Adaptive data recovery from distorted signals

Abstract

This application presents an adaptive data recovery from distorted signals (ADRDS) of original data symbols from intervals or parameters of tone signals derived from a received OFDM signal, including responding to dynamic distortions introduced to the received OFDM signal by an OFDM transmission channel. Such ADRDS is implemented by converting back the derived intervals or parameters into original data symbols corresponding to distinctive sets of the intervals or parameters which the derived intervals or parameters belong to.

Claims

1. A method for adaptive data recovery (ADR) from a received orthogonal frequency division multiplexing (OFDM) signal by an adaptive decoding of data symbols from intervals of tone signals of the received OFDM signal wherein the intervals of the tone signals correspond to cycles or half-cycles of the tone signals, wherein the received OFDM signal is produced by a coding circuit implementing a coding of the data symbols into a transmitted OFDM signal and a transmission link altering the transmitted OFDM signal into the received OFDM signal in accordance to a current transfer function; wherein the method for ADR comprises the steps of: producing, using a received signal processor, the intervals of the tone signals from the received OFDM signal; deriving the current transfer function by utilizing parts of the received OFDM signal corresponding to known parts of the transmitted OFDM signal; wherein the current transfer function includes dynamic distortions introduced to the received OFDM signal by time variant changes of characteristics of the transmission link; defining distinctive sets of the intervals of the tone signals; wherein each of the distinctive sets comprises the intervals of the tone signals expected to represent one of the data symbols; preloading the data symbols to a content addressed memory; identifying ones of the distinctive sets comprising the produced intervals by processing the produced intervals; decoding adaptively the data symbols by reversing both the coding of the data symbols and the current transfer function, wherein the reversing of both the coding of the data symbols and the current transfer function is achieved by reading the data symbols from the content addressed memory addressed with identifiers of the identified distinctive sets.

2. A method as claimed in claim 1, wherein the ADR method further comprises: continuous updating of the current transfer function based on an analysis of the parts of the received OFDM signal corresponding to the known parts of the transmitted OFDM signal.

3. A method as claimed in claim 1, wherein the ADR method further comprises: continuous updating of the current transfer function based on an analysis of the received OFDM signal, in order to accommodate a gradual fading of the received OFDM signal.

4. A method as claimed in claim 1, wherein the defining distinctive sets of the intervals of the tone signals comprises: using a known pattern of the coding of the data symbols and the current transfer function of the transmission link.

5. A method for adaptive data recovery (ADR) from a received orthogonal frequency division multiplexing (OFDM) signal by an adaptive decoding of data symbols from parameters of tone signals of the received OFDM signal, wherein the received OFDM signal is produced by a coding circuit implementing a coding of the data symbols into a transmitted OFDM signal and a transmission link altering the transmitted OFDM signal into the received OFDM signal in accordance to a current transfer function; wherein the method for ADR comprises the steps of: producing, using a real time processor, the parameters of the tone signals from the received OFDM signal wherein the produced parameters correspond to amplitudes or phases of the tone signals, wherein operations of the real time processor are controlled by a background processor; deriving the current transfer function by utilizing parts of the received OFDM signal corresponding to known parts of the transmitted OFDM signal; wherein the current transfer function includes dynamic distortions introduced to the received OFDM signal by time variant changes of characteristics of the transmission link; defining distinctive sets of the parameters of the tone signals; wherein each of the distinctive sets comprises the parameters of the tone signals expected to represent one of the data symbols; preloading the data symbols to a content addressed memory; identifying ones of the distinctive sets comprising the produced parameters by processing the produced parameters; decoding adaptively the data symbols by reversing both the coding of the data symbols and the current transfer function, wherein the reversing of both the coding of the data symbols and the current transform function is achieved by reading the data symbols from the content addressed memory addressed with identifiers of the identified distinctive sets.

6. A method as claimed in claim 5, wherein the ADR method further comprises: continuous updating of the current transfer function based on an analysis of the parts of the received OFDM signal corresponding to the known parts of the transmitted OFDM signal.

7. A method as claimed in claim 5, wherein the ADR method further comprises: continuous updating of the current transfer function based on an analysis of the received OFDM signal, in order to accommodate a gradual fading of the received OFDM signal.

8. A method as claimed in claim 5, wherein the defining distinctive sets of the parameters of the tone signals comprises: using a known pattern of the coding of the data symbols and the current transfer function of the transmission link.

9. A method for adaptive data recovery (ADR) from a received orthogonal frequency division multiplexing (OFDM) signal by an adaptive decoding of data symbols from parameters of cycles or half-cycles of tone signals of the received OFDM signal, wherein the received OFDM signal is produced by a coding circuit implementing a coding of the data symbols into a transmitted OFDM signal and a transmission link altering the transmitted OFDM signal into the received OFDM signal in accordance to a current transfer function; wherein the method for ADR comprises the steps of: producing, using a received signal processor, the parameters of the cycles or half-cycles of the tone signals from the received OFDM signal wherein the produced parameters correspond to amplitudes or phases of the cycles or half-cycles of the tone signals; deriving the current transfer function by utilizing parts of the received OFDM signal corresponding to known parts of the transmitted OFDM signal; wherein the current transfer function includes dynamic distortions introduced to the received OFDM signal by time variant changes of characteristics of the transmission link; defining distinctive sets of the parameters of the cycles or half-cycles of the tone signals; wherein each of the distinctive sets comprises the parameters of the cycles or half-cycles of the tone signals expected to represent one of the data symbols; preloading the data symbols to a content addressed memory; identifying ones of the distinctive sets comprising the produced parameters by processing the produced parameters; decoding adaptively the data symbols by reversing both the coding of the data symbols and the current transfer function, wherein the reversing of both the coding of the data symbols and the current transfer function is achieved by reading the data symbols from the content addressed memory addressed with identifiers of the identified distinctive sets.

10. A method as claimed in claim 9, wherein the ADR method further comprises: continuous updating of the current transfer function based on an analysis of the parts of the received OFDM signal corresponding to the known parts of the transmitted OFDM signal.

11. A method as claimed in claim 9, wherein the ADR method further comprises: continuous updating of the current transfer function based on an analysis of the received OFDM signal, in order to accommodate a gradual fading of the received OFDM signal.

12. A method as claimed in claim 9, wherein the defining distinctive sets of the cycles or half-cycles of the parameters of the tone signals comprises: using a known pattern of the coding of the data symbols and the current transfer function of the transmission link.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) 1. Brief Description of the NFIT Drawings

(2) FIG. 1 shows Block Diagram of Inverse Transformation Method in order to introduce major sub-systems defined in FIG. 2-FIG. 11.

(3) FIG. 12A & FIG. 12B define Timing Clocks driving the sub-systems shown in FIG. 2-FIG. 11, wherein:

(4) FIG. 12A shows time slots assigned for sub-clocks driving consecutive processing stages,

(5) FIG. 12B shows sub-clocks driving consecutive bits of circular processing stages.

(6) Note: the time slots define phase displacements assigned to the sub-clocks; wherein such time slots are filled periodically with circular sub-clock pulses, as it is required for driving specific bits of every circuluar register representing a circular processing stage.

(7) The FIG. 2-FIG. 11 are numbered correspondingly to the flow of processed data.

(8) All interconnect signals between these figures have unique names identifying their sources and destinations explained in the Detailed Description utilizing the same names.

