Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claims 1-20 are pending and have been examined on the merits.
The Group and/or Art Unit location of your application in the PTO has changed. To aid in correlating any papers for this application, all further correspondence regarding this application should be directed to Group Art Unit 2175.
Specification
The title of the invention is not descriptive. A new title is required that is clearly indicative of the invention to which the claims are directed.
Double Patenting
The Examiner has considered the claims of copending U.S. Application No. 18/980,712 for nonstatutory double patenting (MPEP § 804). Those claims are directed to a different invention (current-sharing control of multiple voltage regulation circuits on a common power line) and do not recite syndrome calculation, syndrome weight, or decoding circuitry. Accordingly, no double patenting rejection is made.
Claim Interpretation
Claims are given their broadest reasonable interpretation consistent with the specification (MPEP § 2111). The following constructions are applied:
“preliminary syndrome calculation circuitry” (claims 7, 12). “Preliminary” is a label; the specification describes it as circuitry that calculates a simplified syndrome from a portion of the codeword and a portion of the parity-check matrix, available at an early stage of codeword processing (Spec. ¶¶ [0017], [0055]-[0057]). The claims require only that this circuitry calculate a syndrome (and, in claim 12, a syndrome weight) from a portion of the data and a portion of the matrix.
“syndrome weight … corresponding to a bit error rate” (claim 1). The specification states that the quantity of “1” bits in a syndrome is the “syndrome weight” (Spec. ¶ [0050]) and that a particular layer’s syndrome weight “may demonstrate a strong correlation” with overall BER (Spec. ¶ [0057]). A syndrome weight that increases with, and is used as an indicator of, the number of bit errors in a codeword therefore “corresponds to” the BER.
Alternative (“or”) limitations. Claims 2, 7, 12-15, 17 and 18 recite “a voltage, a clock frequency, or any combination” or “a clock controller or a voltage regulator, or both.” A prior-art teaching of either alternative meets the limitation.
“power management component” / “decoding component” (claims 7, 12). The specification defines these as hardware circuitry (ASIC, FPGA, state machine, or hardware processing device) physically resident on the controller (Spec. ¶¶ [0026]-[0027]). Under BRI each is a structural circuit performing the recited control function.
Claim Rejections - 35 USC § 112
4. The following is a quotation of the second paragraph of 35 U.S.C. 112:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 3, 6, 10 and 12 are rejected under 35 U.S.C. 112(b) as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor regards as the invention:
Regarding claim 3, the claim recites “determining that a syndrome weight of the one or more syndromes exceeds a first threshold syndrome weight” and, in the next step, “determining that a syndrome weight of the one or more syndromes does not exceed a first threshold syndrome weight.” Claim 1 already recites “a syndrome weight of the one or more calculated syndromes,” and the term “one or more syndromes” differs from “one or more calculated syndromes.” It is unclear whether the syndrome weights of claim 3 are the syndrome weight of claim 1 or different ones, and whether the two recitations of “a first threshold syndrome weight” refer to one threshold or two. Applicant should use “the syndrome weight” and “the first threshold syndrome weight” where the same element is meant.
Regarding claim 6, the claim recites “adjusting the voltage of the voltage regulator.” Claim 5 introduces “a regulated voltage of a voltage regulator” and “the voltage associated with operating the one or more decoding circuitries,” but nothing recites “the voltage of the voltage regulator.” It is unclear whether this refers to the regulated voltage or the voltage associated with operating the decoding circuitries, and the claim then refers separately to “the regulated voltage.”
Regarding claim 10, the claim recites “a filter is coupled between the preliminary syndrome calculation circuitry and the power management,” but “the power management” has no antecedent basis (claim 7 recites “a power management component”). The claim also recites “prevent voltages indicative of corresponding to two signals consecutively received,” and the phrase “indicative of corresponding to” is unintelligible, so the scope of what the filter must prevent cannot be determined. Compare Spec. ¶ [0043] (voltage “reflecting” two consecutively received signals).
Regarding claim 12, the claim ends “associated decoding the data based on the syndrome weight of the syndrome,” which omits a word (apparently “associated with decoding”), and so it is unclear what the clock frequency and voltage are “associated” with. Claims 13-20 depend from claim 12 and are rejected for the same reason to the extent they inherit the defect.
Appropriate correction is required. Because the claims are rejected as indefinite, the prior-art grounds below are based on the claims as best understood.
Claim Objections
Claim 1 is objected to for lacking articles: “utilizing data and parity-check matrix” should read “utilizing data and a parity-check matrix,” and “adjusting at least one of operating parameters” should read “adjusting at least one of the operating parameters” or “at least one operating parameter.” Claim 8 is objected to because “control switching circuitry” should read “control the switching circuitry” (antecedent in the first wherein clause). Claim 12 is objected to because “A apparatus” should read “An apparatus.”
Appropriate correction is required.
