Wiegand vs OSDP: The Test Plan a Buyer Should Demand
This article treats wiegand versus osdp test plan as one part of a working system rather than a product in isolation. Two technologies get compared in one sentence in a design meeting, and the sentence ends the discussion. Which of them carries the credential from the reader to the controller is settled, the door schedule is issued, and the meeting moves on to lighting. The consequence surfaces later, at a door, as a ticket that reads "card fails on 214" while the same card works at 213 and 215. By then the reader, the controller, the cable run, the credential encoding and the commissioning record sit in four different scopes, and nobody wrote down which of them was supposed to have proved what.
- Chart the interface — write down what hands over between lock, door and system at wiegand versus osdp test plan
- Trial the handover — record the sequence that proves wiegand-vs-osdp works end to end
- Retain the trace — store the commissioning evidence with the file for one is wiring convention
- wiegand versus osdp test plan runs this sequence, and wiegand vs osdp is the record it produces.
That gap is what this guide closes. It is written for the systems integrator who will sign acceptance, the access-control engineer who owns the panel, and the IT lead who owns the network the controller hangs off. It treats the reader-to-controller path as the object under test, because that is where the two options differ in a way that changes what can be proven afterwards. A legacy wiring convention carries a formatted code over a data line and stops there: no device identity, no acknowledgement, no error signal, no way for the controller to ask the reader a question and receive an answer. An addressed, bidirectional protocol can carry the code and, around it, an address, a reply, a sequence number, a request for the reader's own status, and — where the specific implementation and licence support it — a protected session between two named endpoints.
Direct answer. A wiegand vs osdp test plan is the written set of proofs a buyer attaches to the reader-to-controller path before doors are signed. It names the path boundary, lists the assumptions each party is entitled to hold about the other, and maps every assumption to a numbered test with a pass condition and an evidence record. The tests run in five gates — desk verification, bench run, installed door, fault injection, acceptance pack — covering payload fidelity, format agreement, identity binding, denial handling, cut-wire and power-loss behaviour, loss of the server, clock reconciliation, record content and handover evidence. A plan that cannot produce the five standard answers (what is sent, who owns the clock, what is written down, what happens when something is unavailable, who owns the fix) has not been written yet.
The scope is deliberately narrow. This is not a comparison of the two approaches on merit, not a cost model, and not a product selection. Where a specific SKU is involved, protocol implementation, credential format, event reporting and buffer depth are confirmed on the dated product page for the specific SKU, and — where the far side of the path belongs to another vendor — against that vendor's dated datasheet. Two neighbouring questions are handled elsewhere: Security Questions to Ask About Wiegand vs OSDP takes the threat-model view, and the Wiegand vs OSDP Requirements Document Template is the specification-side companion this plan is written to test against.
One is a wiring convention, the other is a protocol specification
Wiegand is a wiring convention between a reader and a controller. It carries a data line and usually a power line and a ground, and the credential leaves the reader as a burst of clocked pulses on that data line. What the controller does with those pulses is not defined by the convention itself: each panel vendor decides which code widths it will decode and how it interprets the parity bits. The convention says nothing about encryption and says nothing about device identity. A reader on a legacy path cannot introduce itself, and the controller cannot ask it anything. The range this procurement eventually buys from is the smart door locks.
OSDP is different in kind. It is a written protocol specification with an addressed, bidirectional message structure, ordinarily carried over a half-duplex RS-485 pair. Devices on the bus hold addresses, the controller addresses a device, the device answers, and the exchange carries sequence and integrity information so that a damaged or duplicated frame is visible rather than silent. Whether any given product implements it, at which level of the specification, and with which optional features enabled, is a per-SKU question to confirm on the dated product page and against the other party's datasheet.
| Element on the path | Legacy wiring convention | Addressed bidirectional protocol |
|---|---|---|
| Credential identity | A formatted code on a shared data line, width agreed informally between two vendors | A code carried inside a framed message that names its source |
| Device identity | None on the wire; the input channel is the identity | An address on the bus, assigned at commissioning |
| Direction | Reader to controller for credential data; light and buzzer ride separate lines | Both directions, controller may poll and reader may raise events |
| Error visibility | A damaged burst may still decode to a valid-looking code | Sequence and integrity fields make a damaged or duplicated frame visible |
| Several readers | One channel per reader; sharing needs extra hardware | Several devices may share one run, each with its own address |
| Session protection | Not part of the convention | A protected-session mode exists in the specification; implementation is per-vendor and per-SKU |
| What the reader can report | Credential presented, and little else the controller listens to | Credential presented plus device events, depending on implementation |
| Behaviour when the link dies | Whatever the controller's input does when the line goes away | Whatever the protocol's timeout and error handling do, per implementation |
Every row in that table is a difference in what can be proven, not a difference in what is better. The wiring convention is not wrong. It is a system whose failure modes are invisible by construction, so the proofs you demand have to be shaped differently. A test plan written for one path and pasted onto the other produces false confidence, which is worse than no plan because it closes the ticket.
