fl ^ ^t -A**) _re l _a tive to the first codeword's wideband CQI S _hwb . The all sub-band CQI of codeword 2 is reported differentially, after generating it using ^{2 wi} itself as the reference. That frequency differential CQI

δS,-/ - fdiff ψr,wb ' $ rj ) ,,j / ; _ — = l 1 r,, ...... MM,, h is reported to eNB, with fewer bits, and is one example of a differential subbands CQI vector that is generated for each codeword r=2, ...R. The reference CQI is reported with high resolution (e.g., 4-5 bits) and the subband differential CQI has smaller dynamic range reported with lower resolution (e.g., 2-3 bits). In Spatial differential CQI reporting one codeword is selected as reference codeword and its reference CQI value is reported with high resolution (e.g., 4 or 5 bits). CQI of remaining spatial codewords is encoded differentially with respect to the reference CQI of the reference codeword and reported with lower resolution (e.g., 2-3 bits). A variant embodiment relative of FIG. 4 reports individual differential CQIs designated δCQIi _j CQIi, _W b), for codeword CWl using lower resolution (e.g., 2-3 bits) for each subband j (or selected subband j in a Best-m set) for codeword CWl. For higher codewords, the CQI reporting feeds back individual spatial differential CQIs δ CQI _1J CQI]J, differencing subband CQIs pairwise across codewords, using lower resolution (e.g., 2-3 bits) for subband j ( or each selected subband j in a Best-m set) for each codeword CW2, etc. As in FIG. 4, wideband CQIi _iW b is reported at higher resolution (e.g., 4-5 bits) for codeword CWl. Wideband CQI _r>W b for codewords r=2,..M is encoded spatial differentially with respect to wideband CQIi _iW b of codeword CWl, and sent back using fewer bits (e.g., 2-3 bits/codeword) and hence keeps the feedback overhead down. This differential encoding is expressed by δ CQI i,wb), r = 2...R and wb is the wideband CQI.

In FIG. 5, a process embodiment performs additional differential reporting between the differential CQI sequences of each codeword. For example, denote the differential CQI sequence of codeword 1 CWl as ^{δ V ^Mi2 . - ^5I " ^] , differential CQI sequence of codeword 2 as ^ ^21, ^ ^22, --^ 1 _Inste ad of feeding back^ ²¹ -^ ²² --- ⁵ ^ ^] ,

the process feeds back f ^J _d ^ _iff ( V{A"S" ₂ 2 ₁ i A—S ₂ 2 ₂ 2 ^, .—...S ₂ 2 _M M } ^J λ ^{> 1} δ—S _n 11 A—S ₁ 1 ₂ 2 ^, .—...S ₁ ^I _M ^M ) ^J )/ _O r/ _λ /δS _2j , AS ₁ J where the subtraction of vectors between codewords is performed element- wise by subband j. This is one example of a spatial differential between a frequency differential subbands CQI vector and another frequency differential subbands CQI vector, wherein the differential between them is also called a Delta Delta herein.

Note that the definition of differential CQI can be the same or different for different differential quantities. Also, the difference function f^ ) for computing the differential of the differential CQI in some embodiments can be specified differently than difference function f _ώ fl() for computing the differential CQIs for purposes of the subtraction of vectors. Note that "subtraction" refers to is suitably specified difference function for the element wise operation, and is not limited to particular definition of subtraction.