Claim Rejections - 35 USC § 103
5. The following is a quotation of 35 U.S.C. § 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art t which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
6. Claims 1 and 2 are rejected under 35 U.S.C. 103 as being unpatentable over Weng (US 10,050,642, patented on August 14, 2018) in view of Bazarsky (US Pub No. 2017/0269991, published on September 212, 2017).
In order to expedite and avoid piecemeal prosecution, the following rejection is made to the extent that the claims are understood, by considering those elements which are understood and interpreting their function in a manner which is consistent with the recited goals of the claims, and then applying the best available art.
The examiner relies on the entire teachings of Weng and Bazarksy references; the applicant should carefully consider the entire teachings of the above-mentioned references to better understand the examiner’s position.
Regarding claim 1, Weng discloses a method of power saving for a low-density parity check (LDPC) decoder (Weng, Abstract; col. 2:26-32; claim 1) comprising:
calculating one or more syndromes utilizing data and parity-check matrix - Weng describes the parity-check matrix H and the syndrome check “H·Cᵀ = S = 0, wherein H is the parity matrix, C is the hard decision codeword and S is the syndrome,” performed as “a hard decision and a syndrome check” during each decoding iteration (col. 1:63-2:5; FIG. 1). In the decoding engine 200 of FIG. 2 the syndrome check is activated through the Early Termination (ET) check circuit 280, which contains the check syndrome weight circuit 260 (col. 3:46-67); and claim 1 recites “at the end of the initial decoding iteration, determining a syndrome weight indicating a number of error bits of the decoded codeword”; and
PNG
media_image1.png
985
675
media_image1.png
Greyscale
adjusting at least one of operating parameters associated with operating one or more decoding circuitries to decode the data based on a syndrome weight of the one or more calculated syndromes - “The check syndrome weight circuit 260 is used for determining the syndrome weight and using the syndrome weight to update the clock frequency. A signal is output from the check syndrome weight circuit 260 to the oscillator 265, which generates an updated clock signal CLK with a frequency according to the syndrome weight” for the LDPC decoding engine 200 (col. 3:60-67; FIG. 2, elements 260, 265, 200). See also “In each iteration, the syndrome weight is first obtained, and a clock frequency is set accordingly” (col. 3:16-17) and Abstract (“using the determined syndrome weight to set an optimal clock frequency for the LDPC decoding”). The clock frequency is an operating parameter of the decoding circuitry, which includes a bit-flipping engine and an N2 decoding engine (col. 2:67-3:3); and
the syndrome weight corresponding to a bit error rate (BER) of the data - Weng teaches that “the syndrome weight will increase on average with the number of error bits” and so “syndrome weight is a good indicator for when power boost is required” (col. 3:11-15), and Weng’s independent claims recite a syndrome weight “indicating a number of error bits of the decoded codeword” (claim 1; claim 6). The number of erroneous bits in a codeword of fixed length is the bit error rate of that codeword, and the art treats the two as interchangeable. See Bazarsky (US 2017/0269991 A1) ¶ [0077] (“the BER of a codeword and the number of errors of a codeword may be interchangeable”) and ¶ [0073] (BER estimate “equal to the syndrome weight of the codeword”). The syndrome weight of Weng therefore corresponds to the BER of the data.
Weng does not use the term “BER.” In the same field of endeavor, Bazarsky teaches that “Because syndrome weights may generally indicate a number of expected bit errors in a data sequence, calculating a syndrome weight can be considered a form of BER estimation” (¶ [0018]), that the BER estimate may be “equal to the syndrome weight of the codeword” (¶ [0073]), and that the BER estimate “corresponds to a syndrome weight calculated based on the hard bit data” (¶ [0052]; see also ¶ [0036]).
PNG
media_image2.png
945
612
media_image2.png
Greyscale
It would have been obvious to a POSITA to describe or treat Weng’s syndrome weight as a BER estimate, because Bazarsky shows this to be the accepted meaning of the quantity Weng already uses, and doing so merely labels a known relationship between the same known elements with predictable results (KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398 (2007); MPEP § 2143(I)(A), (C)). A reasonable expectation of success exists because no change to Weng’s circuit is required.
Regarding claim 2, Weng discloses the method of claim 1, wherein the operating parameters include a voltage, a clock frequency, or any combination thereof, because the operating parameter that is adjusted is the clock frequency of the LDPC decoding engine 200 (col. 3:60-67; col. 2:22-32; FIG. 2, oscillator 265).
7. Claims 3 are rejected under 35 U.S.C. 103 as being unpatentable over Weng in view of Bazarsky and further in view of Zhang (US No. 9,768,807, patented on September 19, 2017).
The examiner relies on the entire teachings of Weng and Bazarksy and Zhang references; the applicant should carefully consider the entire teachings of the above-mentioned references to better understand the examiner’s position.
