AI coverage analysis · RTL verification

Stop guessing which tests to write.

Photoniq reads your simulation coverage database and predicts the exact test vectors that will hit your remaining uncovered RTL logic — cutting directed-test writing from weeks to hours.

coverage_map.ucdb — Photoniq Analyzer MODULE TOGGLE LINE FSM FUNC alu_core 97% 100% 98% 84% fetch_unit 100% 88% 61% 52% decode_fsm 95% 91% 79% 96% lsu_buffer 76% 63% 48% 41% wb_control 99% 97% 100% 82% phnq Detected 3 critical gaps · Generating ranked test recommendations... RANKED TEST VECTORS (priority order): 01 lsu_buffer_misaligned_access_edge (score: 0.94) 02 fetch_unit_branch_mispred_recovery (score: 0.87) 03 fetch_unit_fsm_stall_sequence_2b (score: 0.81) © 2026 Photoniq — photoniqo.com
~40%
reduction in directed-test writing cycles before tape-out sign-off
≥92%
coverage prediction accuracy on held-out test suites
85%
is where random simulation stalls. Photoniq finds what's left.

The coverage plateau

Random simulation runs out of ideas at 85%.

Every chip team hits it. Coverage climbs fast through the first weeks of constrained-random simulation, then flattens. The remaining 10–15% is functionally uncovered — and the only way in is directed tests written by hand.

Writing directed tests takes weeks. DV engineers read RTL, reason through corner cases, write test scenarios, iterate. It is the slowest and most error-prone part of the verification cycle. Coverage escapes at tape-out are almost always in this zone.

Photoniq breaks the plateau. Our model reads your existing coverage database and predicts the specific test vectors that will reach uncovered logic — ranked by likelihood of finding new bugs.

0% 50% 85% 100% 85% plateau zone simulation time → coverage %

How it works

Point it at your coverage database.

Three steps from simulator output to ranked test recommendations. No new toolchain, no RTL rewrite.

STEP 01
Connect your simulator

Plug in via standard UCDB or SAIF export. VCS, Questa, and Xcelium all supported. One CLI command or API call.

STEP 02
AI analyzes RTL + coverage gaps

Our model reads your RTL source and your existing test suite coverage map. It identifies uncovered logic paths and infers which stimulus patterns would reach them.

STEP 03
Get ranked test recommendations

Photoniq outputs a ranked list of test scenarios most likely to close your remaining coverage gaps — with confidence scores so you can triage and act immediately.

Coverage prediction

Know where you're exposed before tape-out.

Random simulation leaves a predictable residue: rarely-toggled FSM state transitions, out-of-bounds addressing sequences, multi-cycle protocol edge cases. Photoniq's model was trained on RTL coverage patterns from real designs — it learns which code constructs correlate with hard-to-reach logic and which input sequences reliably unlock those paths.

The output is a ranked list: these specific test scenarios will toggle these specific uncovered bins, ordered by predicted bug-discovery probability. Not a hint — a scored, actionable manifest.

Yield impact modeling

Layout choices have a cost. Now you can see it.

Every floorplan decision — block aspect ratio, density allocation, metal layer assignment — affects manufacturing yield. Photoniq estimates the yield impact of layout choices using process-specific defect density models, giving your physical design team an early signal that would otherwise only surface at post-silicon analysis.

Correlate layout regions with functional coverage holes. When an uncovered logic block sits in a high-yield-risk zone, your triage priority is obvious. We do not replace your DRC or LVS toolchain — we flag which coverage gaps carry compounded yield risk.

Simulator integration

Fits in your existing flow.

Photoniq plugs into VCS, Questa (formerly ModelSim), Xcelium, and Riviera-PRO via standard UCDB and SAIF export formats. No changes to your simulation scripts, no proprietary instrumentation.

One CLI command exports your coverage database and sends it to the Photoniq analysis engine. Results return as a structured JSON manifest you can version-control alongside your RTL — and diff across simulation runs to track closure progress.

Terminal — phnq CLI
$ phnq analyze --ucdb sim.ucdb --rtl ./src

Photoniq v0.9.1 · Connecting to analysis engine...
✓ UCDB loaded: 12,847 coverage bins across 34 modules
✓ RTL indexed: 48,213 lines across ./src/**/*.sv
Running gap analysis...
⚠ 3 critical uncovered regions detected
Generating ranked test recommendations...

✓ Analysis complete in 4.2s

Results → ./phnq-manifest.json (10 recommendations)
$ cat phnq-manifest.json | jq '.recommendations[0]'
{
  "test_scenario": "lsu_buffer_misaligned_access_edge",
  "target_module": "lsu_buffer",
  "confidence": 0.94,
  "coverage_bins_hit": ["misaligned_word_rd", "misaligned_dword_wr"],
  "suggested_constraints": { ... }
}

Who uses Photoniq

Built for teams that can't miss bugs at tape-out.

ASIC/SoC design teams

Pre-tapeout verification under time pressure. Photoniq identifies the directed-test scenarios you haven't written yet — the ones most likely to catch functional bugs before mask production. Verification sign-off in fewer cycles with documented coverage confidence.

AI chip startups

Custom accelerator RTL with compressed timelines and smaller DV teams. No budget for six months of directed-test writing. Photoniq's predictions let two or three verification engineers cover what would otherwise require eight — without sacrificing coverage confidence on custom datapath and memory subsystem logic.

See all use cases

Practitioner feedback

From the DV floor

We closed the last 8% of functional coverage in two days instead of two weeks. The ranked test list was immediately actionable — we ran the top five recommendations and hit every flagged coverage gap on the first pass.

Kenji R.
Verification Lead, AI chip startup

The yield impact estimates matched what our DRC tool flagged two weeks later. First time I've seen that correlation outside of tape-out post-analysis. It changes how we think about floorplan review in the verification phase.

Amara S.
Design Verification Manager, fabless semiconductor company

Your next tapeout doesn't have to plateau at 85%.

Angel-funded, Santa Clara-based team. Built by former EDA and RTL verification engineers.