In FIG. 5, remarkably, wideband CQI of CW2 is encoded differentially to wideband CQI of CWl and furthermore the frequency differential subbands CQI vector for CW2 is encoded spatially differentially to the frequency differential subbands CQI vector of CWl. Note that here the differential CQI sequence (in frequency domain) is computed over the sub-band CQI set S, e.g., that is semi-statically configured by higher layer to include all the sub-bands, or only a sub-set of the sub-bands in the system bandwidth. Delta Delta ( δδ ) is likely to have desirably reduced dynamic range compared toδS _rj when there is substantial positive correlation between δ Si _j and A S _2J , for instance. Reduced dynamic range facilitates effective compression. Favorable UE geometry, high wideband SINR for all codewords, and favorable precoding matrix PM at eNB each contribute to a high positive correlation between δ S _lo and δS _2j (i.e., correlation between differential CQIs for different codewords r). Accordingly, some embodiments operate eNB so that precoding matrix PM is checked or even optimized for delivery of favorable wideband SINR for all codewords. If favorable SINR is not achieved at a given MIMO rank R>1, then the rank R can be decremented or otherwise reduced and precoding matrix PM is recomputed until favorable SINR is achieved at a lower MIMO rank or until the rank is reduced to one, R=I.

In a category of embodiments called Joint Sub-band Selection on different codewords, the same set of sub-bands is selected for CQI reporting each of the codewords. The A-suffixed FIGS. 6A, 7A, 8A each depict Joint Sub-band Selection on

different codewords for respective embodiments. Because different codewords select the same set of sub-bands, a single indication of their position suffices for all codewords. Selection of a sub-band is based on a performance metric defined over all codewords, and the result of the selection is represented by SV(j) =1 if a given subband j is selected and otherwise SV(J)=O. Alternatively, the location of the m selected subbands is jointly reported using compressed label using log ₂ (c^ j bits. Some examples of different implementations for the performance metric are 1) maximizing the sum throughput, summed over all codewords, 2) maximizing the arithmetic/exponential average CQI over all codewords, 3) maximizing the mean CQI over all codewords, and 4) minimizing the difference between CQIs of different codewords.

Another category of embodiments performs spatial differential CQI quantization without frequency CQI quantization, where an example is given in FIGS. 6A and 6B for UE-selected (best-m) average CQI and FIGS. 7A and 7B for UE-selected (best-m) individual CQI reporting. FIGS. 6A and 6B are examples of best-m UE-selected average CQI report with spatial differential quantization. In brief, spatial differential reporting is performed between the wideband CQIs of different codewords, and between the best-m average CQI (UE-selected) between different codewords, preferably using fewer bits (e.g., 2-3 bits). In addition, the wideband CQI and best-m average CQI (UE-selected) of the reference codeword is also fed back with high resolution (e.g. 4-5 bits). For instance, consider a

MIMO-OFDMA system with two spatial codewords. FIG. 6A and 6B are each a diagram of subbands in the frequency domain for different codewords, showing wideband CQI report for unselected subbands of the first codeword, wideband CQI report for unselected subbands of the second codeword encoded differentially with respect to the wideband CQI of the first codeword (spatial delta), best-m average CQI report for selected subbands of the first codeword, and best-m average CQI report for selected subbands of the second codeword encoded differentially with respect to the best-m average CQI of the first codeword (spatial delta), and in FIG. 6A where the selected (best-m) subbands have the same subband indices across codewords, and in FIG. 6B where the selected (best-m) subbands have different subband indices when compared across codewords. Some embodiments also perform additional frequency differential CQI compression, for example by reporting the best-m average CQI of the selected subbands of the first codeword encoded differentially with respect to the wideband CQI of the first codeword.

Similarly, such frequency differential CQI report can also be applied to the remaining codeword (r = 2, ... R) .