What actually crosses the path, and what each side believes
sequenceDiagram
participant Cardholder
participant Reader
participant Controller
participant Server
Cardholder->>Reader: Credential presented
Reader->>Controller: Payload framed with source address
Controller->>Server: Credential check against held database
Server-->>Controller: Decision and current credential state
Controller->>Reader: Release indication for the indicator
Reader-->>Controller: Acknowledgement carrying sequence number
Controller-->>Server: Event written with time, source and decision
The payload is not the credential. The card carries a native identifier; what leaves the reader is a formatted representation of it — a code of an agreed width, often with facility bits and parity added. So every reader-to-controller path makes two separate assertions at once: that the payload identifies the credential in the reader's hand, and that the payload originated at the reader mounted on this door. Access control depends on both, and the second assertion is the one that the data path either carries or cannot carry.
The failure that makes this concrete is the format mismatch. A reader emitting one code width into a controller expecting another can still produce something the controller will decode, because the narrower payload sits inside the wider frame with parity in the wrong place. The controller acts on it, a door opens, and the event log shows an entirely ordinary grant — no error, no alarm, nothing for an operator to notice. The mirror case is equally quiet: a genuine credential arriving through a marginal installation with one bit lost is rejected, and the first ticket raised names the reader rather than the cable.
The second thing to settle before any test is written is where the path ends. The device that the controller talks to is often not the device bolted to the door. It may sit behind a converter, a sub-panel, or a bridge that puts the legacy code onto a network. Each of those moves the point at which identity is lost, and the test plan must name its own endpoints or it will prove the wrong segment. Write the two endpoints on the first page, with the door they serve and the panel channel or address they occupy.
The assumption ledger: what each party is entitled to believe
The useful way to write this plan is not as a list of features. It is as a list of assumptions, each with a party entitled to hold it, an artefact that would support it, and a test that either proves it or forces the record to say it is not held. The ledger below is the spine of the plan; the gates that follow simply execute it.
| Assumption | Who may hold it | Evidence that supports it | Test |
|---|---|---|---|
| The payload received identifies the credential presented | Controller | Encoding procedure on both sides, same code width and parity | T-01, T-02, T-03 |
| The payload arrived from the reader at this door | Controller | Physical binding recorded at commissioning | T-04, T-05 |
| The reader is the commissioned device, with expected firmware | Controller | Address or channel assignment record | T-05, T-06 |
| Nothing altered the payload in transit | Controller | Protected session where the implementation and licence support it, otherwise a written statement that the assumption is not held | T-07 |
| The reader raises the events it claims to raise | Controller | Event list in the vendor's dated datasheet | T-08 |
| The controller answers the reader's indicator requests | Reader | Panel behaviour documented for that model | T-09 |
| The credential set the controller holds is current | Reader | Database location and refresh behaviour documented | T-10 |
| The decision belongs to the controller, not the reader | Reader | Panel is the decision point; the reader reports | T-10 |
| The event time is the time the event happened | Both | Clock ownership declared, drift measured | T-11 |
| The event is recorded somewhere that survives the door | Both | Audit log and event memory defined as separate things | T-12 |
| The door's power-loss state is the approved state | Door owner | Approval record, not the product | T-13 |
| The escape side releases by itself, and the path only reports it | Controller | Exit device behaviour, with the request-to-exit signal treated as a report | T-15 |
Two features of that ledger deserve emphasis. Most rows belong to the controller and are not the controller vendor's to grant — they are the buyer's to verify, and the vendor's assurance is not a substitute for the test. And the rows that look like technicalities are the ones that most often have no answer at all: the drift figure, the buffer depth, the power-loss state. A plan that tests only the happy path will pass on a system whose failure behaviour has never been observed.
The ledger is also the honest place to record what nobody is entitled to assume. Where a path carries no integrity protection and no device identity, the correct entry is not a reassuring sentence but a line reading that the assumption is not held, with the compensating control named. That is a finding, and findings are what a test plan is for.
Why the ledger and not the datasheet is what you test
A datasheet describes one device in isolation. Every row of the ledger describes two devices acting together, and nothing in either datasheet can settle it. Three archetypes account for most of what goes wrong on this path, and each one is invisible from the paperwork.
The first is the silent misdecode: a code arrives malformed, is decoded as something valid, and grants access to a credential nobody presented. There is no fault, no alarm and no entry that looks wrong, so the only thing that catches it is a test that deliberately presents a code in the wrong format and insists on a rejection.
The second is identity drift. During construction the reader at a door is swapped, the channel or address is re-assigned, the original device is kept as a spare, and the configuration record is never reconciled. From that point the panel is deciding on events from a device that is not the one on the door, and the door is not the one in the schedule.
The third is the assumption nobody wrote down. The clock owner, the event buffer depth, the fail state on power loss: never specified, therefore never tested, therefore discovered on the night of the outage. That is why the plan begins at the desk and ends at a signed evidence pack rather than living as a commissioning checklist in a technician's pocket.
