eduKateSG · GS1 · How X Works
Find your next learning step
Start with the reliability question you need, then follow the worked example and its connected reading route.
Full chapter index · Work the Hamming example · Check your understanding · Connected article directory
The gold standard of Coding Theory is a reliability claim that states the message set, encoding, channel assumptions, decoding rule and failure boundary. A useful explanation lets the reader reproduce a successful recovery and identify the error patterns for which that recovery is no longer guaranteed.
This subject root continues The Gold Standard of Education | Learning That Transfers into Life under How X Works. It connects the existing Information Theory, Probability, Discrete Mathematics and Cryptography roots with parity, checksums, Reed–Solomon codes, finite-field algorithms and reliable communication.
Scope: this guide focuses on error-control coding. The bit strings, polynomial values, channel probabilities and burst examples are specified teaching cases. Their calculations illustrate mathematical guarantees and counterexamples; they are not measured benchmarks for a device or implementation.
Full chapter index
Chapters 1–4 · Define the communication task
Chapters 5–8 · Read distance and redundancy
Chapters 9–12 · Encode and decode
Chapters 13–16 · Use structure and evidence
Chapters 17–20 · Check limits and transfer
CHAPTER 1 OF 20 · Define the communication task
1. Start with a message that must survive a noisy channel
Coding Theory studies ways to represent messages so that a receiver can recognise or recover them when transmission or storage introduces errors. Its central question is precise: which disturbances can this code and decoder handle, under which assumptions, and at what cost?
Imagine sending a one-bit instruction through a teaching channel that can flip a bit. If 0 and 1 are both valid messages transmitted directly, a received 1 could be an unchanged 1 or a corrupted 0. The receiver cannot distinguish those histories from that observation alone.
An encoder can instead represent 0 by 000 and 1 by 111. The extra structure creates separation between the two valid transmitted words. A decoder then needs an explicit rule for interpreting words such as 001.
The gold standard is a complete chain from the message set through encoding, the channel model and decoding to a justified reliability claim. The examples here are mathematical teaching constructions. Their probabilities and error patterns are specified for inspection, rather than measured performance of an actual communication product.
Continue into: Error-correcting codes and parity.
CHAPTER 2 OF 20 · Define the communication task
2. Separate compression, error control and cryptography
Source coding represents information efficiently, often by exploiting structure or probability. Channel coding introduces structured redundancy to support reliable recovery. Cryptography addresses security properties such as confidentiality or authenticity under its own assumptions.
These jobs can appear in one pipeline. A file might be compressed, protected against unauthorised modification and encoded for noisy transmission. The order and implementation matter, but success at one job does not establish the others.
A threefold repetition code makes its message easy to recognise; it also makes it easy to read. Likewise, passing a parity check says something about a specified error test, not who sent the message.
This root concentrates on error control and connects the wider jobs through the directory. Information Theory explains the probability and capacity framework. Compression articles develop efficient representation. Cryptography provides the separate route for security claims.
Continue into: Information Theory · Cryptography · Useful redundancy.
CHAPTER 3 OF 20 · Define the communication task
3. Name the disturbance before choosing a code
A substitution changes a symbol at a known position. An erasure marks a position as unknown. An insertion or deletion changes alignment. A burst groups corruption within a short transmitted interval. These disturbances provide different information to the receiver.
Compare the repetition words received as 0?0 and 010. In the first, the receiver knows which position is missing. In the second, it sees three definite bits and must decide which, if any, changed.
The elementary Hamming-distance calculations in this guide assume aligned words of fixed length with substitutions or marked erasures. An inserted bit can shift every later position and requires additional framing or synchronisation methods.
Write the channel assumptions before counting errors. Independence, equal error probabilities and known symbol boundaries are modelling choices. A code tested under isolated bit flips needs further examination when the real channel produces long correlated bursts.
| Pattern | Example | What the receiver knows |
|---|---|---|
| Substitution | 000 becomes 010 | A definite bit is present, but its correctness is unknown. |
| Erasure | 000 becomes 0?0 | The missing position is marked. |
| Insertion or deletion | A bit is added or lost | Later alignment can change. |
| Burst | Several nearby transmitted positions change | A clustered pattern needs its own model. |
Continue into: Wireless Communication.
CHAPTER 4 OF 20 · Define the communication task
4. State the alphabet, block length and information rate
A block code maps a set of messages into fixed-length codewords. For a binary linear code with k information bits and n transmitted bits, there are 2 raised to k messages, and the code rate is k/n.