FIGS. 7A and 7B are examples of best-m UE-selected individual CQI report with spatial differential quantization. Spatial differential reporting is performed between the wideband CQIs of different codewords, and between the best-m individual CQI (UE- selected) for each selected subband j between different codewords, preferably using fewer bits (e.g., 2-3 bits). In addition, the wideband CQI and best-m individual CQI (UE- selected) for each of the selected subbands of the reference codeword is also fed back with high resolution (e.g., 4-5 bits). For instance, consider a MIMO-OFDMA system with two spatial codewords. FIGS. 7A and 7B are each a diagram of subbands in the frequency domain for different codewords, showing wideband CQI report for unselected subbands of the first codeword, wideband CQI report for unselected subbands of the second codeword encoded differentially with respect to the wideband CQI of the first codeword (spatial delta), Best-m individual CQI report for each of the selected subbands of the first codeword, and Best-m individual CQI report for each of the selected subbands of the second codeword encoded differentially with respect to the Best-m individual CQI of the corresponding subband of the first codeword (spatial delta). In FIG. 7 A the selected (Best-m) subbands have the same subband indices across codewords, and in FIG. 7B the selected (Best-m) subbands have different subband indices when compared across codewords. Additionally, some embodiments perform additional frequency differential CQI compression to the first codeword (i.e. the reference codeword), by reporting the Best-m individual CQI of each of the selected subbands of the first codeword encoded differentially with respect to the wideband CQI of the first codeword. Also, such frequency differential CQI report can be applied to remaining codeword r = 2, ... R). In FIGS. 8 A and 8B, a differential encoding of the differential CQI(s) (Delta

Delta) is quantized for the second codeword CW2, and any other codewords CW3, etc. The Delta Delta is given by δ δ CQI ₁ = fd _ϋ (fddCQh _φ CQI _r , _wb ), f ^gCQI ₁ , _p CQIi, _wb )) for r = 2, ...R. The Delta Delta is fed back for each higher codeword CW2, etc., using lower resolution (e.g., 2 or 3-bits). The report includes wideband CQI designated CQI] _;H * (e.g., 4-5 bits) and differential encoding f _dl g(CQIi _j , CQIi, _W b) (e.g., 2-3 bits per Best-m subbband) for CWl. Also reporting for each higher codeword has differential encoding Si, _w b), r = 2...R for unselected subbands using lower resolution (e.g., 2-3 bits). Further, the Delta Delta for selected subbands j is fed back as δ δ CQI ₁ = _j

(f _d ,fl(CQI _rj , CQI _r,wb ), Z _d1 J _j (CQI _1J , CQIi _,wb )) for r = 2, ...R using lower resolution (e.g., 2-3 bits). Joint Sub-band Selection in FIG. 8A or Independent Sub-band Selection FIG. 8B can be reported. In the Independent case, a respective indication SV(rj) of the positions of the sub-bands selected is reported for each codeword by operations in FIG. 3. For the Joint case only one indication SV(j) of the best m sub-bands position is reported common to all codewords is reported for each codeword by operations in FIG. 3. Note that the definition of differential CQI is specified the same or different for different differential quantities.

Description now turns to FIG. 9-10 for embodiments that include Scanning -based selective sub-band CQI Reporting herein. For example, the scanning process scans the system bandwidth according to a scanning pattern, to select one or several sub-bands. Large feedback overhead of systems lacking compression is substantially reduced, which allows the amount of CQI reporting to be increased so that the entire system bandwidth is covered in the CQI report. Uplink overhead is consequently decreased, uplink control signaling design is simplified and improved, uplink feedback coverage and MIMO control channel coverage are improved.

In connection with FIGS. 9-10, 1, 2, and 3, configurable scanning pattern CQI reporting embodiments can improve network speed and CQI feedback-related latency in UE and in eNB, and help ameliorate and solve these problems by organizing, coordinating, parallelizing and/or pipelining the CQI feedback processing herein for lower latency and higher system speed and relax constraints. Scanning herein makes the index order of operations, called a scanning pattern over subband/codeword indices (j, r), more uniform from block to block in UE and coordinates with index order of operations in eNB. UE operations, e.g. in FIGS. 11 and 12, such as CQI generation, differencing, writing (storing) and reading (loading), and transmission of the CQI feedback in UEi are coordinated by using a same configurable scanning pattern. The same scanning pattern is applied in eNB, e.g. in FIGS. 11 and 12, uniformly for eNB operations such as CQI reception, writing (storing) and reading (loading) of CQI feedback information in decompression and decoding for low latency recovery of UEi CQI(r,j)== S ₁ ^ ₀ and to facilitate scheduling in eNB. eNB receives encoded CQI feedback and on-the-fly de- differences that CQI feedback from a given UEi by the same scanning pattern in eNB as in UEi by which that CQI feedback streams into eNB. In a Sequential Scanning pattern, the fc-th CQI feedback reports CQI of the j ^' -th sub-band according to a pattern RB# = k +