(9) Inputs supplied from different drawings are connected at the top or left side and outputs are generated on the bottom due to the top-down or left-right data flow observed generally.

(10) Clocked circuits like registers or flip-flops are drawn with two times thicker lines than combinatorial circuits like arithmometers or selectors.

(11) FIG. 2 shows sampling of DMT signal and correction of it's non-linearity.

(12) FIG. 3 shows comb filtering of DMT signal.

(13) FIG. 4 shows Resonating IIR Filter for 129 Tone (129T/RIF).

(14) FIG. 4A shows Resonating IIR Filter 129.5 SubTone (129.5ST/RIF).

(15) FIG. 5 shows integration & amplitudes registration for half-cycles of 129 Tone.

(16) FIG. 5A shows integration & amplitudes registration for half-cycles of 129.5 Sub-Tone.

(17) FIG. 6 shows phase capturing and initialization of tone processing for 129 Tone/128.5 Sub-Tone/129.5 Sub-Tone, wherein block 8 shows the Frame Samples Counter and MTP Start generator common for Real Time Processors for all Tones/Sub-Tones.

(18) FIG. 7 shows retiming & averaging of positive and negative half-cycles for 129 Tone/128.5 Sub-Tone/129.5 Sub-Tone.

(19) FIG. 8 shows amplitude & phase normalization for 129 Tone/128.5 Sub-Tone/129.5 Sub-Tone.

(20) FIG. 9A-FIG. 9B show accessing noise compensation coefficients for 129 Tone.

(21) FIG. 10 shows using these coefficients for compensating expected noise contributions from the 128.5 Sub-Tone & 129.5 Sub-Tone.

(22) FIG. 11 shows Recovery and Registration of 129T Frame Symbols.

(23) 2. Brief Description of the IST Drawings

(24) FIG. 13A shows data recovery from preprocessed signal subspaces by using processed signal parameters.

(25) FIG. 13B shows data recovery from preprocessed subspaces of OFDM signal by using processed signal parameters (DRPS PSP OFDM).

(26) FIG. 13C shows a comparator of signal interval to reference frame (CSR).

(27) FIG. 13D shows data recovery from preprocessed subspaces of OFDM signal (DRPS OFDM).

(28) FIG. 13E shows data recovery from processed parameters of OFDM signal (DRPP OFDM).

(29) 3. Brief Description of the DDR Drawings

(30) FIG. 14 shows Direct Data Recovery using parameters of sub-band cycles (DDR PSBC).

(31) FIG. 15 shows Adaptive Data Decoder (ADD) for the DDR PSBC.

(32) FIG. 16 shows Direct Data Recovery using sub-bands subspaces (DDR SBS).

(33) FIG. 17 shows Adaptive Data Decoder (ADD) for the DDR SBS.

(34) FIG. 18 shows Direct Data Recovery using parameters of sub-bands (DDR PSB).

(35) 4. Brief Description of the ADD and ADRDS Drawings

(36) FIG. 19 shows calculation and registration of integrals of amplitude gradients for 129 Tone.

(37) FIG. 20 shows selecting reference frames and derivation of Deviation Integrals.

(38) FIG. 21 shows detections of minimums of Deviation Integrals and their utilization for data symbols recovery.

(39) FIG. 22 shows Adaptive Data Decoding from intervals of tone signals.

(40) FIG. 23 shows Adaptive Data Decoding from parameters of tone signals.

DETAILED DESCRIPTION

1. Embodiments of NFIT

(41) The Inverse Transformation Method (ITM) is introduced in FIG. 1 as including subsystems shown in blocks 1-7.

(42) These subsystems enable an efficient low-power processing of high-speed oversampled data is enabled by implementing real-time processing units (RTPs) which use simplified algorithms based on variable coefficients.

(43) These RTPs are controlled by a Programmable Control Unit (PCU) which performs a background processing. This background processing includes implementing adaptive non-linear algorithms which analyze received line signal and intermediate processing results, in order to define such coefficients and to download them to content addressed memories such as the Control Register Set for 129 Tone (mentioned further below as 129T_CRS occurring in FIG. 4, FIG. 6 and FIG. 7).

(44) These memories are accessed by the RTPs implementing said ITM outlined in FIG. 1. These RTPs can be implemented as it is detailed below for 129 Tone of DMT Frame.

(45) The RTPs include doing basic sorting of recovered symbols (introduced in the block 7 of FIG. 1 and detailed in FIG. 11) based on symbols occurrence frequencies and noise levels in surrounding sub-tones or tones, while the PCU comprises doing further analysis of such sorted symbols including use of adaptive statistical methods for finalizing selection of most credible symbols.

(46) Said blocks 1-7 are defined in greater detail in FIG. 2-FIG. 11 and corresponding descriptions, as it is referenced below: block 1 comprising the PAAR Correction, is detailed in FIG. 2 and described further below; block 2 with the diagram of frequency magnitude response of its frequency sampling filters (shown on the right side) is detailed in FIG. 3, FIG. 4, FIG. 4A and described further below; block 3 with the diagram illustrating detection of amplitudes and phases of tones & sub-tones (shown on the right side), is detailed in FIG. 5, FIG. 5A, FIG. 6, FIG. 7 and described further below; block 4 is detailed in FIG. 8; block 5 is detailed in FIG. 9A-FIG. 9B and FIG. 10; blocks 6 and 7 are detailed in FIG. 11.

(47) The embodiments presented herein are based on the assumption listed below: DMT OFDM Frame has frequency 4 kHz. DMT Frame comprises OFDM Tones numbered from 32 to 255 (such OFDM Tones have frequencies equal to Tone_NR4 kHz) The sampling clock 0/Clk (see FIG. 2) is kept in phase with the DMT Frame, and has sampling frequency 4 kHz25516=16.32 Mhz.

(48) The NFIT (see FIG. 1) comprises a correction of Peak to Average Amplitude Ratio (PAAR), which reverses a non-linear line signal distortion caused by a gain limitation of line amplification path when a composition of tones having different frequencies & phases ascends into extreme amplitude levels.

(49) The PAAR correction is explained below.

(50) For Ys=Modulus(A/D sample), Ylt=Linearity Threshold, Cs=Compensation Slope;

(51) Yc=Corrected Sample Modulus is calculated as Yc=f(Ys) function defined below:

(52) If Ys>Ylt; Yc=Ys+Cs (Ys-Ylt).sup.2

(53) else; Yc=Ys.

(54) Since such correcting function Yc=f(Ys) maintains continuity of the derivative of the resulting corrected curve, such transformation maintains a smooth transition between the non-corrected and corrected regions while it reverses non-linearity occurring originally in the corrected region due to the gain limitation.

(55) A detailed implementation of such PAAR correction is shown in FIG. 2, wherein the A/D samples are written into the Stage1 of the Synchronous Circular Processor (SCP) comprising A/D_Buffer0/A/D_Buffer1 driven by the circular sub-clocks 1/Clk0/1/Clk1 accordingly (see FIG. 12B explaining circular sub-clocks applications).

(56) Using 2 buffers having separate processing circuits attached enables two times longer processing times for calculating DMT0/DMT1 values with reversed effects of the gain limitation.

(57) The Linearity Threshold (LinThr(D:0)) is subtracted from the amplitude of the attenuated signal sample (i.e. from the Modulus(A/D_Buffer(Sign,D:0)) and such subtraction result is squared and added to the amplitude of the attenuated sample, in order to reverse said gain attenuation.