The repetition example carries one information bit in three transmitted bits, so its rate is 1/3. A binary (7,4) code carries four information bits in seven transmitted bits, giving rate 4/7. Codewords and message bits are different counts.
Symbols need not be bits. Later, a small polynomial example uses the seven-symbol field of integers modulo 7. A rate expressed in information symbols per transmitted symbol must retain that alphabet convention.
Include the decoder in any comparison. Two codes with different rates can also differ in correctable patterns, latency and computation. Rate alone does not rank their reliability or tell you how many useful messages a complete system delivers per second.
Continue into: Discrete Mathematics.
CHAPTER 5 OF 20 · Read distance and redundancy
5. Measure separation with Hamming distance
The Hamming distance between two equal-length words counts the positions where they differ. For 000 and 111 it is 3. For 0110011 and 0110001 it is 1, because only position 6 differs.
The minimum distance of a code is the smallest distance between any two distinct valid codewords. It describes the closest competing pair, which is the pair that limits a worst-case guarantee.
A received word can be compared with every valid codeword to find a nearest candidate. The decoder must also specify what to do with ties and with patterns outside its promised error range.
MIT’s Computation Structures material gives the geometric distance framework. The calculations below make it concrete: count changed positions, identify the set of possible original words and ask whether the receiver has enough information to choose uniquely.
Primary distance reference: MIT Computation Structures: Hamming distance, parity and correction.
Continue into: Linear Algebra.
CHAPTER 6 OF 20 · Read distance and redundancy
6. Derive the detection and correction guarantees
If a code has minimum distance d, fewer than d substitutions cannot change one valid codeword into another. Checking code membership therefore detects every pattern of up to d − 1 substitutions.
Unique correction of every pattern of up to t substitutions requires d ≥ 2t + 1. Otherwise a received word could lie within t changes of two different codewords. The receiver would face an ambiguity that no choice of confident wording can remove.
For a distance-3 code, every single substitution is uniquely correctable. Membership checking can also detect every pair of substitutions, but a decoder designed to correct one error may interpret a two-error word as a different single-error case.
This distinction is essential: a code’s distance property and a decoder’s delivered behaviour are connected but different claims. State whether the task is detection, correction or a combination, and which error patterns the decision rule is allowed to assume.
| Task | Sufficient minimum-distance condition | For d = 3 |
|---|---|---|
| Detect every pattern of at most s substitutions | d ≥ s + 1 | Detect up to two by membership checking. |
| Uniquely correct every pattern of at most t substitutions | d ≥ 2t + 1 | Correct one. |
| Uniquely recover e erasures plus t substitutions | d > 2t + e | Two erasures, or one substitution, are within the bound. |
CHAPTER 7 OF 20 · Read distance and redundancy
7. Work the repetition code and its probability model
With codewords 000 and 111, majority decoding recovers the transmitted bit whenever at most one of the three positions flips. Two flips cause the majority to favour the wrong message, and three flips do so as well.
Assume each bit independently flips with probability p. The probability of a wrong majority decision is 3p²(1 − p) + p³. The first term counts exactly two flips and the second counts three.
At p = 0.1, the result is 0.028. Under these stipulated assumptions, one information bit sent with repetition has a lower decision error probability than the uncoded bit’s 0.1, while requiring three transmitted bits.
The calculation does not compare equal bandwidth, delay or energy budgets, and it does not apply unchanged to correlated errors. Its value is that every assumption and counted event is visible. A fair engineering comparison must specify the resources and channel being held constant.
| Number of flipped positions | Probability under independent crossover p | Decision |
|---|---|---|
| 0 | (1 − p)³ | Correct |
| 1 | 3p(1 − p)² | Correct |
| 2 | 3p²(1 − p) | Wrong |
| 3 | p³ | Wrong |
Continue into: Probability.
CHAPTER 8 OF 20 · Read distance and redundancy
8. Use parity to detect a pattern without inventing a location
An even-parity encoder appends a bit so that the complete word contains an even number of ones. For message 1011, there are three ones, so the appended bit is 1 and the transmitted word is 10111.
Flipping any one position changes the parity. Flipping any two distinct positions preserves it. More generally, the check detects an odd number of flips and misses an even number.
A failed parity check identifies a violated constraint, not the corrupted position. A one-bit error in any of the five positions of this example produces the same overall parity failure.
Passing the check means that this particular constraint is satisfied. It does not prove the received word is the original. The Checksums and CRCs branch develops richer detection methods; its specific guarantees still depend on the chosen construction and error model.