L(j-1) for which k = 1, 2, ...L is repeated for each j = 1, 2, ...M according to a scanning loop executed by processor or other circuit with the sequential scanning pattern. In Down-Sampled Scanning herein, The UE feeds back quantized versions of every L-th CQI, and downsampling has a specified offset k that is incremented to report CQI for each resource block RB# = L(J-I) + k, where j = 1, 2, ...M is repeated for each k = 1, 2, ...L by a process loop.

CQI scanning scans a 2-dimensional matrix or space-frequency grid in both frequency domain and space domain. A configurable scanning covers the frequency- space domain by transmitting a CQI report in more or less compressed form based on CQI of exhaustive non-overlapping (mutually exclusive) proper subsets { CQI(SB _j , CW _r )} or averages over such subsets of the set of CQI having all indices j, r that encompass the frequency-space domain. In FIG. 9, Scanning feedback is performed first in the frequency domain, and then in the spatial domain. At the beginning, the 1 ^st codeword is selected, and scanning feedback is performed in the frequency-domain for this codeword. After codeword 1 has been scanned, codeword 2 is selected and its CQI is fed back according to a scanning feedback pattern (1,2,...1O). In FIG. 10, Scanning feedback is performed first in the spatial domain r, and then in the frequency domain j. At the beginning, a sub- band j is selected, and scanning feedback is performed in the spatial-domain r to feed back CQI(r=l, j), CQI(r=2,j). Then the next sub-band j+1 is selected according the scanning reporting pattern in the frequency domain, and its spatial CQI vector CQI(r=l, j+1), CQI(r=2,j+l) is scanned and fed back (1,2,...1O).

Various other scanning feedback processes 1) combine the scanning reporting processes shown in FIGS. 9 and 10, or 2) handle codewords having different numbers of subbands j per codeword, or 3) handle codewords not only having different numbers of subbands per codeword, but also varying widths of subbands varying with codeword index r; or 4) handle codewords having corresponding numbers M of subbands per codeword, but varying widths L of subbands j among the subbands for each codeword, or 5) CQI of a best-m spatial codeword at a sub-band in some embodiments is given more priority and scheduled to be fed back prior to unselected subbands in the space-frequency grid. Note that any of the various described scanning-based CQI reporting structures and methods of FIGS. 9-10 and other scanning embodiments described herein are applicable to scanning-based CQI reporting of CQIs, differentially encoded CQIs (Deltas), differentially encoded CQI differences (Delta Deltas), codebook indices for joint

quantization of CQI, Deltas, and/or Delta Deltas, as described herein and/or as shown in any of the Figures herein, and further applicable to any combination of any one, some or all of the foregoing. Generalizing, a scanning pattern P ₁ herein for UEi is a mapping from a sequence of the counting numbers (e.g., s= 1, 2, 3, ...N, where N=RM) to a two- dimensional discrete index value space (r, j) for the subbands j of each codeword r, where 1 < r < R and 0 < j < M. (Index j=0 is suitably used for CQI reference value for a given codeword.) In some embodiments, the scanning process includes a scanning loop over a one-dimensional scanning pattern index (e.g., s= 1, 2, 3, ...N, where N=RM) and the loop kernel has a mapping function from the one-dimensional scanning pattern index to generate indices r, j in a two-dimensional discrete index value space (r, j) for the subbands j of each codeword r. The loop kernel further includes a computation and/or read or write to storage that involves variables as a function of the indices in two- dimensional discrete index value space (r, j). UEi and eNB use the same scanning pattern P ₁ so that communication occurs intelligibly between them. In some embodiments, a scanning code is exchanged between UEi and eNB (or vice versa) that identifies a scanning pattern P ₁ .