(58) Any non-linearity can be reversed smoothly (i.e. without derivatives discontinuity) with any accuracy desired by applying polynomial transformation:
Y.sub.reversed=C.sub.s0Y.sub.s; if . . . Y.sub.s(0,Y.sub.t1]
Y.sub.reversed=C.sub.s0Y.sub.s+C.sub.s1(Y.sub.sY.sub.t1).sup.e1; if . . . Y.sub.s(Y.sub.t1,Y.sub.t2]
Y.sub.reversed=C.sub.s0Y.sub.s+C.sub.s1(Y.sub.sY.sub.t1).sup.e1+C.sub.s2(Y.sub.sY.sub.t2).sup.e2; if . . . Y.sub.s(Y.sub.t2,Y.sub.t3]
Y.sub.reversed=C.sub.s0Y.sub.s+C.sub.s1(Y.sub.sY.sub.t1).sup.e1+C.sub.s2(Y.sub.sY.sub.t2).sup.e2+ . . . +C.sub.sN(Y.sub.SY.sub.tN).sup.eN; if . . . Y.sub.s(Y.sub.t(N1),Y.sub.tN] wherein; C.sub.s0, C.sub.s1, . . . C.sub.sN represent slopes of approximations added at 0, Y.sub.t1, Y.sub.t2, . . . Y.sub.tN non-linearity thresholds.

(59) The implementation and equations shown above illustrate a method for reversal of gain non-linearity and/or signal attenuation, wherein such method comprises: identification of dependency between processed signal attenuation and attenuated signal amplitude; defining approximation thresholds and their approximation slopes and approximation exponents; calculating an exponential component for every said approximation threshold exceeded by an attenuated signal sample, by rising a difference, between the attenuated sample and its approximation threshold, to a power defined by its approximation exponent; calculating an approximation component for every such approximation threshold exceeded by an attenuated signal sample, by multiplying such exponential component by its slope coefficient; addition of such approximation component, calculated for the particular approximation threshold, to the approximation result comprising previous approximation components calculated for previous approximation thresholds exceeded by the attenuated signal sample; wherein by such addition of the approximation components calculated for the approximation thresholds exceeded by the distorted and/or attenuated signal sample, said gain-non-linearity and/or signal attenuation is reversed.

(60) This disclosure includes an implementation of a Finite Impulse Response (FIR) filter with a circularly driven register (i.e. consecutive processed samples are clocked in circularly into the register) connected to circuits processing properly delayed samples supplied by the register. Such register based FIR filter is shown in FIG. 3 wherein the FIR filter is exemplified as the 1-z.sup.511 Comb Filter.

(61) The comb filtering based on 1-z.sup.511 begins when N+1=512 samples initializing a new tone are collected in CFR2(S0:S511), wherein: the first filtered sample S(511) is filtered with the collected already samples S(0)/S(509) delayed 511/2 times accordingly, in order to produce the output CFSO(Sign,E:0) fulfilling the difference equation v(n)=x(n)r.sup.Nx(nN)r.sup.2x(n2); and similarly the second filtered sample S(0) is filtered with the collected already samples S(1)/S(510) delayed 511/2 times accordingly, in order to produce the output CFSE(Sign,E:0) fulfilling the same difference equation.

(62) Said corrected DMT0/DMT1 outputs of the 1.sup.st SCP stage are connected to the Comb Filter Register 2 driven by 512 circular clocks 2/Clk0, 2/Clk3, . . . 2Clk511 in order to enable the 1-z.sup.511 Comb Filter of 512.sup.th order implemented by the 2.sup.nd SCP stage.

(63) Such comb filter has 511 zeros assigning 511 Sub-Bands which can be produced by Frequency Sampling Filters constructed by connecting the output of such Comb Filter to 511 resonating filters defined by the equations:
1/(1e.sup.j2k/511z.sup.1) for k=0,1,2, . . . 510.

(64) Such idea is implemented in more practical way in FIG. 3 where all the details are shown and described (such Frequency Sampling Filtering named as Type IV FSF is explained comprehensively on the pages 311-319 in the book Understanding Digital Signal Processing by Lyons, Ed. 2004).

(65) Consequently even zeros from the range of 64 to 510 correspond to even Sub-Bands 64-510 which are considered as facilitating DMT tones numbered from 32T to 255T, while odd zeros correspond to separating them odd Sub-Bands numbered from 63-511 which are considered as facilitating noise sensing Sub-Tones numbered as 31.5ST, 32.5ST, and 33.5ST to 255.5ST.

(66) Such naming convention of the Tones and Sub-Tones is used further on in this Section text and drawings.

(67) The Comb Filter shown in FIG. 3 uses selection circuits, connected to the circularly driven Comb Filter Register 2 (CFR2), for producing consecutive filtered signal samples.

(68) Another possible implementation can use a shifted CFR2 wherein the DMT0/DMT1 signals are clocked into the same segment S0 of the CFR2 and always the same segments S0/S511 can be used, as providing 511 times delay, for producing Comb Filter Output signal.

(69) This disclosure comprises both; the FIR filter, with the circularly driven filter register, using the selection circuits connected to the register for supplying consecutive signal samples, and the FIR filter, with the shifted filter register, utilizing the shifting of the filter register for supplying consecutive signal samples.

(70) This disclosure includes an implementation of an Infinite Impulse Response (IIR) filter with a circularly driven filter register (i.e consecutive filtered samples are clocked circularly into the register) supplying IIR processing circuits with properly delayed samples. Such IIR filter achieves infinite response characteristic by connecting outputs of such IIR processing circuits back to the inputs of the circularly driven register.

(71) Said IIR Filter with circularly driven register (see FIG. 4), uses selection circuits, connected to the outputs of the Resonator Filter Register (129RFR(S0:S3)), for supplying filter processing circuits which produce consecutive filtered signal samples written back circularly into consecutive samples S0-S3 of 129RFR (S0:S3).

(72) Such circularly driven IIR filter exemplified in FIG. 4, is a resonating filter, having idealistic transfer function (F(z)=1/(1e.sup.j2258/511 z.sup.1)) adjusted into the Type IV FSF (explained in the Lyons book mentioned above) for better stability and performance.

(73) Another possible implementation can use a shifted Resonator Filter Register (RFR(S0:S3)) wherein the input signal from the previous stage and outputs of the Resonator Filter Register supply filter processing circuits which produce filtered sample clocked into the same segment S0 of the RFR(S0:S3).

(74) This disclosure comprises both; the IIR filter, with the circularly driven register, using the selection circuits connected to the register outputs for supplying consecutive processed samples, and the IIR filter, with the shifted register, using shifted register outputs for supplying consecutive processed samples to the filter processing circuits.

(75) The odd/even output of the comb filter CFS0(Sign,E:0)/CFSE(Sign,E:0) re-timed in the Comb Filter Reg.3 (CFR3) produces Resonant Filters Selected Input (RFSI(S,E:0)) which is connected to the multiple resonating Infinite Impulse Response (IIR) Filters designated for specific Tones or Sub-Tones.