Continue into: Checksums and CRCs · Barcodes and check digits.
CHAPTER 9 OF 20 · Encode and decode
9. Make the arithmetic and parity checks explicit
Binary linear codes use arithmetic in the two-element field, often written GF(2). Addition is modulo 2: 1 + 1 = 0. A parity sum is therefore the exclusive-or of the participating bits.
For an even-parity check, the sum of its selected codeword bits is zero. Several checks can be collected into a parity-check matrix H. Using column vectors here, a valid codeword c satisfies Hc = 0.
If the channel adds an error vector e, the received vector is r = c + e. Its syndrome is Hr = H(c + e) = He, because Hc = 0. The syndrome depends on the error pattern through these checks.
That identity explains the method without claiming that every syndrome uniquely identifies every error. Different error vectors can share a syndrome. Decoding requires assumptions or a selection rule that says which of those patterns to consider.
Primary algebraic coding reference: MIT 6.02: Linear Block Codes — Encoding and Syndrome Decoding.
Continue into: Abstract Algebra.
CHAPTER 10 OF 20 · Encode and decode
10. Encode a complete seven-bit Hamming example
Fix the positions from 1 to 7 as p1, p2, d1, p4, d2, d3, d4. Put the four message bits in positions 3, 5, 6 and 7. All parity checks in this example use even parity.
Check 1 covers positions 1, 3, 5 and 7. Check 2 covers 2, 3, 6 and 7. Check 4 covers 4, 5, 6 and 7. The check memberships encode each position’s nonzero three-bit binary label.
For message d1d2d3d4 = 1011, the parity bits are p1 = 0, p2 = 1 and p4 = 0. The complete codeword is 0110011. Substituting its bits into all three checks gives zero.
This position convention is part of the example. Other presentations arrange the same kind of code differently. A decoder and encoder must share the bit ordering, check equations and syndrome interpretation before their results can be compared.
| Position | Role | Bit | Participating check labels |
|---|---|---|---|
| 1 | p1 | 0 | 1 |
| 2 | p2 | 1 | 2 |
| 3 | d1 | 1 | 1, 2 |
| 4 | p4 | 0 | 4 |
| 5 | d2 | 0 | 1, 4 |
| 6 | d3 | 1 | 2, 4 |
| 7 | d4 | 1 | 1, 2, 4 |
Parity calculation: p1 = d1 + d2 + d4 = 1 + 0 + 1 = 0 modulo 2; p2 = d1 + d3 + d4 = 1 + 1 + 1 = 1; p4 = d2 + d3 + d4 = 0 + 1 + 1 = 0.
CHAPTER 11 OF 20 · Encode and decode
11. Decode one error and expose a two-error failure
Flip position 6 of 0110011. The receiver obtains 0110001. Its check results are s1 = 0, s2 = 1 and s4 = 1. Reading s4s2s1 as a binary number gives 110, or position 6.
Under the assumption of at most one flipped bit, the syndrome identifies the location. Flipping position 6 back recovers 0110011, and reading data positions 3, 5, 6 and 7 returns 1011.
Now flip positions 1 and 2 of the original instead. The received word is 1010011. Its syndrome points to position 3. A single-error decoder flips that position and returns 1000011, a different valid codeword with message 0011.
This counterexample shows why a nonzero syndrome cannot be read as proof of exactly one error. The algorithm follows its designed rule, but the actual disturbance violates the assumption needed for a correct result.
| Case | Received word | Syndrome s4s2s1 | Decoder action and outcome |
|---|---|---|---|
| Flip position 6 | 0110001 | 110 = 6 | Flip 6; recover 0110011 and message 1011. |
| Flip positions 1 and 2 | 1010011 | 011 = 3 | Flip 3; return 1000011 and wrong message 0011. |
Continue into: Error-correcting codes and parity.
CHAPTER 12 OF 20 · Encode and decode
12. Add an overall parity bit and define the decision rule
Append an eighth bit that makes the entire seven-bit Hamming word plus that bit have even parity. The example 0110011 already has four ones, so its extended form is 01100110.
Use both the three Hamming check results and the overall parity. Assuming at most two flipped bits in the eight-position word, a nonzero syndrome with odd overall parity identifies a single error among the first seven positions.
A zero syndrome with odd overall parity identifies an error in the eighth position. A nonzero syndrome with even overall parity signals two errors, which the stated decoder detects and does not attempt to correct.