In FIG. 11, a process flow for Mean-Delta and/or Delta Delta CQI reporting can be compared with the other Figures. In the UE, a step 2510 quantizes the mean or median or reference or wideband CQIs Fo _,r across all subbands collectively for each codeword r. Then a step 2515 generates wideband Deltas Do, _r = Fo, _r - Fo,i for codeword index values r=2,...R. Next a step 2520 generates subband deltas D _{j r} = S _J>r - F _0,r individually for each of the subband index values j= 1, 2,... M and for codeword index values r=l,...R. A succeeding step 2525 generates differentials of the differentials (Delta Deltas) δD _{j r} = D _{j r} - D _j i for each of the subband index values j= 1, 2,...M and codeword index values r=2,...R. In FIG. 11, a further step 2530 sends uplink feedback 2540. Uplink feedback 2540 includes CQI feedback vector (Fo, i , D ₀ , _r , D _j i, δD _{j r} , R) respectively representing wideband CQI for codeword CWl; wideband Deltas for codeword index values r=2,...R; subband Deltas for codeword CWl; Delta Deltas for codeword index values r=2,...R; and the rank value R. In FIG. 11, an alternative form of uplink feedback 2540 includes a CQI feedback vector (Fo, i , D ₀ , _r , Ji, J _r ). The elements of this feedback vector respectively represent wideband CQI for codeword CWl; wideband Deltas for codeword index values r=2,...R; joint quantized vector subband Deltas compressed and delivered by step 2530 as a codebook index Ji for codeword CWl. Moreover, step 2535 compresses and

delivers the Delta. Deltas delivered as R- 1 respective codebook indices J _r determined by joint quantizing each of R-I Delta Delta CQI vectors over M subband index values j= 1, 2,...M For the R-I codeword index values r=2,...R. The rank value R is implicit in the UE reporting. eNB counts the number of feedback indices J in the feedback or counts the number of wideband deltas D _0,r therein plus one, or counts all the feedback values and divides by two (by count shifting rightward one), or executes some other suitable counting process. Alternatively, the rank value R is explicitly fed back to eNB.

In FIG. 11, for the base station eNB, operations commence with a BEGIN 2601 and execute a decision step 2605 that determines whether any codebook feedback is involved. In the meantime, feedback vectors are incoming to eNB from numerous UEs indexed i. If Yes at step 2605, operations proceed to a step 2610 and 2614 for codebook accesses. Step 2610 uses codebook index J _1I (Ji from UE i) to retrieve the subband deltas D _j i for UEi for each of the subband index values j= 1, 2,...M for codeword CWl. Step 2614 uses each of the R-I codebook indices J ₁₁ (J _r from UE i) to codebook-retrieve the Delta Delta vector δD _{j r} for UEi for each of the subband index values j= 1, 2, ...M and R- 1 codeword index values r=2,...R, whereupon a step 2618 is reached. In FIG. 11, eNB step 2618 is reached after step 2614 or in case codebook feedback is not applicable at decision step 2605. Step 2618 uses the Delta Deltas and the CWl deltas and recovers the differentials (subband deltas) for each UEi and for all the codewords by a summing process expressed as D _{lj r} = δD _{lj r} + D _1J j for each of the subband index values j= 1, 2, ...M and codeword index values r=2, ...R. A succeeding step 2620 recovers the wideband CQIs F ₁ Q ₁₁ = D ₁ Q ₁₁ + F ₁ Q _1I for codeword index values r=2,...R and all UEs i. Next a step 2625 recovers the original CQIs S _1J11 - for all UEs i by a process expressed as S _{1J 1} . = D _{1J 1} - + F ₁ Q _j for each of the subband index values j= 1, 2,...M and for codeword index values r=l,...R. At this point the CQI information is recovered, and operations proceed to the eNB scheduler 2630 to allocate subbands to UEs followed by a precoder that establishes precoding matrices for the UEs based on the subband allocations. In a step 2640, the base station eNB transmits by downlink to the UEs using composite precoding matrix PM based on the precoding matrices thus established. A variant of FIG. 11 at UE quantizes CQI for a reference subband F _REF,γ for subband J=REF for each codeword r, reference subband Deltas δ _REF , _γ = F _REF , _γ - F _REF . _I for codeword index values r=2,...R; adjacent subband differential vectors for the adjacent subband deltas pairwise expressed as δ _{j r} = S _J>r - S _j _ _1>r for each of the subband index

