From Bit to Block: Capacity Achievement via Code Concatenation

Bin Zhang

From Bit to Block: Capacity Achievement via Code Concatenation

T0 review · 0 major / 4 minor · reviewed 2026-08-02 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Forney's concatenation can turn bit-level reliability into block-level reliability on any binary memoryless symmetric channel, without sacrificing rate.

desk verdict A clean and correct reduction: Forney concatenation upgrades bit-level to block-level reliability under a large-deviation condition, with BCH outer codes requiring epsilon=o(1/log log n); the main caveat is that the paper leaves the inner-code hypothesis uninstantiated. read the letter →

arxiv 2607.05816 v2 pith:G32FT6EX submitted 2026-07-07 cs.IT math.IT

Bin Zhang This is my paper · ORCID

classification cs.ITmath.IT

keywords capacity-achievingcodesbinarymemorylesssymmetricchannelsconcatenatedbit-errorprobabilityblock-errorForneyconcatenationBCHlargedeviations

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proves that, on any binary memoryless symmetric channel, a code whose per-bit error probability is small can be upgraded—by Forney's concatenation with a high-rate outer code—into a code whose whole-block error probability is small, without changing the rate. The proof identifies a precise sufficient condition: the outer code must correct a fraction δ of errors larger than the inner bit-error fraction ε, and the outer block length n2 must make n2 D(δ‖ε) (a binomial large-deviation exponent) grow faster than log n1, the union-bound cost of the number of columns. When the outer code is a binary BCH code, this condition is met whenever the inner bit-error probability is o(1/log log n1), which is dramatically slower than the o(1/n1) required by a direct union bound. The upshot: any inner family with bit-error decay o(1/log log n1) at rates below capacity is automatically converted into a capacity-achieving concatenated family. The existence of such an inner family is assumed, not constructed.

What carries the argument

Forney's concatenation Con(C2,C1): codewords are n2×n1 binary matrices whose rows lie in inner code C1 and columns in outer code C2. The two-stage decoder first decodes each row with D1 (bit level), leaving per-column independent residual errors, then decodes each column with D2 (block level). The core identity is the Chernoff bound: a column survivor count Sj ∼ Bin(n2,εℓ) exceeds the correction radius tℓ with probability ≤ exp(−n2 D(δℓ‖εℓ)). The outer BCH code is the algebraic engine: primitive narrow-sense BCH codes of length 2^s−1 have designed distance 2t+1, redundancy ≤ s t, and a bounded-distance decoder; choosing s so that t ≈ 2ε n2 yields rate 1−o(1) and δ = 2ε+o(ε).

What would settle it

Find a BMS channel W and an inner code–decoder sequence with rate approaching some R<C(W) and bit-error probability εℓ = o(1/log log n1,ℓ) for which the two-stage concatenation with the BCH outer code of Corollary 1 has block-error probability bounded away from zero. One direct experimental route: simulate the two-stage decoder for finite n and verify the bound n1 exp(−n2 D(δ‖ε)) — if block-error exceeds it substantially and consistently, the union bound or independence step is wrong.

Watch

Extended reading notes

Core claim

The paper's central claim is Theorem 1: fix a BMS channel W and rate R < C(W). Let (C1,ℓ,D1,ℓ) be inner codes/decoders of length n1,ℓ, rate R, uniform bit-error probability εℓ = Pbit(C1,ℓ,D1,ℓ;W). Let (C2,ℓ,D2,ℓ) be outer codes/decoders of length n2,ℓ, rate tending to 1, whose decoder corrects every error pattern of weight at most tℓ = δℓ n2,ℓ. If εℓ < δℓ and n2,ℓ D(δℓ‖εℓ)/log n1,ℓ → ∞, then the concatenated code Con(C2,ℓ,C1,ℓ) decoded by row-then-column has rate tending to R and block-error probability at most n1,ℓ exp(−n2,ℓ D(δℓ‖εℓ)) → 0. Corollary 1 specializes the outer code to primitive narrow-sense BCH codes, which have rate at least 1 − st/n2 and correct t errors; the proof shows that

Load-bearing premise

The paper's capacity-achieving conclusion depends on the existence, for each rate below capacity, of an inner code–decoder sequence with rate approaching the target and bit-error probability o(1/log log n); the paper proves the conversion but never constructs or supplies a family with that decay.