(76) Such resonating IIR filter designated for the 129Tone (129T) is shown in FIG. 4, wherein: the reference (from 129T_CRS) indicates that any following constant is provided by its register (belonging to the Control Register Set for 129 Tone), wherein this register is loaded by PCU in order to control operations of the Real Time Processor for 129 Tone (129T_RTP); the coefficient k equals to 2129=258 for the 129 Tone; the resonating IIR filtering begins after the CFR3(S0) is produced after collecting N+1=512 samples in CFR2(S0:S511); the Resonator Filter Register is reset by the signal RESET_RFR(S0:S3) before any new tone IIR filtering begins, and furthermore such IIR filtering of an entire sequence of N+1=512 samples is completed before using resulting RFR outputs for any further signal processing.

(77) Similar resonating IIR filter designated for the 129.5Sub-Tone (129.55T) is shown in FIG. 4A, wherein: the reference (from 129.5ST_CRS) indicates that any following constant is provided by its register (belonging to the Control Register Set for 129.5ST), wherein this register is loaded by PCU in order to control operations of Real Time Processor for 129.5Sub-Tone (129.5ST_RTP); the coefficient k equals to 2129.5=259 for the 129.5Sub-Tone; the resonating IIR filtering begins after the CFR3(S0) is produced after collecting N+1=512 samples in CFR2(S0:S511); the Resonator Filter Register is reset by the signal RESET_RFR(S0:S3) before any new tone IIR filtering begins, and furthermore such IIR filtering of an entire sequence of N+1=512 samples is completed before using resulting RFR outputs for any further signal processing.

(78) This disclosure comprises implementation of integrating and/or averaging time domain filter with a circularly driven register (i.e consecutive processed samples are clocked in circularly into the register) supplying such filter's integrating/summating circuits with a proper set of integrated/summated samples.

(79) Such time domain filter achieves integration/summation over a consecutive set containing a required number of samples, by circular replacing of the first sample of a previous set, stored in the circular register, with a new sample following the last sample of the previous set. Resulting consecutive set of samples on the circular register outputs is supplied to the filter integrating/summating circuit producing filter output.

(80) Such time domain filter is exemplified in FIG. 5 where it is used for integration of 129T Half-Cycles and for detecting phases of such Half-Cycles (HC) ends, wherein an end of the present HC occurs at the beginning of the next opposite HC.

(81) Since the input to such HC integrating filter has already been filtered by the previous stages FSF, such input must have sinusoidal shape. Therefore resulting integral of amplitudes of 129T HC represents filtered indicator of original amplitude of the 129T sinusoid. Such integral is used for the recovery of the original tone amplitude as it explained later on.

(82) Since such time domain filter and all the previous filters belong to the SCP operating in phase with the Tones Frame (DMT Frame), such detected HC phase is used for recovering phase of the originally transmitted HC of 129T.

(83) The outputs of the 129T Resonator Filter Register (129RFR(S0:S3)) are clocked in circularly into the Stage5/129 Half Cycle Register (129HCR(S0:S15)) which comprises 16 samples covering an approximated HC interval.

(84) The outputs of 129HCR are connected to the summating circuits producing an integral of the last 16 samples long sequence (named Next Integer (NI)).

(85) While Next Integral (NI) of amplitudes of HC long interval is calculated and fed to the Integral Register (IR(0:K)); it is also compared with the Previous Integral (PI) kept in Previous Integral Buffer (PIB(0:K)), in order to verify if Half-Cycle end is reached.

(86) Such HC end occurs when NI<PI/NI>PI is detected following positive/negative HC accordingly.

(87) When the end of positive/negative HC is detected, the integral of amplitudes over positive/negative HC is loaded into 129 Positive Amp1. Reg. ((129PAR((0:K))/129 Negative Amp1. Reg. (129NAR(0:K)) by signal 129Ld_PA/FE/129Ld_NA/RE accordingly.

(88) Signals 129Ld_PA/FE/129Ld_NA/RE are generated by DecCLK/IncCLK strobes, only if IncCTR>5/DecCTR>5 condition is met. The purpose of such preconditioning is prevention of oscillations (such as caused by computational instability at small signal amplitudes), by providing histeresis introduced by Inc.Counter(0:2)/Dec.Counter(0:2) for positive/negative HC accordingly.

(89) Said IncCTR>5/DecCTR>5 conditions are possible only when the multi-tone processing inhibition signal MTP_Inh is de-activated after initial 640 sampling periods of every new DMT frame (see FIG. 5 and FIG. 6).

(90) The 642 Decoder (shown in the block 8 in FIG. 6) decodes such 640 samples delay introduced by waiting until 640+2 sampling intervals are counted by Frame Samples Counter (FSC), wherein the additional 2 intervals account for the 2 sampling intervals occurring between the MTP_Inh generation in the SCP stage 4 and its actual application in the SCP stage 6.

(91) In addition to the prevention of IncCTR>5/DecCTR>5 conditions, MTP_Inh signal inhibits any generation of 129Ld_PHC/129Ld_NHC (see FIG. 6), and thus MTP_Inh makes sure that no time domain processing takes place before valid signals are supplied by the frequency domain filters.

(92) Circuit shown in FIG. 5A providing Half Cycles Integration & Amplitudes Registration for 129.5 Tone, performs the same operations as the circuit shown in FIG. 5 described above.

(93) Since SCP operations are driven by clocks and sub-clocks having known phase and frequency relation to DMT Frame, results produced by SCP stages have known phase relations to DMT Frame as well. Therefore such detection of an end of positive/negative HC can be used to detect phase of Tone cycle producing such HC.

(94) As such detection of positive/negative HC end signals detection of falling/rising edge of 129 Tone sinusoid as well, signal 129Ld_PA/FE/129Ld_NA/RE is used in FIG. 6 for capturing phase of such falling/rising edge by capturing 129 Tone Phase in 129 Falling Edge Reg. (129FER(0:5))/129 Rising Edge Reg. (129RER(0:5) accordingly.

(95) This FIG. 6 shows Phase Capturing and Tone Processing Initialization for the 129T/128.5ST/129.5ST, wherein: the reference (from 129T_CRS) indicates that any following constant is provided by its register (belonging to the Control Register Set for 129 Tone), wherein this register is loaded by PCU in order to control operations of Real Time Processor for 129 Tone (129T_RTP); since 129RisingEdgeReg./129FallingEdgeReg. captures the end of a negative/positive half-cycle, it represents phase of the rising/falling edge accordingly of a sinusoid represented by such negative/positive half-cycle.

(96) Such 129 Tone Phase is produced by subtracting 129 Last Cycle Phase Reg. (such 129LCPR(0:13) specifies nr. of sampling intervals corresponding to the beginning of the presently expected cycle of 129 Tone) from Frame Samples Counter (such FSC(0:12) specifies nr. of sampling intervals which past from the beginning of the present DMT Frame).

(97) Consequently such capture of the 129 Tone Phase defines phase of presently detected cycle of 129 Tone measured in number of sampling intervals which occurred between the beginning of the expected cycle (having 0 phase) and the detected 129T cycle.

(98) Content of 129LCPR is derived by comparing if FSC-LCPR=129 Cycle (129 Cycle represents number of sampling intervals expected during consecutive 129Tone cycle), and by loading FSC into LCPR whenever such equality condition is fulfilled.

(99) In order to avoid accumulation of digitization errors during such multiple comparisons (involving fractional numbers expressing expected lengths of 129Tone cycles); a method using fractional bit staffing (described also in public domain) can be applied by adding consecutive bits from Fractional Bits Register (FBR(0:128)) to 129CycleBase(0:4).

(100) These additions provide consecutive values of 129Cycle(0:5) keeping total digitization error below single sampling interval.