values j= 2,... M and for codeword index values r=l,...R; and differentials (Delta Deltas) δ δ _{j r} = δ _{j r} - δ _j i for each of the M-I adjacent subband differentials for each of the codeword index values r=2,...R. eNB recovers the reference subband CQIs F _{l REF} , _r = Aj _1REF11 + F _{1 REF1I} for codeword index values r=2,...R and all UEs I; recovers the differentials (adjacent-subband deltas) for each UEi and for all the codewords by a summing process expressed as δ _{lj r} = δ δ _{lj r} + A _{1J 1} for j= 2, ...M and r=2,...R; and reconstructs the original subband CQIs for subbands j>REF by the repeated addition S _1Jir = S _1J - _Ij +δ _{lj r} for r=l,...R and reconstructs the original subband CQIs for subbands j<REF by the repeated subtraction S ₁₀ , _r = S ₁₀₊₁ , _r - δ ₁₀ , _r for codeword index values r= 1 , ... R. A Best-m category of variants of FIG. 11 variously provide reporting for selected and unselected subbands, e.g. Best-m Average CQI or Best-m Individual CQI reporting, with joint quantization or not. The selected subband positions are recorded by setting corresponding bits of subband vector SV(j) =1 and/or by joint quantization using a compressed label with bits. Unselected subbands have SV(J)=O. The mean or median or wideband CQI is designated _s F _r for the selected subbands "s" for each r-th codeword CWl, CW2, etc., and the mean or median or wideband CQI _u F _r for the unselected subbands "u" for each r-th codeword CWl, CW2, etc. Deltas for the selected subbands _s D _r = _s F _r - _s Fi and for the unselected subbands _u δ _r = _u F _r - _u Fi are generated for each codeword CW2, etc. When spatial CQIs are quite similar in magnitude across code words r, such CQI reporting delivers compression because both the spatial CQI differences _s D _r and the spatial CQI differences _u δ _r are small and easily represented with just a few bits. Alternatively, the process generates selected-subband deltas δ _h ι = S _j j - Fo, i and selected-subband deltas expressed by δ _J>r = S _hI - Fo, _r for each codeword CW2, etc., i.e. for selected subbands j indicated by subband vector SV and r = 2,...R. Subband spatial deltas can be generated instead, as represented by D _{j r} = S _J>r - S _J; i, for selected subbands j and code words indexed r=2, ...R. Some embodiments form Delta Deltas by a process represented by δD _{j r} = D _{j r} - D _j j or otherwise for the selected subbands j and r = 2, ...R. In still another variant of the process of FIG. 11 , the CQI report vector has subband vector SV and mean/median/wideband collective CQI _u Fi _, spatial CQI differences u δ _r relative to CQI _u Fi for the unselected subbands for codeword CWl. A horizontal CQI difference _su D _r = _s F _r - _u F _r delivers CQI differences in a given row for each codeword, r = 1,2,... R. Variants of the second embodiment _su D' _r = _s F _(r+ i _)modR - _u F _r for

each codeword, r = 1,2, ...R, also can provide useful correlation in a zig-zag or other scanning pattern.