Editorial extensions

If this is right

For any inner family with vanishing bit-error probability that satisfies condition (3), the concatenation has the same asymptotic rate and vanishing block-error probability on the same BMS channel.
For inner bit-error probability o(1/log log n), primitive narrow-sense BCH outer codes make the conversion explicit and keep the outer rate 1−o(1).
The required bit-error decay is much weaker than the union-bound threshold o(1/n), so the construction extends the class of decoders that can be promoted to block reliability.
The concatenated block-error probability obeys the quantitative bound n1 exp{−n2 D(δ‖ε)}, giving a finite-length design rule for choosing n2 relative to n1 and ε.
If such inner families exist for every rate below capacity, the concatenation construction achieves Shannon capacity on any BMS channel.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The result is essentially a reduction: the whole status of “capacity-achieving concatenated codes” is pinned on the existence of inner codes with bit-error decay o(1/log log n). Establishing or refuting that decay for any known capacity-approaching family would settle whether the route is real.
For channels or families where bit-error decay is slower, one could test whether the BCH condition is genuinely necessary or an artifact of the proof; a threshold lower than o(1/log log n) would require a different outer code, perhaps with δ decaying slower than ε.
The clean row-independence after row decoding relies on the memoryless channel and per-row decoders; extending the argument to channels with memory or to iterative/turbo row decoders would require a new independence argument, not covered here.
A numerical check: on a binary symmetric channel, concatenate a short inner code with a known BCH code, compare measured block-error to n1 exp(−n2 D(δ‖ε)); matching would support the exponent, while a mismatch would expose overlooked dependencies.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Bin Zhang's note gives a clean sufficient condition under which Forney concatenation converts bit-level reliability of an inner code into block-level reliability of the concatenated code on any BMS channel. The main theorem is a standard Chernoff-plus-union-bound argument, but the statement is useful and, as far as I can tell, not in the literature. The BCH specialization is also correct: if the inner bit-error probability epsilon_n = o(1/log log n), then rate-one BCH outer codes with correction radius slightly above epsilon_n do the job. The algebra checks, including the choice of alpha and the rate loss calculation.

What I like: the proof is honest and self-contained. The independence of residual errors across rows follows straight from memorylessness, and the stochastic domination by Bin(n2, epsilon) is valid. The paper does not pretend to construct the inner code; it gives a reduction. That is a real contribution, though a modest one.

The soft spots are the things you'd expect. First, the abstract says any family with vanishing bit-error probability below capacity can be converted. That is technically true only if one also allows an arbitrary outer code sequence chosen to satisfy (3). The paper itself only proves the BCH outer code works when epsilon log log n -> 0. For a slowly decaying epsilon, the general theorem would require an existence argument (e.g., GV-type codes) that is not stated. That is a presentation gap, not an error. Second, the paper never supplies an actual inner family with the required decay. Known polar or RM bit-error bounds almost certainly satisfy it, but the paper should have said so. As written, the headline 'capacity achievement' is conditional on an unstated instantiation. Third, the BCH proof uses Theta(...) a bit loosely at the large-deviation step; the conclusion is right, but a referee may want the constants spelled out.

Who is this for? People working on capacity proofs via concatenation or on bit-to-block conversions. It is a modular tool, not a breakthrough. It deserves a proper referee: it is short, correct, and gives a genuine reduction that others may use. I would send it to review. It may need minor revisions: fix the abstract's overclaim, cite an inner family with the required decay, and tighten the Theta(...) step.

Referee Report

0 major / 4 minor

Summary. The paper studies Forney's concatenation as a method to convert bit-level reliability into block-level reliability. Theorem 1 proves that if the outer code corrects a fraction δℓ > εℓ of residual errors and n2,ℓ D(δℓ ∥ εℓ)/log n1,ℓ → ∞, then the concatenated code has block-error probability at most n1,ℓ exp{−n2,ℓ D(δℓ ∥ εℓ)} and rate tending to the inner rate. Corollary 1 specializes the outer code to primitive narrow-sense BCH codes and shows that a sufficient condition is εℓ log log n1,ℓ → 0. The paper then concludes that any inner family with vanishing bit-error probability satisfying this condition yields a capacity-achieving concatenated family.