(101) SCP combines in-phase processing in frequency domain with in-phase processing in time domain.

(102) Therefore SCP detects time/phase dependence between noise sub-bands and DMT Tones.

(103) Consequently SCP enables estimating and compensating impact of neighbor noise sub-bands and neighbor tones on specific cycles of particular tones.

(104) Such estimates and compensation use data from training session and from adaptive wave-form sampling and screening for identifying noise patterns and for programming compensation and inverse transformation coefficients by PCU.

(105) Such detection of phase relations is facilitated by capturing a falling edge of positive HC of 129T/128.5ST/129.5ST in 129FER(0:5)/128.5FER(0:5)/129.5FER(0:5) by signal 129Ld_PA/FE/128.5Ld_PA/FE/129.5Ld_PA/FE.

(106) Similarly a rising edge of negative HC of 129T/128.5ST/129.5ST is captured in 129RER(0:5)/128.5RER(0:5)/129.5RER(0:5) by signal 129Ld_NA/RE/128.5Ld_NA/RE/129.5Ld_NA/RE.

(107) In order avoid using incomplete HC detected at a beginning of DMT Frame, second appearance of signal LD_PA/FE/LD_NA/RE is required in order to produce signal 129Ld_PHC/129Ld_NHC enabling further processing of 129Tone shown in FIG. 7.

(108) This FIG. 7 shows Retiming & Averaging of Positive and Negative HC for 129T/128.5ST/129.5ST, wherein: the content of 129FER/128.5FER/129.5FER is processed by the Modulo-Cycle Adder of Half-Cycle converting a phase of falling edge into a corresponding phase of rising edge, wherein this phase of rising edge represents phase of sinusoid defined by a positive half-cycle ending at the time instant captured in 129FER/128.5FER/129.5FER.

(109) The 7/Clk shown in FIG. 7 generates single PHC/Clk/CYC/Clk impulse if it detects Ld_PHC/Ld_NHC timed by 6/Clk. Such PHC/Clk re-times 129RER(0:5)/128.5RER(0:5)/129.5RER(0:5) by re-loading them into 129REB(0:5)/128.5REB(0:5)/129.5REB(0:5), which are: averaged with 129FER(0:5)/128.5FER(0:5)/129.5FER(0:5) converted into cycle edge by Modulo-Cycle Addition of Half-Cycle); and re-timed with CYC/Clk loading them into 129AER(0:5)/128.5AER(0:5)/129.5AER(0:5).

(110) The positive amplitude registers 129PAR(0:K)/128.5PAR(0:K)/129.5PAR(0:K) are averaged with 129NAR(0:K)/128.5NAR(0:K)/129.5NAR(0:K) accordingly and loaded into the averaged amplitude registers 129AAR(0:K+1)/128.5AAR(0:K+1)/129.5PAR(0:K+1).

(111) SCP comprises using every positive or negative HC as separate data used for recovering a tone symbol. Such ability of using singular Half-Cycles for data recovery provides a huge data redundancy which facilitates use of statistical methods much more reliable than conventionally used DFT averaging over DMT Frame.

(112) Nevertheless, in order to illustrate implementation having lower power dissipation; SCP exemplified by this embodiment has 7th stage (see FIG. 7) combining amplitudes and phases of positive and negative HC into averages per cycle (which still provide significant redundancy).

(113) The NFIT comprises an inversion of frequency related distortions in a transmission channel (such as DMT link), by applying different normalizing coefficients to different Carrier Frequencies (such as DMT Tones) wherein such normalizing coefficients are adjusted to equalize amplitude and phase distortions of the transmitted Carrier Freq. including distortions introduced by a signal processing applied; such inverse normalization of amplitudes and phases comprises: identification of the frequency related distortions occurring on the Carrier Frequencies (or DMT Tones) by using training sessions or adaptive wave-form sampling/screening controlled by PCU; calculating normalizing coefficients, for such Carrier Frequencies or DMT Tones, by PCU; using such normalizing coefficients, supplied by PCU, by real-time processing unit for equalizing such frequency related distortions in the processed Carrier Freq. or DMT Tones.

(114) Such amplitude and phase normalization for 129T/128.5ST/129.5ST is shown in FIG. 8, wherein it includes normalization of noise sensing Sub-Tones (128.5ST/129.5ST) neighbor to the data carrying 129T.

(115) 129 Tone phase defined by 129 Tone Averaged Edge Register (129AER(0:5)), is normalized by multiplying by the 129T Phase Normalizing coefficient (129PhaNor) and by adding the 129T Phase Adjusting coefficient (129PhaAdj).

(116) Since sinusoidal noise contribution from such neighbor sub-tones is dependent on phase differences between the tone and the sub-tones, such phase differences are normalized by multiplying them by the Phase Normalizing coefficient.

(117) 129T Averaged Amplitude Register (129AAR(0:K+1)) and its 128.5ST/129.5ST counterparts ((128.5AAR(0:K+1)/(129.5AAR(0:K+1)) are normalized by multiplying them by the 129T Amplitude Normalizing coefficient (129AmpNor).

(118) All such normalizing coefficients are taken from the 129 Tone Control Register Set (129T_CRS) which is pre-loaded by PCU implementing adaptive distortion reversing techniques.

(119) While SCP comprises performing signal processing operations which are synchronized by the processed incoming signal, such approach comprises two different synchronization methods specified below and exemplified by the embodiments shown herein.

(120) When SCP stages (such as previous 7 stages) perform processing of belonging to frequency domain DMT Tones (or Multi-Band carriers); they are synchronized by DMT Frame (or channel frame), as such stages are driven by the clocks or sub-clocks synchronous to the sampling clock which is phase locked to DMT Frame (or channel frame).

(121) When SCP stages (such as this 8.sup.th stage and next stages) perform processing of already detected tone (or band) cycles belonging to time domain; they are synchronized by such cycles detection events instead, as such stages are driven by clocks generated when information about cycle detection is passed from a higher level stage to the next level.

(122) Such second synchronization method does not do (discontinues) any further processing if a new cycle of the tone (or band) is not detected.

(123) SCP comprises both synchronization methods defined above.

(124) The cycle detection signal CYC/Clk enables using leading edge of 8/Clk/8 (having frequency 8 times lower than the sampling clock) for the one time activation of AS1/Clk signal which drives all the registers of the SCP 8th stage presented in FIG. 8.

(125) Such AS1/Clk signal remains active (for about 1 sampling period) until the leading edge of the next 9/Clk signal activates the AS1_RST signal (see FIG. 9A).

(126) Such AS1_RST signal enables using leading edge of the next 8/Clk/8 for the one time activation (for about 1 sampling period) of the signal which initiates reading of amplitude and noise compensation coefficients from Memory of Noise Compensation Coefficients (MNCC).

(127) Such timing enables Address Decoders for Memory of Noise Compensation Coeff. (AD_MNCC) to have processing time extended to 8 sampling intervals in order to use normalized amplitudes and phases provided by the previous 8th stage for decoding Address(0:8)/NS_MNCC before AS2/Read_MNCC is activated.

(128) The NFIT comprises an efficient non-linear reversing of transmission channel distortions and non-linear noise compensation in over-sampled signals, by implementing real time processing units (RTPs) using simplified algorithms applying variable coefficients, wherein such RTPs are controlled by the back-ground processing PCU which implements adaptive non-linear algorithms by analyzing received line signal and intermediate RTPs processing results and by defining and down-loading such coefficients to content addressed memories accessed by RTPs such a 129 Tone Control Registers Set (129T_CRS) or Memory of Noise Compensation Coefficients (MMCC).

