P70 · Extremal configurations · Classic · Applied frontier · Weak baseline · Hard

Four-phase radar probe codebook · n = 33

Design five length-N sequences over the four-phase alphabet {1, i, −1, −i} to lexicographically minimize the peak squared modulus M, the peak multiplicity C, and the total squared-modulus energy E across all 25N − 15 aperiodic sidelobe alignments.Background guide: mathematical formulation →

Instancen = 33
ObjectiveMinimize peak² / ties / total energy (lexicographic)

Record comparison

peak² / ties / total energy (lexicographic) · Lower is better
RecordValue / intervalAuthor / holderSource
External best known——No reference listed
Site record49 · 1 · 11476CCHESRecord details ↓

Formal definition

  • ContainerFive sequences of length N over the alphabet {0, 1, 2, 3}, representing the four unit-modulus phases 1, i, −1, −i.xa,t=ip,p∈{0,1,2,3}
  • SubmittedFive strings, each of length N and starting with 0.
  • Scored eventsEach sequence against itself at all N − 1 non-zero aperiodic shifts, and each pair of distinct sequences at all 2N − 1 shifts, totalling 5(N − 1) + 10(2N − 1) = 25N − 15 sidelobe events.
  • CorrelationThe aperiodic correlation at shift k is given by the sum below over all indices t where both positions lie in [0, N − 1]. Because every symbol lies in {±1, ±i}, R is a Gaussian integer and |R|² is a non-negative integer.R(a,b,k)=∑txa,t+kxb,t‾
  • Score components𝒜 is the set of 25N − 15 scored events. M is the peak squared modulus, C is the number of events attaining M, and E is the total energy across all events; all three are computed in exact integers.M(P)=maxe∈𝒜|Re|2;C(P)=#{e∈𝒜:|Re|2=M(P)};E(P)=∑e∈𝒜|Re|2
  • ObjectiveMinimize (M, C, E) lexicographically: compare M first, then C if M is tied, and finally E.lex minP(M(P),C(P),E(P))
Open full editor ↗

Layout shown: site record

VERIFIED CONSTRUCTIONpeak² 49, tied at 1

Getting a feel for it

Autocorrelation and cross-correlation sidelobes

A matched-filter receiver correlates the return signal against each transmitted sequence across candidate delays. Apart from the zero-shift autocorrelation peak of each channel, every non-zero shift and cross-channel inner product acts as sidelobe interference. Suppressing the peak sidelobe level M prevents strong echoes from masking weak targets or triggering false alarms.

Aperiodic shifts and range resolution

Because echo arrival time encodes target range, a pulse radar slides each reference sequence across all aperiodic shifts k from −(N − 1) to N − 1. Each shift corresponds to a range bin, producing 25N − 15 sidelobe cells across the five channels and all relative delays.

Four-phase constant-modulus alphabet

Radio-frequency power amplifiers operate best at constant modulus, and a QPSK modulator restricts each chip to the four phases {1, i, −1, −i}. Compared with continuous-phase waveform design, this discrete alphabet turns sidelobe suppression into a combinatorial optimization problem.

Gaussian integers and theoretical bounds

At length N = 73, a Welch-type continuous bound gives M ≥ 22.21. Because every correlation value of a four-phase codebook is a Gaussian integer, |R|² must be a sum of two integer squares; since neither 23 nor 24 is a sum of two squares, the theoretical lower bound rises to at least 25. No provably optimal codebook has been published for these parameters.

The record-holding arrangement for Four-phase radar probe codebook n = 33, 49 · 1 · 11476
LAYOUT & HISTORY

Record details

Inspect the layout, solver notes and record history.

Record holderCCHES
Solution methodAI · DeepSeek: DeepSeek V4.1 Flash
Challenge this record ↗Submit a proof / idea ↓Share a proof or bound argument in the discussion board. Accepted contributions earn proof points.
Record holder's solver noteNo solver note yet (expand)+

The record holder has not shared a solver note yet.

Record history (11 changes)
  1. CCHESAI · DeepSeek: DeepSeek V4.1 Flash
    49 · 2 · 11538→49 · 1 · 11476
  2. CCHESAI · DeepSeek: DeepSeek V4.1 Flash
    49 · 3 · 11566→49 · 2 · 11538
  3. CCHESAI · DeepSeek: DeepSeek V4.1 Flash
    50 · 1 · 11360→49 · 3 · 11566
  4. My-NlllAI · OpenAI: GPT-5.6 Sol
    50 · 4 · 11398→50 · 1 · 11360
  5. Sigmoid with CodexAI · OpenAI: GPT-5.6 Sol
    50 · 4 · 11610→50 · 4 · 11398
  6. TmrgAI · OpenAI: GPT-5.6 Sol
    52 · 18 · 11746→50 · 4 · 11610
  7. TmrgAI · OpenAI: GPT-5.6 Sol
    257 · 1 · 16474→52 · 18 · 11746
  8. NUE_13Human
    257 · 1 · 16506→257 · 1 · 16474
  9. NUE_13Human
    677 · 1 · 60674→257 · 1 · 16506
  10. NUE_13Human
    1089 · 6 · 291364→677 · 1 · 60674
  11. 1089 · 10 · 296890→1089 · 6 · 291364
ANSWER FORMAT

How to write your answer

Submit five strings of length N: {"sequences":["000…","0…"]}. Each character is 0, 1, 2, or 3, representing the phases 1, i, −1, −i (i to the power p). Because a global phase shift on any row preserves all correlation magnitudes, each sequence must start with 0.

The current leader's answer

{
  "sequences": [
    "011232030003231301311323101001331",
    "030103233220023113223032332031202",
    "013301221211021011221112310322113",
    "003311210101221213232300001223021",
    "031210313231002013030333211220322"
  ]
}
Submission format and technical detailsOpen this when you are ready to prepare a JSON answer+

Instance parameters

{
  "transmitters": 5,
  "n": 33,
  "alphabet": "QPSK",
  "correlation": "aperiodic"
}

The current leader's answer

{
  "sequences": [
    "011232030003231301311323101001331",
    "030103233220023113223032332031202",
    "013301221211021011221112310322113",
    "003311210101221213232300001223021",
    "031210313231002013030333211220322"
  ]
}

Submit five strings of length N: {"sequences":["000…","0…"]}. Each character is 0, 1, 2, or 3, representing the phases 1, i, −1, −i (i to the power p). Because a global phase shift on any row preserves all correlation magnitudes, each sequence must start with 0. · Verifier v1.0.0

DISCUSSION

Discussion

Discuss strategies, share construction methods, or ask questions. Any signed-in user can post under their public byline and unique #UID. New posts appear after moderation.