Significance. The result is a clean and useful reduction. The proof of Theorem 1 is a standard Chernoff-plus-union-bound argument and is correct. The BCH specialization provides explicit outer codes with rate tending to one and corrects a fraction 2εℓ of errors, and the resulting large-deviation algebra, with the correction noted below, supports the claimed condition. The result is substantially weaker than the direct union-bound requirement n1 εℓ → 0. The main caveat is that the capacity-achieving conclusion is conditional on the existence of inner code–decoder sequences with the stated bit-error decay; the paper neither constructs nor cites such a family. This is a missing instantiation rather than an internal inconsistency, but it should be made explicit.

minor comments (4)

[IV, proof of Corollary 1] The displayed identity D(δℓ∥εℓ) = δℓ log(δℓ/εℓ) + o(δℓ) is not correct. From δℓ = 2εℓ + o(εℓ), the second term of the relative entropy is (1−δℓ) log((1−δℓ)/(1−εℓ)) = −εℓ + o(εℓ), not o(δℓ). The correct expansion is D(δℓ∥εℓ) = (2 ln 2 − 1)εℓ + o(εℓ). The conclusion D(δℓ∥εℓ) = Θ(εℓ) remains valid, but the displayed equation should be corrected.
[Abstract and Section I] The claim that “a family with vanishing bit-error probability for rates below capacity can be converted into a capacity-achieving concatenated-code family” is unqualified. Theorem 1 requires the large-deviation condition (3), and Corollary 1 additionally requires (6). The abstract should either state these conditions explicitly or the paper should cite a known inner family (e.g., polar or Reed–Muller codes) that satisfies them. Without this, the capacity-achieving statement is only a conditional reduction.
[IV, proof of Corollary 1] Typo: “Choose an integers ℓ” should read “Choose an integer sℓ.”
[III, proof of Theorem 1] The step bounding Pr{Ei,j = 1} ≤ εℓ and concluding stochastic domination by Bin(n2,ℓ, εℓ) is standard but uses a coupling that is not spelled out. A one-sentence justification would improve readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the derivation is a self-contained sufficient-condition proof; the capacity claim is explicitly conditional on an assumed inner-code family, not an input masquerading as a prediction.

full rationale

Theorem 1 derives a block-error bound from the Chernoff bound and a union bound. Condition (3), n_{2,ell} D(delta_ell || epsilon_ell) / log n_{1,ell} -> infinity, is exactly the condition under which the derived bound n_{1,ell} exp{-n_{2,ell} D(delta_ell || epsilon_ell)} tends to zero. This is a sufficient condition obtained by the proof, not an input that is renamed as an output; the bound (4) is not assumed but proved from the independence of residual errors, stochastic domination by Bin(n_2, epsilon), and the bounded-distance correction property. Corollary 1 independently chooses BCH outer-code parameters from the inner bit-error probability epsilon_ell and verifies the hypotheses of Theorem 1 using standard BCH redundancy bounds and the relative-entropy estimate. No fitted parameter is later relabeled as a prediction: the inner-code family with epsilon_ell = o(1/log log n_{1,ell}) is explicitly a hypothesis of Corollary 1, and the paper does not claim to construct or explain such a family. There are no self-citations by the sole author that carry the argument, and no uniqueness theorem or prior-work ansatz is imported to force the construction. The advertised capacity-achievement is conditional on the existence of suitable inner sequences, which is a genuine limitation in instantiation but not circular reasoning.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

No empirical fitting; α is an existential proof device. The only substantive pull from outside the paper is the assumed existence of an inner family with fast-decaying bit-error, plus standard BCH and Chernoff facts.

free parameters (1)

alpha_ell
Auxiliary sequence in Corollary 1 chosen to satisfy α ≫ ϵ(log log n1 + log 1/ϵ) and α ≪ 1; existence follows from ϵ=o(1/log log n). It is a proof device, not a fitted constant.