A zero syndrome with even overall parity means no error under the at-most-two assumption. Larger error patterns can defeat these interpretations. The extended code’s distance is 4, supporting the specified single-error-correction and double-error-detection behaviour, rather than unlimited recovery.
| Hamming syndrome | Overall parity | Interpretation and action |
|---|---|---|
| Zero | Even | No error under the stated bound. |
| Zero | Odd | Correct the overall parity bit. |
| Nonzero | Odd | Correct the indicated position among the first seven. |
| Nonzero | Even | Detect two errors; do not apply single-error correction. |
CHAPTER 13 OF 20 · Use structure and evidence
13. Use known erasure positions as additional information
An erasure tells the receiver where information is missing. For a code with minimum distance d, up to d − 1 marked erasures leave a unique valid codeword when all remaining positions are correct.
The reason is direct. If two valid words matched every unerased position, they could differ only in the erased positions. With fewer than d such positions, that would contradict the code’s minimum distance.
A distance-3 code can therefore recover two erasures, while its guaranteed unknown-substitution correction radius is only one. The receiver spends less uncertainty on locating what went wrong when positions are already marked.
For a mixture of t unknown substitutions and e marked erasures, d > 2t + e is a sufficient worst-case uniqueness condition. The decoder still needs a method for achieving that recovery and a defined response when the condition is exceeded.
Continue into: Side information.
CHAPTER 14 OF 20 · Use structure and evidence
14. Build a small polynomial evaluation code
Work in arithmetic modulo 7. Encode two information symbols a and b as the values of f(x) = a + bx at the four distinct points x = 0, 1, 2 and 3.
For a = 2 and b = 3, the codeword is (2,5,1,4). Each value is reduced modulo 7. If the values at x = 0 and x = 2 are erased, the remaining pairs are (1,5) and (3,4).
The slope is (4 − 5)/(3 − 1) modulo 7. The inverse of 2 is 4 because 2 × 4 is congruent to 1. Thus b = 6 × 4 = 3 modulo 7, and a = 5 − 3 = 2. Both erased values can then be reconstructed.
Two distinct degree-at-most-one polynomials agree at no more than one field point. Their four-symbol codewords therefore differ in at least three positions; examples attain three, so the minimum distance is 3. This small Reed–Solomon-style construction connects finite fields, interpolation and recovery without reproducing a production specification.
| Evaluation point x | Raw integer value | Transmitted symbol modulo 7 |
|---|---|---|
| 0 | 2 | 2 |
| 1 | 5 | 5 |
| 2 | 8 | 1 |
| 3 | 11 | 4 |
Further algebraic coding route: William A. Geisel’s NASA tutorial on Reed–Solomon error correction coding.
Continue into: Reed–Solomon codes · Finite-field factorisation.
CHAPTER 15 OF 20 · Use structure and evidence
15. Spread a burst across codewords through interleaving
Take three four-symbol codewords A, B and C. Sending each complete word in turn gives A1, A2, A3, A4, then B1 through B4 and then C1 through C4.
Instead, send by columns: A1, B1, C1, A2, B2, C2, and so on. A burst corrupting three consecutive transmitted symbols then strikes one symbol in each of the three words, provided alignment and the stated transmission order remain intact.
After deinterleaving, a decoder capable of correcting one symbol error per word can handle that stipulated burst. Interleaving changes the distribution of errors across words; it adds neither code distance nor extra information by itself.
The receiver needs buffering, an agreed ordering and suitable handling of lost alignment. Longer bursts require a new analysis. NASA’s historical concatenated-coding report gives a concrete specialist route into interleavers and layered codes.
Historical engineering reference: NASA NTRS: Bandwidth efficient CCSDS coding standard proposals (1992). This is a reference to the described concatenated architecture, not a statement that it is the current specification of a particular spacecraft.
Continue into: Telecommunications.
CHAPTER 16 OF 20 · Use structure and evidence
16. Let reliability information change the decoding calculation
Majority decoding treats all received bits equally. Suppose instead that a three-bit repetition word crosses three independent binary channels with known crossover probabilities 0.01, 0.4 and 0.4, and the received word is 011.
The likelihood of receiving 011 if 000 was sent is 0.99 × 0.4 × 0.4 = 0.1584. If 111 was sent, it is 0.01 × 0.6 × 0.6 = 0.0036. Maximum-likelihood decoding therefore chooses 000, although an unweighted majority chooses 111.
The first channel’s much greater reliability changes the evidence. These are stipulated channel probabilities, not confidence numbers invented from the received bits. With equal message priors, the likelihood comparison also gives the maximum-posterior choice.