Regarding claim 3, Weng teaches the method of claim 2 as set out above. Weng further teaches increasing the clock frequency … responsive to determining that a syndrome weight of the one or more syndromes exceeds a first threshold syndrome weight: in each iteration the syndrome weight is obtained and evaluated for an increase over the earlier iteration, and “when the syndrome weight of a specific iteration … indicates a number of error bits of the decoded codeword has increased, the updated clock frequency will be increased accordingly” (claim 1; col. 3:16-33); Weng also identifies error-count levels of about 25 and about 40 bits above which throughput falls and the decoder should be boosted or switched to a more reliable mode (col. 2:11-20; col. 3:1-6). These prior-iteration and error-count levels are “threshold syndrome weights” under BRI.
Weng does not teach decreasing the clock frequency, the voltage, or any combination thereof responsive to determining that a syndrome weight does not exceed a first threshold syndrome weight. In Weng the updated clock frequency is “the same as or higher than” the first clock frequency (claim 1), and the clock returns to its initial value only “at the end of decoding” (col. 3:34-37), not in response to a syndrome weight that does not exceed a threshold.
In the same field of endeavor, Zhang teaches that “[i]f the syndrome weight continuously decreases for a couple of iterations, then most likely the syndrome weight will keep decreasing until becoming zero in the following iterations. This information can be used to predict the number of remaining iterations and scale down the clock frequency and decoder voltage supply” (col. 6:63-7:2), where the syndrome weight “is the number of nonzero syndromes” (col. 6:49-50). Bazarsky teaches comparing a syndrome-weight-based BER estimate 124 with thresholds 131-133 and, when the estimate is below the first threshold, selecting the lowest-power decoding mode (¶¶ [0038]-[0042], [0109]; FIG. 3; see also ¶ [0073]). Zhang thus supplies the decrease of clock frequency and voltage when the syndrome weight is low, and Bazarsky supplies the explicit comparison to a threshold.
PNG
media_image3.png
920
550
media_image3.png
Greyscale
It would have been obvious to a POSITA before the effective filing date to modify Weng so that its oscillator-controlling circuit also lowers the clock frequency (and, per Zhang, the supply voltage) when the syndrome weight does not exceed a threshold. Weng’s stated goal is to “minimize the tradeoff between power consumption and throughput” (col. 2:22-26) and its scheme is an “auto-tuning” of clock frequency (col. 2:59-62); a tuning scheme that only ever raises the clock would leave power wasted on lightly corrupted codewords. Zhang states that the improvement in “power consumption [is] dependent on accurate value of the syndrome weight” (col. 7:2-5) and teaches scaling down on a low or falling weight, and Bazarsky teaches that using a lower-power configuration for low-BER data “may save energy” (¶ [0044]). Combining these is the use of known techniques to improve similar devices in the same way, with predictable results (KSR; MPEP § 2143(I)(C), (G)). A reasonable expectation of success exists because Weng’s oscillator 265 already receives its frequency command from the syndrome-weight circuit 260, so commanding a lower frequency requires no new circuit element.
8. Claims 5 are rejected under 35 U.S.C. 103 as being unpatentable over Weng in view of Bazarsky and further in view of Zhang and further in view of Parry (US No. 10,838,807, patented on November 17, 2020).
The examiner relies on the entire teachings of Weng, Zhang, Bazarksy and Parry references; the applicant should carefully consider the entire teachings of the above-mentioned references to better understand the examiner’s position.
Regarding claim 5, Weng teaches the method of claim 2, including adjusting the clock frequency of the decoding circuitry based on the calculated syndrome weight (col. 3:60-67).
Weng does not teach adjusting a regulated voltage of a voltage regulator to adjust the voltage associated with operating the one or more decoding circuitries based on the calculated syndrome weight; Weng adjusts only the clock frequency.
In the same field of endeavor, Zhang teaches using the syndrome weight to “scale down the clock frequency and decoder voltage supply” (col. 6, l. 63 - col. 7, l. 2), i.e., adjusting the supply voltage of the decoder 512 (FIG. 5) from the syndrome weight. Parry teaches a memory controller that dynamically adjusts the voltage level of the ECC subsystem “between 0.7 and 1.0 Volts,” where “less power is used by using lower voltages and frequencies” (col. 5:49-59; claim 1, col. 16:61-17:6; FIG. 2, step 204), and where “hardware blocks can set thresholds and/or directly regulate voltage and frequency” (col. 5:59-67) and the ECC logic can operate “with variable voltages and clock frequencies” (col. 3:17-28); the regulated supply of the ECC logic is what is adjusted. Parry’s monitored metrics include a bit error rate monitor (col. 3:3-8).
PNG
media_image4.png
439
904
media_image4.png
Greyscale
It would have been obvious to a POSITA to modify Weng by also adjusting the regulated supply voltage of the decoder, as Zhang and Parry teach, in addition to the clock. A lower clock frequency permits a lower supply voltage, and lowering both reduces dynamic power more than lowering the clock alone; Parry says so directly (“less power is used by using lower voltages and frequencies,” col. 5:56-59) and Zhang pairs the two in the same sentence as the syndrome-weight criterion. The modification applies a known power-saving technique (voltage scaling alongside frequency scaling, i.e., DVFS) to a known decoder control scheme with predictable results (KSR; MPEP § 2143(I)(A), (G)), and a POSITA would have expected success because voltage regulators with adjustable output are conventional and Parry already applies them to ECC logic.
9. Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Bazarsky in view of Weng.
The examiner relies on the entire teachings of Weng and Bazarksy references; the applicant should carefully consider the entire teachings of the above-mentioned references to better understand the examiner’s position.
Regarding claim 7, Bazarsky discloses an apparatus (data storage device 110, FIGS. 1 and 8) comprising: preliminary syndrome calculation circuitry configured to calculate a syndrome utilizing data and a portion of a parity-check matrix - the BER estimator 123 in memory interface 121 of controller 120 determines a BER estimate 124 for each sub code 140-143 “‘on-the-fly’ based on the hard bits as the hard bits are received, and prior to the hard bits being provided to the ECC decoder 125 for decoding” (¶ [0036]; FIG. 1); the estimate corresponds to a syndrome weight calculated as S = H·V and SW = Σ sᵢ, where H is the parity-check matrix and V is the data received from the memory (¶¶ [0036]-[0037], [0052], [0073]). Because the calculation is made on an individual sub code - a portion of ECC codeword 116 (¶¶ [0027], [0034]) protected by its own parity bits (¶ [0028]) - it uses the parity-check relations covering that portion, not the whole codeword-level matrix (¶ [0035]).
Bazarsky further discloses a power management component configured to control … a decoding component to decode the data based on a syndrome weight of the syndrome calculated at the preliminary syndrome calculation circuitry: data path management unit (DPMU) 130 “configured to control the overall flow of data and sequence of operations” (¶ [0031]) receives BER estimate 124 from estimator 123 (FIG. 8; ¶ [0072]-[0073]) and “may use the BER estimate 124 … to select an initial decoding mode 134 of the ECC decoder 125 … and may instruct the ECC decoder 125 to initiate decoding” (¶ [0040]; ¶¶ [0038]-[0043]; FIG. 3), among ultra-low-power, low-power and full-power modes of the ECC decoder 125 (decoding component) that differ in power consumption (¶¶ [0019], [0034], [0044], [0063]).
Bazarsky does not teach that the power management component controls a clock controller or a voltage regulator, or both, to respectively adjust a clock frequency, a voltage, or both, associated with operating the decoding component. Bazarsky changes decoder power by choosing a decoding algorithm and message resolution, not by changing the clock or supply provided to the decoder.
In the same field of endeavor, Weng teaches an LDPC decoder in which a check syndrome weight circuit 260 determines the syndrome weight and “[a] signal is output from the check syndrome weight circuit 260 to the oscillator 265, which generates an updated clock signal CLK with a frequency according to the syndrome weight” for the decoding engine 200 (col. 3:60-67; col. 2:36-45; FIG. 2). Oscillator 265 is a clock controller that sets the clock frequency of the decoder, and syndrome-weight circuit 260, with the processor circuit that controls the decoder (col. 2:40-45), is the power management function that commands it.
It would have been obvious to a POSITA before the effective filing date to modify Bazarsky’s DPMU 130 so that, in addition to selecting the decoding mode from the syndrome-weight BER estimate, it uses that same estimate to command the frequency of the clock supplied to the ECC decoder 125, as Weng teaches. Both references use the syndrome weight as an indicator of the number of bit errors in an LDPC codeword (Bazarsky ¶ [0018]; Weng col. 3:7-15), and both aim at saving power without losing throughput (Bazarsky ¶¶ [0044], [0063]; Weng col. 2:22-26). Bazarsky’s estimate is available “prior to the hard bits being provided to the ECC decoder” (¶ [0036]), which gives the clock time to be retuned before decoding starts, and Weng’s clock tuning supplies the throughput-preserving boost that Bazarsky’s higher-power modes need. The combination applies a known technique (syndrome-weight-driven clock tuning) to a known system (syndrome-weight-driven decoder-mode selection) to obtain a predictable power/throughput improvement (KSR; MPEP § 2143(I)(A), (D), (G)). A reasonable expectation of success exists because Weng’s oscillator control needs only a syndrome-weight value as input, and Bazarsky already produces one.
Regarding claim 11, Bazarsky as applied to claim 7 teaches the apparatus of claim 7, wherein the decoding component comprises a plurality of decoding circuitries respectively configured to decode data - ECC decoder 125 supports ULP mode 126, LP mode 127 and FP mode 128 (FIG. 1; ¶ [0034]), and in the FIG. 11 embodiment a first decoder 1120 and a second decoder 1122 (¶¶ [0088]-[0089]) - and wherein the syndrome weight of the syndrome is indicative of which decoding circuitry among the plurality is to decode the data - the syndrome-weight BER estimate 124 is compared with thresholds 131-133 to choose the decoder/mode that will decode the codeword or sub code (¶¶ [0040]-[0043], [0085], [0104]; FIG. 3). Bazarsky does not teach the clock/voltage control of claim 7, which Weng supplies for the reasons and with the motivation stated for claim 7, so no further gap exists for this claim.