assumptions (5)

domain assumption For every R < C(W), an inner code–decoder sequence exists with rate → R and uniform bit-error probability ϵℓ = o(1/log log n1,ℓ)
Required for the abstract/Corollary conclusion that the concatenation achieves capacity; the paper proves the conversion but does not construct or cite such a family.
domain assumption Primitive narrow-sense BCH codes of length 2^s − 1 with designed distance 2t+1 have redundancy ≤ st and a bounded-distance decoder correcting t errors
Standard fact cited to [10]–[12]; the paper relies on it in Corollary 1 to obtain outer rate → 1 and correction radius δ = 2ϵ + o(ϵ).
standard math Chernoff bound (1): for S ∼ Bin(n,b), a > b, Pr{S ≥ an} ≤ exp(−nD(a∥b)); and the estimate D(a∥b) = a log(a/b) − O(a) when a/b → ∞ and a → 0
Used directly in Theorem 1 and Corollary 1 to control the tail of residual column errors.
standard math BMS channel is memoryless and symmetric; capacity achieved by uniform input
Definition 2 and the preliminary discussion; needed so row outputs are independent across rows and the inner code can be used at rates below capacity.
standard math Con(C2,C1) has rate R(C1)R(C2)
Dimension formula for row-column concatenation used to show the concatenated rate tends to R.

how reviews work

0 comments

Cite this review

Pith. "Pith review of From Bit to Block: Capacity Achievement via Code Concatenation." pith.science (2026). https://pith.science/paper/G32FT6EX

@misc{pith2026260705816,
  author       = {Pith},
  title        = {Pith review of: From Bit to Block: Capacity Achievement via Code Concatenation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G32FT6EX}},
  note         = {Machine review of arXiv:2607.05816}
}

read the original abstract

This paper shows that bit-level reliability can be converted into block-level reliability for binary codes over BMS channels through Forney's concatenation scheme. The construction concatenates an inner code of rate R-o(1) and vanishing bit-error probability \epsilon with an outer code of rate 1-o(1) equipped with a bounded-distance decoder. We show that, when the outer correction radius exceeds \epsilon and a suitable large-deviation condition holds, the concatenated code has rate R-o(1) and vanishing block-error probability. Hence, a family with vanishing bit-error probability for rates below capacity can be converted into a capacity-achieving concatenated-code family. We further show that, when binary BCH codes are used as outer codes, inner codes with bit-error probability \epsilon=o(1/\log\log n) can be converted into a capacity-achieving concatenated family.

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

13 extracted references · 1 linked inside Pith

[1]

A mathematical theory of communi- cation,

C. E. Shannon, “A mathematical theory of communi- cation,”Bell System Technical Journal, vol. 27, no. 3, pp. 379–423, 1948

1948
[2]

Channel polarization: A method for con- structing capacity-achieving codes for symmetric binary- input memoryless channels,

E. Arikan, “Channel polarization: A method for con- structing capacity-achieving codes for symmetric binary- input memoryless channels,”IEEE Transactions on In- formation Theory, vol. 55, no. 7, pp. 3051–3073, 2009

2009
[3]

Reed–Muller codes achieve capacity on erasure channels,

S. Kudekar, S. Kumar, M. Mondelli, H. D. Pfister, E. Sasoglu, and R. L. Urbanke, “Reed–Muller codes achieve capacity on erasure channels,”IEEE Transactions on Information Theory, vol. 63, no. 7, pp. 4298–4316, 2017

2017
[4]

A proof that Reed–Muller codes achieve Shannon capacity on symmetric channels,

E. Abbe and C. Sandon, “A proof that Reed–Muller codes achieve Shannon capacity on symmetric channels,” in Proc. 64th IEEE Symposium on Foundations of Computer Science (FOCS), 2023, pp. 177–193

2023
[5]

Reed–Muller codes on BMS channels achieve vanishing bit-error probability for all rates below capacity,

G. Reeves and H. D. Pfister, “Reed–Muller codes on BMS channels achieve vanishing bit-error probability for all rates below capacity,”IEEE Transactions on Information Theory, vol. 70, no. 2, pp. 920–948, 2024

2024
[6]

A family of capacity- achieving Abelian codes for the binary erasure chan- nel,

L. P. Natarajan and P. Krishnan, “A family of capacity- achieving Abelian codes for the binary erasure chan- nel,” inProc. National Conference on Communications (NCC), 2022, pp. 425–430