(129) Such NFIT noise compensation method comprises RTP operations listed below: frequency domain and/or time domain processing of data carrying signal and/or neighbor tones or frequency bands in order to derive estimates of parameters influencing distortion or noise components in the signal, wherein such parameters may include amplitudes and/or phase of data carrying tone or freq. band and/or surrounding noise or interference from neighbor tones or bands; converting such parameters into an effective address of said content addressed memory in order to access coefficients providing most accurate compensation of said channel distortion or noise; applying such coefficients to a sequence of predefined arithmetic and/or logical operations performed on the received signal in order to reverse transmission channel distortions and/or to improve signal to noise ratio.

(130) Such noise compensation method is illustrated in FIG. 9A-FIG. 9B and FIG. 10 showing stage 9.sup.th and 10.sup.th of the SCP embodiment.

(131) It is shown in FIG. 9A that: said noise affecting parameters supplied by 129T/129.5ST Normalized Amplitude Registers (abbreviated to 129NAR(0:P)/129.5NAR(0:P)) and 129.5ST Averaged Phase Difference Register (abbreviated to 129.5APDR(0:L) are used together with the 129 Tone Amplitude Thresholds/Next Sub-tone Amplitude Thresholds and Next-Sub-tone Phase Difference Thresholds, in order to decode address to the Next Sub-tone MNCC (Address(0:8)/NS_MNCC).

(132) It is detailed in FIG. 9A that: said 129 Tone Amplitude Thresholds facilitating use of different coefficients programmed adaptively by PCU, are applied as T/AT1, T/AT2, . . . T/AT6, T/AT7; said Next Sub-tone Amplitude Thresholds facilitating use of different coefficients depending on 129.5ST Amplitude, are applied as NS/AT1, NS/AT2, . . . NS/AT6, NS/AT7; said Next Sub-tone Phase Difference Thresholds facilitating use of different coefficients depending on 129.5 ST Phase Difference, are applied as NS/PDT1, NS/PDT2, . . . NS/PDT6, NS/PDT7; said Address(0:8)/NS_MNCC selects reading & loading of coefficients, compensating expected noise contribution from the 129.5ST, to their registers 129.5Amp1. Addition Reg/129.5 Amp1. Multiplication Reg./129.5 Phase Addition Reg./129.5 Phase Multiplication Register.

(133) Very similar circuits and methods (shown in FIG. 9B) addressing the Previous Sub-tone MNCC (Address(0:8)/PS_MNCC) are applied in order to select & load coefficients, compensating expected noise contribution from the 128.5ST, to their registers 128.5Amp1. Addition Reg/128.5 Amp1. Multiplication Reg./128.5 Phase Addition Reg./128.5 Phase Multiplication Register.

(134) These registers, loaded from NS_MNCC and PS_MNCC, supply coefficients producing estimates of noise compensating components which are added to 129T amplitude and to 129T phase (as shown in FIG. 10), in order to provide compensated amplitude in 129 Compensated Amplitude Register (129CAR(0:P,ERR)) and compensated phase in 129 Compensated Phase Register (129CPR(0:L,ERR)).

(135) Such noise compensating coefficients are derived by PCU based on evaluations of noise patterns occurring in tones frequency region and their contributions to signal noise acquired during training session and supported by adaptive wave-form sampling and screening utilizing wide coverage of almost entire spectrum by Tones and Sub-tones detected with said FSFs.

(136) The NFIT comprises: detecting noise patterns occurring in frequency domain by using frequency domain processing such as Frequency Sampling Filters for noise sensing in a wide continous frequency spectrum incorporating data carrying tones or frequency bands; detecting noise patterns occurring in time domain by using time domain processing for noise sensing over time intervals including tone (or band) reception intervals; using back-ground PCU for analyzing such detected noise patterns and for creating deterministic and random models of occurring noise patterns; using such models of noise patterns for deriving noise compensation coefficients used by the Real Time Processors for improving signal to noise ratios in received data carrying signal; taking advantage of the recovered symbols redundancy (assured by the RTPs time domain processing ability of recovering data symbol from every tone cycle) by applying such noise models for estimating probability of symbols recovered and/or for dismissing symbols accompanied by high noise levels close in time; using such probability estimates and/or dismissals of unreliable symbols for applying statistical methods which are more reliable than conventional DFT averaging of tone signal received.

(137) Such ability of said symbol dismissal, if detected in a vicinity of high noise, is illustrated in FIG. 10, wherein: the comparison is made if the sum of 129T Amplitude Noise Components (128.5/129.5Amp.NoiseComp.) exceeds 129 Maximum Amp1. Noise (abbreviated to 129Max.AmpNoise) pre-loaded to 129T-CTRS by PCU as a total limit on both acceptable compensations from 128.5ST and 129.5ST taken together. if the comparison 128.5ANC+129.5ANC>129MAN, the ERROR bit marking such symbol for dismissal is written to the 129CAR(0:P,E).

(138) Similarly for the 129T Phase Noise Components (128.5/129.5Pha.NoiseComp.): if the comparison 128.5PNC+129.5PNC>129MPN, the ERROR bit marking such symbol for dismissal is written to the 129CAP(0:L,E).

(139) The NFIT comprises a method for recovery of data symbol transmitted by a singular half-cycle/cycle of said DMT or Multiband tone, wherein: an amplitude measure of said singular half-cycle/cycle, such as integral of amplitude over the half-cycle/cycle time period, and a phase measure of the half-cycle/cycle, are applied to a symbol decoder transforming such combination of amplitude and phase measures into a number representing said recovered data symbol.

(140) Such symbol recovery method further comprises: comparing said amplitude measure to predefined amplitude thresholds, in order to decode an amplitude related factor in a recovered symbol definition; comparing said phase measure to predefined phase thresholds, in order to decode a phase related factor in recovered symbol definition; wherein such amplitude and phase comparators produce their parts of a binary address to a content addressed memory storing a counter of half-cycles/cycles detecting said symbol occurrences during said DMT or Multi-band signal frame; wherein such symbols counters memory (SCM) can accommodate different symbols, detected during said DMT or Multi-band frame, varying during the same frame due to said channel distortions and changing in time noise distribution; sorting symbols, carried by singular half-cycles/cycles, detected during said DMT or Multi-band frame, in accordance to their detections numbers and/or distributions; application of statistical methods for selecting data symbol recovered, from said DMT or Multiband tone, such as selection of a symbol having higher detections number in a range outlined with statistical distribution models.

(141) Implementation of such data symbol recovery, is exemplified in FIG. 11. A DMT control registers set (DMT_CRS) programmed adaptively by PCU, supplies said amplitude thresholds (AT1, AT2, AT3, AT4) and said phase thresholds (PT1, PT2, PT3, PT4) to address decoder for symbols counts memory (AD_SCM); wherein: said AT1, AT2, AT3, AT4 (programmed adaptively by PCU) represent Amplitude Thresholds digitizing recovered amplitude; said PT1, PT2, PT3, PT4 (programmed adaptively by PCU) represent Phase Thresholds digitizing recovered phase.

(142) AD_SCM digitizes compensated amplitude/phase provided by CAR(0:P)/CPR(0:L) by comparing them with said amplitude and phase thresholds, in order to produce address ADR(0:3) equal to binary code of symbol detected.