Soft-information decoders can retain more evidence than an early hard bit decision. The exact metric depends on the channel model and observations. A larger score is useful only when its meaning and calibration are supported.
| Candidate sent word | Likelihood calculation | Result |
|---|---|---|
| 000 | 0.99 × 0.4 × 0.4 | 0.1584 |
| 111 | 0.01 × 0.6 × 0.6 | 0.0036 |
Continue into: Viterbi decoding.
CHAPTER 17 OF 20 · Check limits and transfer
17. Read capacity as a limit under a model
For a binary symmetric channel with independent crossover probability p, capacity is C = 1 − H2(p) bits per channel use, where H2(p) = −p log2 p − (1 − p) log2(1 − p). Endpoint values use the usual limiting convention.
At p = 0.1, H2(p) is approximately 0.4690 and C approximately 0.5310. The number is a property of the specified channel model and input/output convention, not a measured speed in megabits per second.
The channel coding theorem gives an asymptotic possibility: at suitable rates below capacity, increasingly long codes can achieve arbitrarily small error probability under the theorem’s conditions. It does not certify that a chosen short code, decoder or implementation already achieves that performance.
The repetition code’s rate 1/3 is below this model’s capacity, yet its majority-decoding error probability at the stipulated p remains 0.028. A rate comparison and a finite-code performance calculation answer different questions.
Capacity reference: MIT Information Theory, Lecture 8: binary symmetric and erasure channels.
Continue into: Information Theory.
CHAPTER 18 OF 20 · Check limits and transfer
18. Compare the complete cost of a reliability design
Reliability can consume transmitted symbols, delay, buffering, computation and energy. A useful comparison fixes the task, channel and resources, then measures the outcomes that matter for that task.
A receiver might detect corruption and request retransmission, or use forward error correction to recover without a new request. Feedback delay, availability of a return path and remaining undetected errors affect the resulting design.
Likewise, a code can provide a strong minimum-distance guarantee while a practical decoder uses an approximation. Report the implemented decision rule and observed failure behaviour rather than substituting the best possible decoder’s theoretical promise.
Keep random-error protection separate from adversarial authenticity. A checksum or a publicly known error-correcting code can be recomputed by someone deliberately changing a message. Use the Cryptography branch for the distinct question of validating origin or authorised integrity.
Continue into: Cryptography · Hard disk drives.
CHAPTER 19 OF 20 · Check limits and transfer
19. Verify a small code exhaustively before extending it
The specified seven-bit encoder accepts 16 four-bit messages. Enumerate them, check that all codewords satisfy the parity equations and calculate the minimum distance between distinct words. The result for this construction is 3.
For every codeword, test the unchanged word and all seven single-bit flips. That gives 16 × 8 = 128 cases, and the stated single-error syndrome decoder should recover the correct message in every one.
Then include the explicit two-error counterexample rather than treating the promised range as universal. For the extended eight-bit decoder, enumerate each unchanged word, every single flip and every pair of flips; check the separate correction and detection outcomes.
Record the bit convention, check ordering, decoder rule and expected outcome with each case. An exhaustive finite test can establish these finite claims for the specified implementation. Broader channel performance still needs the probability model, metrics and resource comparison.
Verification for this guide: direct enumeration of the declared construction found 16 seven-bit codewords with minimum distance 3, and recovered all 128 unchanged-or-single-flip cases. The extended decision rule flagged all 448 two-position-flip cases. Enumeration of the 49 polynomial codewords over GF(7) also gave minimum distance 3. These are finite checks of the specified examples, not channel measurements.
Continue into: Berlekamp–Massey.
CHAPTER 20 OF 20 · Check limits and transfer
20. Transfer the standard into reading and designing reliable systems
The useful habit is to attach every claim to a recoverable chain of reasoning. What are the valid messages? What redundancy was added? Which disturbances are possible? What does the receiver observe, and what rule converts that observation into a decision?
A parity failure, a nonzero syndrome and a low decoding likelihood convey different information. None should be translated automatically into “the original message has been recovered”. Preserve the exact meaning of the check.
Use the directory to follow the next bottleneck: finite-field arithmetic, distance, probability, checksums, decoding algorithms or communication constraints. Return to a complete worked example after each specialist branch.
The gold standard of Coding Theory is a reliability claim that remains correct at its boundary: the reader can demonstrate a successful recovery, explain the conditions that make it possible and show what can happen when those conditions are exceeded.