2022
[7]

Berman codes: A generalization of Reed–Muller codes that achieve BEC capacity,

L. P. Natarajan and P. Krishnan, “Berman codes: A generalization of Reed–Muller codes that achieve BEC capacity,”IEEE Transactions on Information Theory, vol. 69, no. 11, pp. 6956–6975, Nov. 2023

2023
[8]

From bit to block: Decoding on erasure channels,

H. D. Pfister, O. Sprumont, and G. Zemor, “From bit to block: Decoding on erasure channels,” arXiv:2501.05748, 2025

arXiv 2025

Show all 13 references

[9]

Tensor Reed– Muller codes: Achieving capacity with quasilinear de- coding time,

E. Abbe, C. Sandon, and O. Sprumont, “Tensor Reed– Muller codes: Achieving capacity with quasilinear de- coding time,” arXiv:2601.16164, 2026

2026
[10]

Codes correcteurs d’erreurs,

A. Hocquenghem, “Codes correcteurs d’erreurs,” Chiffres, vol. 2, pp. 147–156, Sep. 1959

1959
[11]

On a class of error correcting binary group codes,

R. C. Bose and D. K. Ray-Chaudhuri, “On a class of error correcting binary group codes,”Information and Control, vol. 3, no. 1, pp. 68–79, 1960

1960
[12]

Lin and D

S. Lin and D. J. Costello,Error Control Coding, 2nd ed. Upper Saddle River, NJ, USA: Prentice Hall, 2004

2004
[13]

G. D. Forney,Concatenated Codes. Cambridge, MA, USA: MIT Press, 1966

1966

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.

[1] [1]

A mathematical theory of communi- cation,

C. E. Shannon, “A mathematical theory of communi- cation,”Bell System Technical Journal, vol. 27, no. 3, pp. 379–423, 1948

1948

[2] [2]

Channel polarization: A method for con- structing capacity-achieving codes for symmetric binary- input memoryless channels,

E. Arikan, “Channel polarization: A method for con- structing capacity-achieving codes for symmetric binary- input memoryless channels,”IEEE Transactions on In- formation Theory, vol. 55, no. 7, pp. 3051–3073, 2009

2009

[3] [3]

Reed–Muller codes achieve capacity on erasure channels,

S. Kudekar, S. Kumar, M. Mondelli, H. D. Pfister, E. Sasoglu, and R. L. Urbanke, “Reed–Muller codes achieve capacity on erasure channels,”IEEE Transactions on Information Theory, vol. 63, no. 7, pp. 4298–4316, 2017

2017

[4] [4]

A proof that Reed–Muller codes achieve Shannon capacity on symmetric channels,

E. Abbe and C. Sandon, “A proof that Reed–Muller codes achieve Shannon capacity on symmetric channels,” in Proc. 64th IEEE Symposium on Foundations of Computer Science (FOCS), 2023, pp. 177–193

2023

[5] [5]

Reed–Muller codes on BMS channels achieve vanishing bit-error probability for all rates below capacity,

G. Reeves and H. D. Pfister, “Reed–Muller codes on BMS channels achieve vanishing bit-error probability for all rates below capacity,”IEEE Transactions on Information Theory, vol. 70, no. 2, pp. 920–948, 2024

2024

[6] [6]

A family of capacity- achieving Abelian codes for the binary erasure chan- nel,

L. P. Natarajan and P. Krishnan, “A family of capacity- achieving Abelian codes for the binary erasure chan- nel,” inProc. National Conference on Communications (NCC), 2022, pp. 425–430

2022

[7] [7]

Berman codes: A generalization of Reed–Muller codes that achieve BEC capacity,

L. P. Natarajan and P. Krishnan, “Berman codes: A generalization of Reed–Muller codes that achieve BEC capacity,”IEEE Transactions on Information Theory, vol. 69, no. 11, pp. 6956–6975, Nov. 2023

2023

[8] [8]

From bit to block: Decoding on erasure channels,

H. D. Pfister, O. Sprumont, and G. Zemor, “From bit to block: Decoding on erasure channels,” arXiv:2501.05748, 2025

arXiv 2025