(143) Such ADR(0:3) is applied (as ADR/SCM) to the symbols counts memory (SCM) when the read-write signal (Rd-Wr/SCM) initializes read-write cycle in 129T symbol counts memory (129SCM).

(144) In response to such Rd-Wr/SCM signal said 129SCM provides a content of a symbol counter (129Symb.Count(0:8)) identified by said ADR/SCM.

(145) 129Symb.Count is increased by 1 and is written back to the same symbol location in SCM (as updated counter CNT-UPD(0:8)/SCM), if 129SymAcc=1 (i.e. if both Error Bits CAR(E) and CPR(E) are inactive).

(146) However; 129Symb.Count remains unchanged when it is written back to the same SCM location, if 129SymAcc=0 (i.e. if CAR(E) or CPR(E) is active).

(147) Maximum Count of detections of the same symbol discovered in present 129T, is stored in 129Max.CounterReg. (129MCR(0:8)) which is read by PCU at the end of DMT frame.

(148) Any consecutive updated counter CNT-UPD/SCM (abbreviated as 129SC+1) is compared with such 129Max.CounterReg. (abbreviated as 129MCR).

(149) If (129SC>129MCR)=1; the newly updated counter is loaded to said 129Max.CounterReg., and the address of the newly updated counter (equal to the binary code of the symbol detected) is loaded to 129Max.Cont.Addr.Reg. (129MCAR(0:3) which is read by PCU at the end of DMT frame.

(150) Otherwise if (129SC>129MCR)=0; both 129Max.CounterReg. and 129Max.Cont.Addr.Reg. remain unchanged.

(151) In order to simplify further PCU operations; there is a 129T detected symbols map register (129DSMR(1:16)) which has 16 consecutive bits designated for marking occurrence of the 16 consecutive symbols during DMT frame, wherein particular marking bit is set to 1 if corresponding symbol occurs one or more times. Such 129DSMR(1:16) is read by PCU at the end of DMT frame.

2. Embodiments of IST

(152) The DRPS OFDM described in configuration 6 and shown in FIG. 13D, is designed to compare half-cycles/cycles of tones/sub-bands received from the resonating filters (see FIG. 4) with reference frames having expected sinusoidal contours in order to detect which of such frames is the closest one.

(153) Such comparison can be accomplished using comparator shown in FIG. 13C (explained further below) as producing deviation integrals for continuing sinusoidal outputs produced by the resonating filters.

(154) Averaged values of such deviation integrals can be used to select one of reference frames applied to this tone/sub-band as being the closest one and thus being useful for recovering amplitude related component of the data symbol.

(155) Occurrence of minimum values may indicate phase of the tone/sub-band and thus can be useful for recovering phase related component of the symbol.

(156) Samples of an interval of said received or preprocessed signal may be compared with elements of a reference frame as it is shown in FIG. 13C and explained below.

(157) For a signal interval ending with a sample Sk of a signal, earlier samples S.sub.k+1 of said sample S.sub.k, may be defined by using 1 ranging from 0 to 15 if it is assumed that this interval is 16 samples long.

(158) For such interval a deviation of its sample S.sub.k+1 from its corresponding element R.sub.1 of a reference frame may be calculated as Modulus of (S.sub.k+1-R.sub.1).

(159) Consequently for every such interval, its deviation integral DI.sub.k may be calculated as equal to:

(160) ${\mathrm{DI}}_k=\underset{l=0}{\overset{15}{.Math.}}.Math.S_{k+l}-R_l.Math.$

(161) Estimates of minimums of such deviation integrals may be used to verify if: the interval comprises a data carrying contour (such as an edge of PAM or a half-cycle or cycle of tone or sub-band of OFDM); the frame is close enough to estimate a range of amplitude and phase represented by the contour in order to identify received or preprocessed signal subspace which this contour belongs to, wherein this particular subspace is predefined as carrying specific data by the inverse transformation algorithm.

(162) As it has been indicated in the NFIT embodiment, the direct data recovery may be achieved by using such contours subspace identifiers for addressing Content Addressed Memory, pre-loaded with data implementing said inverse transformation algorithm.

(163) Since this embodiment deviation integrals result from adding positive deviations between single samples and their mask counterparts, minimum values of such integrals indicate edge occurrences.

(164) Such frame (having amplitude and phase attributes assigned to it) can recover specific data symbol encoded originally into a particular half-cycle/cycle by addressing a Content Addressed Memory.

(165) The DRPS PSP OFDM described in configuration 9 and shown in FIG. 13B, is designed to use: amplitude integrals, produced by NFIT circuits for amplitude integration and registration (see FIG. 5) and phases of half-cycles/cycles, supplied by NFIT circuits for phase capturing (see FIG. 6), for selecting reference having amplitudes and phases close enough to limit number of comparisons made simultaneously for high order OFDM codes applied.

(166) Such selection of close frames can be followed by comparisons interpretations and symbol recovery similar to those described above.

(167) The DRPS PSP OFDM configuration shown in FIG. 13E, is optimized (compared to the configuration 12 and another presented in Embodiments of NFIT) by direct addressing of Content Addressed Memory with amplitude integrals and phase captures in order to recover data symbols, instead of comparing these integrals and phase captures with their references (as it's shown in configuration 12) or thresholds (as it's shown in Embodiments of NFIT) before applying the inverse transformation of sub-ranges identified by these comparisons.

3. Embodiments of DDR and ADD

(168) The embodiment of DDR PSBC described in the section 2. Summary of DDR is shown in FIG. 14 and explained below.

(169) The programmable control unit (PCU) reads received signal samples supplied by a waveform screening and capturing circuit (WFSC).

(170) The PCU performs background processing in order to derive and keep updating a relation between said data transmitted originally and said subranges of amplitudes/phases of cycles or half-cycles of sub-band signals.

(171) Such derivation is based on theoretical models of transmission channels and/or training sessions and an analysis of said received signal samples supplied by the WFSC.

(172) The WFSC is described in the sec. 2. Summary of NFIT on page 47 and its configuration with PCU is described in the same section on page 30.

(173) The PCU specifies such relation by: producing said references defining said amplitudes/phases subranges and an assignment of said specific transmitted symbols as corresponding to said specific amplitudes/phases subranges; supplying said assignment of specific transmitted symbols as corresponding to specific subranges of amplitudes/phases detected by the Received Signal Processor (RSP) performing real time processing.

(174) The PCU may also control said real time processing performed by the RSP and modify coefficient used in RSP operations in order to improve their efficiency.

(175) Such RSP implementations are exemplified herein with the circuits described in the sec. 1. Embodiments of NFIT on pages 51 and 53-59 and shown in FIG. 3-FIG. 6.

(176) The Adaptive Data Decoder (ADD) is shown in FIG. 15.

(177) The References Register is loaded from PCU with references defining subranges of parameters (amplitudes/phase of cycles or half-cycles).

(178) Such signal parameters are compared to these references by the Comparator in order to detect which references is this parameter the closest to.

(179) The Comparator uses such closest references to define a binary address of a cell of the Content Addressed Memory (CAM) which contains data symbol loaded to the CAM by the PCU signal assigning data symbols to parameters sub-ranges.

(180) The ADD recovers data symbol corresponding to the signal parameters supplied to it from the RSP, by simply reading this cell from the CAM.