Instructions for the various processes are suitably stored in whole or in part in flash memory, or volatile or nonvolatile memory on or off chip relative to microprocessor core(s) or other processor block(s). Various embodiments are suitably applied to data modulation or multiple access schemes that utilize some type of frequency-domain multiplexing, such as OFDM/OFDMA, frequency-domain multiple access (FDMA), single-carrier FDMA (SC-FDMA), and multi-carrier code division multiple access (MC- CDMA) and any suitable systems. FIG. 12 is a flow diagram for depicting process embodiments for eNB reconstructing original CQIs S _jr , or intermediately reconstructing CQI differences D _jr , from the subband vector SV generated in Best-m CQI reporting. FIG. 12 provides substeps for CQI reporting processes described elsewhere herein. Operations in FIG. 12 commence with a BEGIN 5105 that sets codeword index r=l, and operations proceed to a step 5110 that initializes indices j=l and L=I. A decision step 5120 is part of a process of scanning subband vector SV. For the A-suffixed FIGS. 6A, 7A, 8A, SV(j) is independent of codeword index r. For the B-suffixed FIGS. 6B, 7B, 8B, SV(j,r) is fed back for each spatial codeword r and used in FIG. 12 . (The dependence on codeword index r is shown in FIG. 12 and may be omitted for embodiments where it is not applicable.) A particular example of a 10- element subband vector SV is shown below the flow. Decision step

5120 determines whether a given subband vector element SV(j,r) is one (1) or not. If not, the full-length (M) CQI, or differential CQI, vector D of FIG. 12 is set equal to zero (0) at element j so that D(j,r)=O at step 5125. (In some Best-m Average embodiments, instead set D(j ,r) equal to the differential encoding _u δ _r or CQI _u F _r at step 5125 for the unselected subbands.) Then a step 5130 increments index j. Then a decision step 5140 determines by the criterion j>j max whether the entire subband vector SV has been scanned in a configured scanning pattern such as in FIGS. 9 and 10. If not, operations loop back to decision step 5120. At decision step 5120, if the given subband vector element SV(j,r) is one, then operations proceed to a step 5160 to access short differential CQI vector DV(L,r) element L and multiply to generate a full-length differential CQI vector element D(j ,r) = SV(j ,r) x DV(L,r). Some embodiments simply use the logic IF SV(j,r)=l, THEN D(j ,r) = DV(L,r) at this point. (Note: D(j,r) and DV(L,r) correspond to and are suitably made to take the position of various CQI S _ljr or S _1J51 designations or differential D or δ

designations used.) If the scan across the short CQI, or differential CQI, vector DV is completed, L = Lmax at a decision step 5170 and operations reach decision step 5185. (Lmax = m, the number of Best-m selected subbands, and is a constant in some embodiments or is a function Lmax(r) of codeword r in some other embodiments.) Otherwise, operations proceed from decision step 5170 to a step 5180 that increments the index L that scans the short differential CQI vector DV. (Both steps 5170 and 5180 are omitted for Best-m Average CQI reporting.) Operations go from step 5180 to step 5130 and the process goes on as already described. If Yes at step 5140 or Yes at step 5170, then step 5185 determines whether codeword index r=Rmax, i.e. equals rank R. In Joint Sub- band Selection in FIGS. 6A, 7A, 8A, Rmax=l and operations at FIG. 12 step 5185 immediately determine that r=Rmax and pass directly to a RETURN 5195. In Independent Sub-band Selection FIGS. 6B, 7B, 8B, operations go from FIG. 12 step 5185 to a step 5190 to increment codeword index r=r+l and loop back to step 5110. In due course, all the codewords r are handled in FIG. 12, and step 5185 determines that codeword index now has r=Rmax, and operations reach the RETURN 5195.

In terms of FIG. 12 , an example of the scanning process of FIG. 9-10 provides the earlier-hereinabove described mapping function F(s) = = (j, r) ahead of the loop kernel 5120, 5125, 5160 that is written in terms of indices (j, r). The mapping function F(s) instantiates the scanning pattern and is retrieved from a stored codebook in memory indexed by the scanning code. The mapping function and the loop kernel are embedded together in a one-index loop on index s. When the loop executes, its operations are performed according to the scanning pattern. The scanning pattern process is similarly applied to any process loop on indices (r,j) to instantiate a configured scanning pattern therein (e.g., FIG. 11). Illustrative embodiments are not to be construed in a limiting sense. It is therefore contemplated that the appended claims and their equivalents cover any such embodiments, modifications, and embodiments as fall within the true scope of the invention.