10. Claims 9 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Bazarsky in view of Weng, Zhang and Parry.
The examiner relies on the entire teachings of Weng, Zhang, Bazarksy and Parry references; the applicant should carefully consider the entire teachings of the above-mentioned references to better understand the examiner’s position.
Regarding claim 9, Bazarsky as modified by Weng teaches the apparatus of claim 7, in which the power management function signals the clock source according to the syndrome weight (Weng col. 3:60-67).
Bazarsky and Weng do not teach that the power management component provides one or more signals to the voltage regulator to cause the voltage regulator to increase or decrease the voltage of the voltage regulator.
In the same field of endeavor, Zhang teaches scaling the “decoder voltage supply” from the syndrome weight (col. 6:63-7:2), and Parry teaches that a controller dynamically adjusts the voltage of the ECC subsystem between 0.7 V and 1.0 V, with “hardware blocks” that “directly regulate voltage and frequency,” lowering voltage and frequency to save power and raising them on demand (col. 5:49-67; col. 2:36-47; claim 1). Raising or lowering the regulated voltage on command from the controlling circuit is the voltage-regulator control the claim recites.
It would have been obvious to extend the Bazarsky/Weng control path to the supply voltage, for the same reasons given for claim 5: frequency and voltage are scaled together in DVFS, Zhang expressly pairs them with a syndrome-weight criterion, and Parry shows that a controller can command regulator output up or down for ECC logic. The result is the predictable additional power saving of voltage scaling (KSR; MPEP § 2143(I)(A), (G)), with a reasonable expectation of success.
Regarding claim 15, the combination of claim 14 teaches increasing above one threshold and decreasing at another, which leaves a band between them. Bazarsky, Weng and Zhang do not teach controlling the voltage regulator to maintain its voltage at a predetermined voltage level in that band. Weng teaches that in subsequent iterations the syndrome weight is used “to maintain or increase the clock frequency” (col. 3:19-22), i.e., holding the setting when no change is called for, and Parry teaches that the ECC voltage is dynamically adjusted within a defined range (0.7-1.0 V, col. 5:49-52) using “scaling factors, thresholds, hysteresis” (col. 3:8-11) - hysteresis meaning that the output is held between the trigger thresholds. It would have been obvious to hold the regulator output at its current predetermined level when the syndrome weight lies between the two thresholds, because it avoids needless regulator transitions, which Parry’s hysteresis is designed to do, and it is the conventional behavior of a thresholded control loop (KSR; MPEP § 2143(I)(D), (G)), with a reasonable expectation of success.
11. Claims 12-14, 16-19 are rejected under 35 U.S.C. 103 as being unpatentable over Bazarsky in view of Weng and further in view of Zhang.
The examiner relies on the entire teachings of Weng, Zhang, Bazarksy references; the applicant should carefully consider the entire teachings of the above-mentioned references to better understand the examiner’s position.
Regarding claim 12, Bazarsky discloses an apparatus comprising: a power management component (DPMU 130, as mapped for claim 7; FIG. 1; ¶¶ [0031], [0040]); and preliminary syndrome calculation circuitry (BER estimator 123 in memory interface 121) configured to: calculate a syndrome utilizing a portion of data corresponding to a codeword - the hard bits of one sub code of ECC codeword 116, taken on-the-fly before decoding (¶¶ [0027], [0036]) - and a portion of a plurality of sets of bit patterns aligned in a first orientation in a parity-check matrix - the parity checks of H that apply to that sub code (¶¶ [0035], [0037]; a row of H is a bit pattern aligned in a first orientation, Spec. ¶ [0054]); determine a syndrome weight of the calculated syndrome - SW = Σ sᵢ (¶ [0037]; ¶ [0073]); and provide a signal indicative of the syndrome weight to the power management component - BER estimate 124, equal to or a function of the syndrome weight, is provided from estimator 123 to DPMU 130 (FIG. 8; ¶¶ [0038], [0072]-[0073]).
Bazarsky does not teach wherein the power management component is configured to control a clock controller or a voltage regulator, or both, to respectively adjust a clock frequency, a voltage, or both, associated [with] decoding the data based on the syndrome weight, and it does not state in terms which rows (bit patterns aligned in the first orientation) of the parity-check matrix the sub-code syndrome uses.
In the same field of endeavor, Weng teaches the clock control: syndrome-weight circuit 260 signals oscillator 265 to generate a clock with a frequency according to the syndrome weight for the decoding engine 200 (col. 3:60-67; FIG. 2). Zhang teaches computing syndromes and syndrome weight from a portion of the hard-decision bits and a block of the parity-check matrix at a time: in a partial-parallel QC-LDPC decoder “the hard decisions for one or a few blocks of Z columns are updated at a time” (col. 9:6-12), the syndrome and syndrome weight are generated “on-the-fly” from “a portion of the hard decision bit values … without first waiting to complete generation of all updated hard decision bit values” (col. 14:5-14), and the shift amount is determined “based on which block row and block column of the parity check matrix 169 are being processed” (col. 13:45-52; FIGS. 1-2), i.e., each step works on a portion of the rows of H.
