FA-72235 / Error-correcting codes / Member archive

LLR quantizer floors negative half values · case 05

Negative samples are quantised one step less reliable than positive samples of equal magnitude.

Member previewVariant 5 · 3 implementations · 8 checks per implementation

Case contract

Quantise soft demodulator samples to signed integer LLRs for a hard/soft decoder. Each sample y (integer) is scaled (v = y * scale, in hundredths), rounded to the nearest integer with halves rounded away from zero, and saturated symmetrically to +-(2^(bits-1) - 1). Positive LLR means bit 0, negative bit 1, zero is an erasure (None). Return [llrs, hard decisions].

Why this case matters

Fixed-point receivers quantise LLRs before feeding Viterbi or LDPC decoders; bias at this stage costs coding gain.

One recorded failure

Sample boundary fixture

This sample comes from the broken implementation of a controlled reproducer.

Boundary fixture	Actual	Expected	Outcome
regression [[250, -250, 150, -150, 49, -50], 1, 4]	`[[3, -2, 2, -1, 0, 0], [0, 1, 0, 1, null, null]]`	`[[3, -3, 2, -2, 0, -1], [0, 1, 0, 1, null, 1]]`	Failed

MEMBER ARCHIVE

The complete case is available to members.

This record includes three runnable implementations, regression fixtures, execution results, and source hashes.

Member access is invitation-based. Sign in with your invited account to inspect the sources.