Lunaris / Evidence you can
inspect.
Evidence you can
inspect.
A validation strategy for a research UE: test mechanisms in isolation, reproduce signals, exercise protocol paths, and verify selected live network profiles with independent observations.
RELEASE 17 BASELINE / FR1 PAIRED FDD / TERRESTRIAL + NTN
The strongest accepted RF result
20 trials. 400/400 probes.
The documented OCUDU campaign on 25 October 2026 accepted 16 candidate RF trials with 320/320 traffic probes, followed by four post-deployment RF smoke trials with 80/80 probes. The selected setup used a single-chain USRP X410, 5 MHz bandwidth, 15 kHz subcarrier spacing, FR1 paired FDD.
Terrestrial and GEO-delay-emulated cases were run separately. The GEO cases combined physical RF with software propagation delay and received SIB19 assistance. The peer deployment included explicit PUCCH, SIB19 and shutdown fixes; these results apply to that deployment, not every stock OCUDU version.
| Mechanism | Selected profile | Trials | Probes |
|---|---|---|---|
| Repetition profiles | PUSCH 2 / PDSCH 2; three TN and one GEO-delay case | 4 | 80/80 |
| Configured grant Type 1 | TN 20/4 and 80/8; GEO-delay 80/8 and 160/8 profiles | 4 | 80/80 |
| CU/DU transport variants | TN/GEO-delay, plaintext and DTLS-protected F1-C | 4 | 80/80 |
| PUCCH repetition | TN and GEO-delay, two-slot HARQ-ACK repetition | 2 | 40/40 |
| Two-step random access | Eight-frequency MsgA: TN success; GEO-delay forced fallback | 2 | 40/40 |
| Post-deployment smoke tests | Four representative deployed profiles | 4 | 80/80 |
| Total accepted RF trials | 20 | 400/400 | |
Fourteen focused offline checks also passed. No UHD transport failure, UE error or access interruption was logged in these accepted trials. Earlier failed profiles remain part of the engineering record; passing probes do not erase a strict audit failure elsewhere.
Counts refer to individual campaign trials and their documented traffic probes. They are not throughput, latency, availability or reliability guarantees.
A second interoperability environment
Amarisoft over ZMQ.
ACCESS & TRAFFIC
A licensed independent peer.
Selected campaigns use licensed Amarisoft through a bidirectional software-radio ZMQ link and an isolated private/local core. Evidence includes four-step access, registration and IP traffic.
MSG3 REPETITION
Procedure and transport agree.
A 24 October 2026 two-slot Msg3 repetition case registered, completed three ping probes and released cleanly. UE redundancy versions 0 then 2 matched peer repetition count 2, final RV 2 and CRC OK; the ideal-channel audit was clean.
QUALIFICATION DISCIPLINE
Failures remain visible.
A single-Msg3 control passed its traffic/audit checks but correctly did not qualify as repetition 2. Broader repetition counts have grant-decode coverage; they are not all live-qualified.
Some Type 2 configured-grant profiles passed 20/20 traffic probes while failing strict periodic-UCI audits. The validation approach keeps these outcomes separate and uses bounded qualification profiles. Amarisoft two-step procedures also require an explicit, default-off SLIV compatibility setting. The campaigns described here do not establish Amarisoft RF or NTN two-step qualification.
How testing works
Four layers of confidence.
01 / UNIT
Check a mechanism.
C++ and Python checks cover coding, protocol state, timers, queues, security primitives, configuration rejection and lifecycle behaviour. Assertions remain active in Release builds.
02 / REFERENCE
Check against a separate derivation.
Specification-derived mappings, vectors and calculators check selected standardized structures. Optional MATLAB-generated fixtures are a separate reference source, not a normative compliance certificate.
03 / REPLAY
Repeat the signal path.
Saved IQ and metadata make acquisition, decode boundaries and failure cases reproducible without depending on a new live radio run.
04 / INTEGRATION
Exercise connected components.
Controlled local ZMQ peers exercise access and protocol paths. Hardware, OCUDU/core and Amarisoft campaigns have separate explicit procedures and independent evidence.
Examples from the reference catalogue
Go deeper than a ping.
- Fourteen FR1/FDD PDCCH profile geometries based on 3GPP TS 38.521-4 V17.4.0, including codeword digests, noiseless decode and wrong-RNTI checks. This is not a conformance-performance test campaign.
- Thirteen default-MCS PDSCH profiles expanded into 26 coded-bit-budget cases, and 20 FR1/FDD CP-OFDM PUSCH reference-measurement-channel rows with noiseless payload-recovery checks.
- 351 independently constructed direct-DFT PUCCH format 3/4 reference grids.
- Milenage, NEA2 and NIA2 vector checks; replay, integrity, COUNT and procedure-state checks complement primitive-level tests.
- Reference timing, orbit/geometry, mapping and scheduling calculators help separate standardized transmitted structures from implementation-specific receiver algorithms.
Reproducibility & review
Make the evidence
part of the deliverable.
The required CI baseline uses an audited test inventory. Missing, duplicate, unexpected, failed or skipped required checks fail the baseline. Optional backends and unavailable generated fixtures are reported separately. Runs retain JUnit results, full logs, the inventory, software revision and dependency records.
A paid validation campaign can agree its radio and channel profile, peer build, traffic, fault conditions and acceptance criteria before execution. Procedure logs, gNB/core observations and packet evidence are correlated so a traffic success cannot conceal a failed protocol requirement.
BEFORE THE RUN
Define the question.
Freeze the intended profile and success criteria. Establish which observations are needed and what would count as a failure.
DURING THE RUN
Observe both ends.
Correlate UE state, independent peer observations and user traffic. Preserve deviations and recovery behaviour.
AFTER THE RUN
Review a bounded result.
Deliver a reproducible report with passed cases, failed cases, limitations and the next action. Discuss detailed evidence under an appropriate agreement.
What this does not establish
Evidence has boundaries.
The published results do not establish 3GPP certification, calibrated RF or power performance, field availability, production real-time readiness or a live satellite link. Moving-orbit campaigns need a consistent time-varying delay/Doppler channel. Band-specific RF compliance is separate from generic NR processing.
The DTLS cases test CU/DU F1-C endpoint behaviour, framing and SCTP-AUTH integration. The UE does not terminate DTLS. These runs are not independent cryptographic validation or a full F1 impairment campaign.
Commission a campaign for your profile ↗