The motivation to combine Weng with Bazarsky is that given for claim 7.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to compute Bazarsky’s preliminary sub-code syndrome from only the rows (block rows) of H that cover the sub code and only the corresponding data bits, as Zhang’s partial-parallel architecture does, because Zhang teaches that processing a portion of the matrix at a time yields syndrome weights early, “without first waiting” for the full codeword (col. 14:5-14), which is exactly the early, low-latency estimate Bazarsky wants (¶ [0036]). This is a choice among a small number of known ways to compute a syndrome weight with predictable results (KSR; MPEP § 2143(I)(E)), and a POSITA would have expected success because Zhang’s architecture is a form of the LDPC syndrome calculation Bazarsky already performs.
Regarding claim 13, Bazarsky as applied to claim 12 compares the syndrome-weight-based BER estimate with a first threshold and selects the higher-capability FP mode when it is at or above the threshold and the lowest-power ULP mode when it is below (¶¶ [0040]-[0043]; FIG. 3; ¶ [0109]).
Bazarsky does not teach increasing or decreasing the clock frequency or voltage in response to that comparison. Weng teaches “increas[ing] the clock frequency” when the syndrome weight indicates more error bits and a power boost is required (col. 3:1-6, 11-33; claim 4), and Zhang teaches scaling down the clock frequency and decoder voltage supply when the syndrome weight is low or falling (col. 6:63-7:2). It would have been obvious to apply Weng’s increase when the syndrome weight exceeds Bazarsky’s threshold, and Zhang’s decrease when it does not, for the reasons given for claims 3 and 7: both directions of adjustment are needed to hold throughput when errors are many and to avoid wasted power when they are few (Weng col. 2:22-26; Zhang col. 7:2-5; Bazarsky ¶ [0044]), with predictable results and a reasonable expectation of success (KSR; MPEP § 2143(I)(A), (G)).
Regarding claim 14, Bazarsky teaches several thresholds - first, second and third thresholds 131-133 - against which the syndrome-weight BER estimate is compared, giving a low band (ULP), middle bands (LP, FP) and a high band (soft bits and FP) (¶¶ [0039]-[0043]; FIG. 3).
Bazarsky does not teach clock/voltage adjustment. Weng supplies the increase of clock frequency when the weight is high (col. 3:1-6, 23-33; Weng also distinguishes error-count levels of about 25 and about 40 bits, col. 2:11-20), and Zhang supplies the decrease of clock frequency and voltage when the weight is low (col. 6:63-7:2). Using Bazarsky’s high threshold to trigger the increase and its lower threshold to trigger the decrease is the same combination and rationale as claim 13; Bazarsky’s multiple thresholds are the natural place to set different trigger points for the two directions so that the adjustment does not oscillate, a predictable design choice (KSR; MPEP § 2143(I)(E)).
Regarding claim 16, Bazarsky as applied to claim 12 teaches the apparatus of claim 12 further comprising a plurality of decoding circuitries having a first decoding circuitry and a second decoding circuitry capable of correcting errors on data having a higher BER than the first - ECC decoder 125 with ULP/LP/FP modes (¶ [0034]) and, in the FIG. 11 embodiment, first decoder 1120 and second decoder 1122 (¶¶ [0088]-[0089]), where “the second mode 822 may have an increased error correction capability as compared to the first mode 820” (¶ [0074]) and FP mode is selected for codewords with a higher BER estimate (¶¶ [0042]-[0043]). Weng likewise teaches a bit-flipping engine and an N2 engine that is “much more reliable” above about 40 error bits (col. 2:67-3:3). No gap beyond that addressed for claim 12 exists for this claim; the combination and motivation are those of claim 12.
Regarding claim 17, Bazarsky as applied to claim 16 teaches that a high syndrome-weight BER estimate indicates that the data are to be decoded in the FP mode (second decoding circuitry) (¶¶ [0042]-[0043]).
Bazarsky does not teach increasing the clock frequency or voltage in response to that indication. Weng teaches that “changing the mode requires a power boost, which means the clock frequency should be increased,” and that the clock is raised when the decoder switches from bit-flipping to the more reliable hard-decision soft decoding (col. 3:1-6, 25-33; claims 4 and 9; col. 4:1-10). It would have been obvious to raise the clock frequency when Bazarsky’s weight indicates the second (more capable) decoder, because Weng teaches that the more reliable mode needs the boost to “prevent unstable speed” and keep throughput (col. 4:5-10) and Bazarsky’s FP mode is the slower, higher-power one (¶¶ [0019], [0063]). The result is predictable (KSR; MPEP § 2143(I)(A), (G)), with a reasonable expectation of success.
Regarding claim 18, Bazarsky as applied to claim 16 teaches that a low syndrome-weight BER estimate indicates that the data are to be decoded in the ULP mode (first decoding circuitry) (¶ [0040]; FIG. 3).