Continue into: Education parent · How X Works.
Connected article directory
Use the subject root to orient the question, then follow the existing branch that supplies the next concept, algorithm or application.
Foundations
| Article | Continue here to |
|---|---|
| Information Theory | Connect coding to information measures and channel limits. |
| Probability | Read the independence and error-event assumptions. |
| Discrete Mathematics | Develop finite structures and counting. |
| Linear Algebra | Understand matrix representations and constraints. |
| Abstract Algebra | Follow the mathematical route into fields and algebraic structure. |
Communication and security
| Article | Continue here to |
|---|---|
| Cryptography | Follow the separate security and authenticity route. |
| Telecommunications | Connect reliability to a communication service. |
| Wireless Communication | Study the surrounding transmission system. |
Error control
| Article | Continue here to |
|---|---|
| Error-correcting codes and parity | Start from reliable transfer and simple checks. |
| Checksums and CRCs | Read the detection rules and their limits. |
| Reed–Solomon codes | Continue into symbols, fields and recovery. |
| Barcodes and check digits | See detection in a familiar representation. |
| Making noisy messages arrive correctly | Connect redundancy with the communication purpose. |
Decoding and algebra
| Article | Continue here to |
|---|---|
| Viterbi decoding | Follow a path-based decoding algorithm. |
| Berlekamp–Massey | Explore finite-field sequences and recurrences. |
| Finite-field factorisation | Extend the algebraic toolkit. |
Representations and storage
| Article | Continue here to |
|---|---|
| Useful redundancy | See why an efficient system may add structure back. |
| Side information | Ask what the receiver already knows. |
| Hard disk drives | Read a storage-system context. |
| Barcode geometry | Connect representation to scanning conditions. |
| Braille and binary patterns | Explore finite pattern counts without equating every representation with error control. |
Continue learning
| Article | Continue here to |
|---|---|
| Education parent | Return to learning that transfers into life. |
| How X Works | Continue through the wider mechanism library. |
Check your understanding
Use the seven-position convention from Chapter 10. Encode message 0101. If position 5 flips, what syndrome is obtained and which message should the single-error decoder return? Then state whether two unknown substitutions are within its correction guarantee.
Worked check
The message bits occupy positions 3, 5, 6 and 7 as 0, 1, 0 and 1. The parity bits are p1 = 0, p2 = 1 and p4 = 0, giving codeword 0100101. Flipping position 5 produces 0100001. The syndrome is s1 = 1, s2 = 0, s4 = 1, so s4s2s1 = 101, or position 5. Correcting that bit restores message 0101. Two unknown substitutions exceed this decoder’s single-error guarantee.
Questions readers ask
Is Coding Theory the same as learning to program?
Here it means the mathematical study of codes and reliable representation, especially error control. Programming can implement an encoder or decoder, but it is a separate skill.
Does a passing check prove the message is unchanged?
Only the stated detection property is established. Some changed words satisfy the same checks, and a correction attempt can return a different valid word when the disturbance exceeds its assumptions.
Why can the same code detect two errors yet fail when correcting them?
Membership checking can flag a non-codeword. A single-error decoder additionally chooses a nearby valid word, and a two-error received word may be near the wrong one.
Why do erasures permit more recovery than unknown errors?
Their positions are already identified. The receiver needs to recover the missing values without also locating the corrupted positions.
Is the polynomial example a full production Reed–Solomon system?
It is a small evaluation-code construction over GF(7), selected so the interpolation can be checked by hand. Production systems require their exact field, representation, framing and decoder conventions.
Where should a learner begin?
Count Hamming distances, work repetition and parity, then follow the complete Hamming example. Use the algebra and probability routes when the next derivation needs them.
Sources and scope
The primary teaching and technical references linked at the relevant chapters were checked on 22 October 2026. MIT provides the distance, linear-code and capacity background; NASA’s archived technical reports provide the Reed–Solomon and concatenated-coding routes.
MIT: Computation Structures — distance and parity · MIT: Linear Block Codes · MIT: Information Theory, Lecture 8 · NASA: Tutorial on Reed–Solomon Error Correction Coding · NASA: Concatenated coding architecture.
All bit-string walkthroughs, channel probabilities, GF(7) calculations, practice questions and finite enumeration checks in this guide describe the explicitly stated examples. A reliability claim for another system needs its own code specification, channel assumptions, decoder and resource comparison.
Contents · Return to the Education parent · The Gold Standard master hub · Continue through How X Works