Here’s a simple visual to lock it in:

Indirect dimension (evolution time, t)

UNIFORM (full sampling):
t=0 |x|x|x|x|x|x|x|x|x|x|x|x|x|x|x|x|  ... up to TD/2-1  → AQ(full)

NUS (good schedule: dense early, sparse late, but reaches long t):
t=0 |x|x| |x| | |x| | | |x|       |   |   |  x           |   → AQ(effective) ≈ desired
             (Poisson-gap / exp-biased; includes some long-t points)

NUS (bad schedule: truncated early by too-small NusT2):
t=0 |x|x| |x| | |x| | | |x|                           (stops halfway)
                                          ^ 
                                          longest sampled point << TD/2-1
                                          → AQ(effective) ≪ desired (peaks broaden)

• TD sets the full grid (max digital resolution if fully sampled).
• AQ = TD / (2·SW) is the full-evolution time if you sampled all points.

• NusT2 caps how far along t you allow the NUS scheduler to go.

  • If NUSLIST’s max index ≈ TD/2−1 → you’re reaching the designed long times.
  • If it tops out much earlier → your effective AQ shrinks → broader peaks.

---

4D NUS Pre-Flight Checklist (TopSpin) — for HCCH-NOESY and HCNH-NOESY

A) Set chemical targets per dimension

• Choose SW to only cover what you need

  • ¹H aliphatic: ~0.5–5.5 ppm
  • ¹³C aliphatic: ~0–80 ppm
  • ¹⁵N (if present in other 4Ds): ~90–130 ppm

• Pick desired AQ (ms) from relaxation + crowding

  • ¹H(indirect): 4–6 ms at ~25 kDa (3–8 ms typical)
  • ¹³C(indirect): 12–16 ms (10–20 ms typical)

B) Convert to TD (TopSpin relation)

• AQ = TD / (2·SW) ⇒ TD = 2·SW·AQ (SW in Hz)
• Round TD to convenient values (e.g., 64/96/128/160/192).

C) Generate NUS with correct reach

• Set NusT2 = 2 × desired AQ for each indirect dim (critical to reach long t).
• Use Poisson-gap / exp-biased schedules (dense early, sparse late, but not flat).
• %NUS (total samples): start 2–5 % for 4D NOESY at 25 kDa.

D) Balance %NUS vs NS (time budgeting)

• Keep NS per increment minimal (e.g., 8 scans).
• Prefer more unique points (higher %NUS) over cranking NS for better time-efficiency.
• S/N scales ≈ √NS; resolution comes from TD/AQ, not %NUS.

E) Sanity-check the actual NUSLIST (do this every time)

• For each indirect dim, ensure the max index ≈ TD/2 − 1 appears (at least a few entries).
• Density should decay with t (not flat).
• If max index ≪ TD/2 − 1 → increase NusT2 and regenerate.

F) Processing guards for fast-decaying ¹H(indirect)

• Apodization: QSINE (SSB 2–3) to tame t₁-noise.
• Linear Prediction (optional, conservative): forward or forward-backward, extend +10–20 % max; skip if S/N is weak or there’s a water ridge.
• Zero-fill to next power of two; mild baseline correction.

G) Quick numeric templates (25 kDa, aliphatic 4D)

• If F1 = ¹H(indirect): SW ≈ 5 kHz, AQ 5 ms → TD ≈ 50 → use 64; set NusT2 ≈ 10 ms
• ¹³C(indirect) dims (F2/F3): SW ≈ 12 kHz, AQ 14 ms → TD ≈ 336 → use 320/352; NusT2 ≈ 28 ms
• Direct ¹H (F4): TD 1024–2048 (AQ ~0.10–0.15 s)

H) One-minute pre-launch pilot

• Start a short pilot (a few minutes), reconstruct, and check:

  • F1/F2 linewidths (no truncation)
  • S/N acceptable
  • No obvious NUS artefacts → if issues, adjust %NUS / NusT2 / apodization.

Same rules apply whether F1 is ¹H in HCCH or HCNH NOESY: ¹H(indirect) has short T₂ → keep AQ modest, ensure NUS reaches long-t via NusT2 = 2×AQ, and bias sampling toward early times while still touching the tail.