Bazarsky does not teach decreasing the clock frequency or voltage in response to that indication. Zhang teaches scaling down the clock frequency and decoder voltage supply when the syndrome weight is low or falling (col. 6:63-7:2), and Weng shows that the clock is returned from its boosted value to the initial value after decoding (col. 3:34-37), i.e., a lower clock is the normal setting for easy data. It would have been obvious to lower the clock frequency and voltage whenever Bazarsky’s weight selects the efficient first decoder, because the efficient decoder does not need the speed or supply that the reliable decoder needs and lowering them saves power (Bazarsky ¶ [0044]; Zhang col. 7:2-5; Parry col. 5:56-59). The result is predictable (KSR; MPEP § 2143(I)(A), (G)), with a reasonable expectation of success.
Regarding claim 19, Bazarsky as applied to claim 12 teaches a parity-check matrix H whose rows are parity-check equations (¶ [0037]).
Bazarsky and Weng do not teach that each set of bit patterns (row) corresponds to a respective layer, and each layer stores one or more numerical values indicative of a quantity of bits by which the codeword or a corresponding syndrome is to shift.
In the same field of endeavor, Zhang teaches a QC-LDPC code whose H matrix consists of “rows and columns of square submatrices … that are either zero or cyclically shifted identity matrices” (col. 7:5-9), “M rows of submatrices of dimension Z×Z” (col. 9:22-24) - each block row being a layer - in which each non-zero submatrix has an offset by which the syndromes or hard decisions must be cyclically shifted (col. 9:25-39; col. 12:14-21), the shift amount “based on a portion of a parity check matrix (H) 169” and “determined … based on which block row and block column of the parity check matrix 169 are being processed” (col. 10:5-14; col. 13:45-52; FIG. 2, shift 206 of k bits; FIG. 1, matrix 169 in ECC engine 136). It would have been obvious to implement Bazarsky’s and Weng’s LDPC parity-check matrix as the quasi-cyclic matrix of Zhang, with a block row (layer) holding shift values, because QC-LDPC is the standard structure for flash LDPC decoders and Zhang teaches it enables efficient partial-parallel decoding with improved throughput and area (col. 7:5-12); Weng’s decoder already includes a permutator 270 that performs such shifts (col. 3:55-58; FIG. 2). The substitution is of one known code structure for another with predictable results (KSR; MPEP § 2143(I)(B), (G)), with a reasonable expectation of success.
Examiner's note:
Closest Issues and Potential Allowable Subject Matter
The Examiner notes the following so that Applicant can see where the rejections are firm and where they depend on a contested reading.
Claims 7 and 12 (“preliminary” circuitry; “portion” of the matrix). Bazarsky’s estimator 123 works on a sub code before decoding (¶ [0036]), but Bazarsky does not say which rows of H it uses. For claim 7 this is read as the parity checks covering the sub code; for claim 12, which requires a portion of the “sets of bit patterns aligned in a first orientation,” Zhang (col. 9:6-12; col. 13:45-52; col. 14:5-14) is added. If the claim is amended to require that the preliminary calculation use a single layer, or fewer than all layers, of H and that it complete before the decoder starts, the pre-decode timing is still in Bazarsky ¶ [0036] and Zhang col. 14, but the single-layer, zero-shift selection is not.
“Power management component.” It is mapped to Bazarsky’s DPMU 130 as modified, and to Weng’s circuit 260 with the processor circuit that drives oscillator 265. If Applicant argues the DPMU is not a “power management” component, Weng alone supplies a circuit that manages decoder power by clock control, and the claims recite the component only by its function.
Claims 4 and 8. The selection of one regulator among several is not taught. The Examiner has not taken Official Notice that hybrid LDO/switching-regulator selection is known, because no evidence is of record; a further search on that point could change the indication.
Claims 6 and 10. No reference of record limits the voltage step or smooths successive control signals. A further search of DVFS slew-rate and filtering art is recommended before allowance.
Claim 20. This is the narrowest indication. A rejection over Bazarsky, Weng and Zhang is conceivable, because Zhang teaches that a zero shift needs no shifting (col. 10:10-14; col. 14:56-63) and that barrel shifters lengthen the critical path (col. 9:17-21; col. 12:1-6); the Examiner has not made it because Zhang never selects a zero-shift portion for a preliminary syndrome.
Inventive core. The Specification’s distinguishing feature is a preliminary syndrome computed from one layer of H that needs no barrel shifting, giving an early syndrome weight with a strong BER correlation that drives power adjustment (Spec. ¶¶ [0017], [0057]-[0058], [0070], [0073]). Claims 12, 19 and 20 together come closest to that feature, and claim 20 is the only claim that reaches it.
Examiner has cited particular columns and line numbers in the references applied to the claims above for the convenience of the Applicant. Although the specified citations are representative of the teachings of the art and are applied to specific limitations within the individual claim, other passages and figures may apply as well. It is respectfully requested from the Applicant in preparing responses, to fully consider the references in entirety as potentially teaching all or part of the claimed invention, as well as the context of the passages as taught by the prior art or disclosed by the Examiner.