(181) Such DDR configuration can accommodate instantly fast changing characteristics of the received signal and/or transmission channel, since the RSP can produce signal parameters characterizing channel interferences and/or history of a particular signal, in addition to parameters characterizing said particular signal received presently.

(182) Such additional parameters can be compared with their references derived and supplied by PCU, in order to modify said CAM address to one which selects a cell containing appropriate data symbol.

(183) Such additional parameters can be derived to characterize distortions introduced by signals having adjacent frequencies contributed by the environment of a particular sub-band including external noise or other components of received signal.

(184) The frequency sampling configurations disclosed in the RSP's RTP (see pages 51, 53-59) can perform on chip spectrum analysis exemplified therein by detecting and characterizing parameters of intermediate frequencies occurring between data carrying sub-bands.

(185) This application includes using such parameters of intermediate frequencies for reversing effects of distortions caused by a noise sampled at these intermediate frequencies, by applying them to and using by the same ADD-PCU configuration as the parameters related to specific sub-bands only described above.

(186) The embodiment of DDR SBS described in the section 2. Summary of DDR is shown in FIG. 16 and explained below.

(187) This embodiment of DDR SBS preserves the features of DDR PSBC described above with the exception of two structural differences described below.

(188) It uses sub-band signals as referencing signals for identifying referencing subspaces instead of using signal parameters for identifying parameters subranges, in order to address CAM and recover data symbols from it.

(189) Therefore it needs deviation integrals calculators and analyzer and reference frames selector/identifier shown in the ADD presented in FIG. 17.

(190) Such deviation integrals calculators needed for decoding data from OFDM sub-bands have been already described and shown in FIG. 13C of IST.

(191) Such utilization of deviation integrals instead of straight parameters can allow more reliable data recovery which can be more desirable in Base Stations communicating with mobile devices transmitting less powerful signals.

(192) The embodiment of DDR PSB described in the section 2. Summary of DDR is shown in FIG. 18.

(193) It keeps general features of the PSBC but it does not require oversampling or recovery of individual cycles or half-cycles of OFDM sub-bands.

(194) Therefore its applications include FFT based OFDM receivers used commonly as it is shown in FIG. 18.

(195) The embodiments of the ADD systems and circuits presented in FIG. 13C, FIG. 16 and FIG. 17 comprise solutions shown in the FIG. 19. FIG. 20 and FIG. 21; wherein integrals of tones amplitude gradients instead of the integrals of tones amplitudes are calculated and used as tones parameters utilized for identifying the referencing subspaces and recovering data which these subspaces correspond to.

(196) This identification of referencing subspaces and the data recovery are conducted by: using these gradient integrals for selecting reference frames (defining the referencing sub-spaces) which are expected to be the closest to the received tones; calculating integrals of deviations between the received tones and the selected reference frames; detecting minimums of the deviation integrals in order to identify the closest reference frame and the referencing sub-space defined by the closest reference frame; wherein the selection of the reference frames may be accomplished by using the gradient intervals for addressing a Content Addressed Memory (preloaded by the PCU based of the signal analysis) and reading the closest frames; wherein the closest frame may be used for addressing another Content Addressed Memory (preloaded by the PCU as well) in order to read data corresponding to the referencing subspace defined by the closest frame.

(197) It is shown in FIG. 19 that the calculation and registration of the gradients integrals may be implemented with the circuit very similar to that shown in FIG. 5 and described in &3/page 57-&1/page 58.

(198) The only differences are described below.

(199) An integral of gradients around a specific middle signal sample is calculated by: subtracting amplitudes of previous samples from amplitudes of corresponding following samples shifted forward by the same number of sampling periods as the previous samples are trailing the specific middle sample; adding all the subtraction results derived over a half-cycle surrounding the specific middle sample.

(200) In order to implement such algorithm with signals available in FIG. 19 the following equation needs to be implemented:

(201) $\mathrm{Next}\mathrm{Integral}\left(\mathrm{NI}\right)=\underset{k=0}{\overset{k=7}{.Math.}}5/\mathrm{QCR}\left(\mathrm{Sk}\right)-\underset{k=0}{\overset{k=7}{.Math.}}6/\mathrm{QCR}\left(\mathrm{Sk}\right)$

(202) Therefore the half-cycle sub-tone interval is split into two quarter-cycle intervals which are pre-loaded into two different stages 5.sup.th and 6.sup.th instead of being accommodated in the 5.sup.th. stage as previously.

(203) Consequently the registrations of maximum gradient integrals in the 129Positive Grad.Register and minimum gradient integrals in the 129Negative Grad. Register are delayed by one stage and are done in the 7.sup.th stage instead of the 6.sup.th .stage.

(204) FIG. 20 shows selecting reference frames by reading them from the Content Address Memory of Reference Frames (CAM RF), wherein: the maximum and minimum gradients represented by the outputs of the gradient registers named 129PGR(0:K) and 129NGR(0:K) (selected by signals 129Ld_PG/FE and 129Ld_NG/RE accordingly) provide suitable addresses to the CAM RF; the selected reference frame (Read Ref. Frame RRF(0:15)) is read and loaded to selected frame register (Selected Ref. Frame RF(0:15)) synchronously to the tone half-cycle currently inserted into the shift register (ShiftReg SR(0:15)) from the half-cycle register (7/129HalfCycleReg. 7/129HCR(0:15)).

(205) Such synchronous operations essential for providing corresponding arguments to the Deviation Integral Arithmometer are controlled by synchronization circuit generating read request signal when maximum and minimum gradients are detected, wherein the 129Ld_PG/FE and 129LD_NG/RE are activating proper bit in the register 7/129RRR(0:15) prompting the read request signal Read Request and causing the data ready signal Data Ready to appearing by one half-cycle before the half-cycles having these maximum and minimum gradients.

(206) Such by half-cycle forward displacement allows reference frame application to the half-cycles preceding detections of maximums and minimums occurring at the end of half cycles.

(207) The minimum of deviation integrals (see 129MinimumDeviationRegister 129MDR(0:D)) is detected as shown in FIG. 21 with the circuit very similar to that explained in the &3/page 57-&1/page. 58.

(208) These minimums of deviation integrals are utilized for recovering data symbols from tone half-cycles by using the configuration of the Content Address Memory of Data Symbols (CAM DS) and supporting circuits in the configuration very similar to that explained above as used for reading selected reference frames from the CAM RF.

(209) It shall be noted that using such integrals of half-cycles of gradients instead of amplitude has the advantage of eliminating problems with any floating of DC level in OFDM tones.

(210) The above gradient based solution exemplifies more complex method where subranges of gradients (used as tones parameters) are used to select closest reference frames detecting tones subspaces which are inversely transformed to detect original data.

(211) It shall be noted the ADD solutions contributed herein also include simpler solutions wherein: subranges of gradients (used as tones parameters) are subjected directly to said inverse transformation recovering original data without even using said reference frames; or saids reference frames are applied directly to the tones and use of parameters for these frames selection is avoided.

(212) The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein, but is to be accorded the full scope consistent with the claims, wherein reference to an element in the singular is not intended to mean one and only one unless specifically so stated, but rather one or more. All structural and functional equivalents to the elements of the various embodiments described throughout this disclosure that are known or later come to be known to those of ordinary skill in the art are expressly incorporated herein by reference and are intended to be encompassed by the claims. Moreover, nothing disclosed herein is intended to be dedicated to the public regardless of whether such disclosure is explicitly recited in the claims.