Allowable Subject Matter
12. Claims 4, 8, 20 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
13. Claims 6 and 10 would be allowable if rewritten to overcome the rejection(s) under 35 U.S.C. 112, 2nd paragraph, set forth in this Office action and to include all of the limitations of the base claim and any intervening claims.
14. The following is an Examiner's statement of reasons for the indication of allowable subject matter: Claims 4, 6, 8, 10, 20 are allowable over the prior art of record because the prior arts, cited in its entirety, or in combination, do not teach
Claims 4 and 8. None of Weng, Bazarsky, Zhang or Parry, alone or in combination, teaches or fairly suggests selectively operating one voltage regulator among a plurality of voltage regulators to supply the decoding circuitry based on the calculated syndrome weight (claim 4), or controlling switching circuitry to allow one of a plurality of regulators to provide power while the other regulators are prevented from providing power, based on the syndrome weight of the preliminary syndrome (claim 8). Parry’s per-channel and generation-versus-checking voltages (col. 3:24-28; claims 14, 16) set different voltages for different logic; they do not select one of several regulators for the same decoder in response to a syndrome weight.
Claims 6 and 10. None of the applied references teaches or fairly suggests preventing the regulated voltage from being adjusted by more than a threshold amount while adjusting it based on the syndrome weight (claim 6), or a filter between the preliminary syndrome calculation circuitry and the power management component that smooths the syndrome-weight signals so that the voltages they indicate do not differ by more than a threshold for two consecutively received signals (claim 10). The closest teachings, Parry’s “hysteresis” (col. 3:8-11) and “glitch-less clocking to support dynamic voltage changes” (col. 3:21-24; Example 13), do not limit the size of a voltage step or smooth successive inputs.
Claim 20. None of the applied references teaches or fairly suggests that the portion of the parity-check matrix used by the preliminary syndrome calculation circuitry corresponds to a particular layer storing shift values of zero. Zhang discloses that a shift amount may be zero so that no bit shift is performed (col. 10:10-14; col. 14:56-63) but does not select the zero-shift portion for an early, preliminary syndrome that drives power management. Support: Spec. ¶ [0070] (layer 432-8, all shift values “0,” FIG. 4, so no barrel shifters are needed) and ¶ [0073] (strong correlation with BER).
Conclusion
15. Claims 1-3, 5-7, 9-19 are rejected. Claims 4, 8, 20 are objected.
16. The prior arts made of record and not relied upon are considered pertinent to applicant's disclosure.
US 11,184,030 B2 (Samsung) - selects a high-level or low-level ECC decoder by comparing a syndrome weight with a reference value; pertinent to claims 11 and 16-18 (no clock/voltage control).
US 12,132,499 B2 (Dell) - uses syndrome weights from several hard decodes to choose the next error-correction stage in a storage device; pertinent to syndrome-weight-driven decoder staging.
Ahn, Park and Chung, EURASIP J. Wireless Commun. & Networking 2013:255 - DVFS of an LDPC decoder with multiple voltage/frequency operating points chosen from an SNR estimate; pertinent to claims 2, 5 and 9 (SNR, not syndrome weight, drives the choice).
US 12,301,254 B2 - early stopping of bit-flip LDPC decoding based on syndrome weight (title/abstract only).
US 10,419,024 B2 - early termination of LDPC decoding (title/abstract only).
US 9,136,877 B1 - syndrome layered decoding for LDPC codes; pertinent to claims 19-20 (title/abstract only).
US 8,127,209 B1 - QC-LDPC decoder with list-syndrome decoding (title/abstract only).
US 10,892,777 B2 / US 2020/0252079 A1 - fast error recovery with ECC syndrome weight assist (title/abstract only).
US 2005/0062345 A1 - dynamic control of power-converter output-voltage slew rate, and US 2015/0022272 A1 - closed-loop DVFS; candidates for claims 6 and 10 (title/abstract only; not yet reviewed against the claim language).
17. Any inquiry concerning this communication or earlier communications from the examiner should be directed to examiner Raymond Phan, whose telephone number is (571) 272-3630. The examiner can normally be reached on Monday-Friday from 6:30AM- 3:00PM. The Group Fax No. (571) 273-8300.
Communications via Internet e-mail regarding this application, other than those under 35 U.S.C. 132 or which otherwise require a signature, may be used by the applicant and should be addressed to [raymond.phan@uspto.gov].
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Andrew Jung can be reached at (571) 270-3779. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
All Internet e-mail communications will be made of record in the application file. PTO employees do not engage in Internet communications where there exists a possibility that sensitive information could be identified or exchanged unless the record includes a properly signed express waiver of the confidentiality requirements of 35 U.S.C. 122. This is more clearly set forth in the Interim Internet Usage Policy published in the Official Gazette of the Patent and Trademark on February 25, 1997 at 1195 OG 89.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see hop://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
Any inquiry of a general nature or relating to the status of this application should be directed to the TC 2100 central telephone number is (571) 272-2100.
/RAYMOND N PHAN/
Primary Examiner, Art Unit 2175