Gate A — Desk verification, before a reader is ordered
The plan runs in five gates, and each one closes a different class of uncertainty. A gate is not a stage of construction, it is a place where a specific kind of evidence is produced, and the plan loops back whenever a ledger row is left unproven.
flowchart TD
GA["Gate A · desk verification"] --> GB["Gate B · bench run"]
GB --> GC["Gate C · installed door"]
GC --> GD["Gate D · fault injection"]
GD --> GE["Gate E · acceptance pack"]
GA --> Q{"Any ledger row still unproven"}
GB --> Q
GC --> Q
GD --> Q
Q -->|"yes · name the missing evidence"| GA
Q -->|"no · every row proved or recorded as not held"| GE
This gate produces the document set the later gates are run against. Each question below changes which tests are even possible, so an unanswered question at Gate A is a blocker, not a gap to be closed on site.
| Question to put in writing | Why it changes the plan | Evidence that closes it |
|---|---|---|
| What code format, width and parity does the reader emit, and how is it derived from the credential? | Decides the shape of T-01 to T-03 | Dated reader datasheet plus the encoding procedure |
| What does the controller do when it receives a format it does not accept? | Decides whether a mismatch is a rejection or a silent grant | Dated panel datasheet, stated behaviour |
| Is the path addressed, or is identity carried by the wiring channel? | Decides whether T-04 and T-05 are testable at all | Dated datasheet from both parties |
| Which specification version, and which optional features does each side implement? | Determines which ledger rows can be closed | Written conformance statement naming the version |
| Is a protected session available between these endpoints under this licence? | Decides whether the in-transit assumption can ever be held | Written answer from the supplier and the panel vendor |
| What events does the reader raise, and what does the controller write for each? | Decides T-08 and T-12 | Reader event list and the panel event dictionary |
| What is the reader's event buffer depth, and what flushes it? | Decides what is lost when the server is unreachable | Dated datasheet and the panel's flush behaviour |
| Whose clock is authoritative for a credential event? | Decides the method and the tolerance of T-11 | Written answer naming the clock owner |
| What is the declared state of the door and the reader on power loss? | Decides T-13 | Dated datasheet plus the door's approval |
| How is a denial reported, and does the reason code reach the record? | Decides T-10 | Panel event dictionary |
Answers of the form "it works with our panel" are not answers. Ask for the document, the version and the date. Where a capability depends on a specific implementation rather than on the standard, ask the supplier to state it in writing for the exact model — CONFIRM-PER-SUPPLIER — rather than accepting a family-level assurance that will be read across every SKU on the schedule.
Gate B — The bench run, before the door exists
Reader, controller, the actual cable type, and the same power source, on a bench, with the real credential set in hand. The bench is where the expensive discoveries are cheap, because a bench can be rewired in minutes and a door cannot.
| Bench case | What you do | A correct result looks like | What it proves |
|---|---|---|---|
| Payload repeatability | Present one credential many times, including at the near edge of the read range | Every read yields an identical payload | The installation does not manufacture variation |
| Deliberate format mismatch | Encode a credential in the other supported width and present it | The controller rejects it and says so | A mismatch is a rejection, not a silent grant |
| Marginal read | Present a credential worn, misaligned or partly withdrawn | A clean rejection, never a different valid payload | The read margin does not generate false identities |
| Reader power cycle | Power the reader while the controller runs | The controller's state is defined, not merely unchanged | The controller's recovery behaviour on the reader side |
| Data line opened | Open the data line at the reader end, then at the controller end | The controller's input state, any alarm and any event are documented | What the path does when it is broken |
| Controller powered down | Remove controller power with the reader live | The door takes its approved state; the reader's behaviour is known | The power-loss case before it is an emergency |
| Controller reboot, no server | Reboot the controller with the server unreachable | Credentials are still present, the door still decides, the buffer accepts writes | Whether the door depends on the server to work |
Two of these deserve to be run twice: the deliberate format mismatch, and the power-cycle case. A single pass on a bench that has just been assembled tends to capture the configuration the technician intended rather than the configuration in front of him.
Gate C — On the installed door, with the real cable
Only now does the path acquire a door, a strike, a door position contact, an exit device and a cable route that may be longer, shared or badly earthed relative to the bench. Address or channel assignment is fixed here, and it must be recorded as it is made, not reconstructed afterwards from memory or from a panel export.
| Installed-door case | Pass condition | Record to keep |
|---|---|---|
| Mapping verification | The panel's channel or address for this door matches the physical door and the schedule | Signed mapping record, one line per door |
| Credential set parity | Every credential behaves the same at the issuing desk and at this door | Credential reconciliation list |
| Repeat presentation | Two presentations of one credential produce two events with the correct source and time | Event extract with the raw fields shown |
| Door position contact | The contact reports open and closed as states, and the panel records which one it saw | Contact configuration note |
| Exit-side behaviour | The escape side releases by itself; the exit signal reaches the system as a report | Exit device wiring note and observed behaviour |
CONFIRM-WITH-LOCAL-EGRESS — whether the assembled door's behaviour on loss of power and on a fire-alarm signal is acceptable for the role that door plays in this building, and which authority signs that off, is a question for the approving authority and the local fire-safety regime rather than for a product page. Put it in writing before the plan is signed.
The door position contact is worth a sentence of its own because it is routinely over-read. It reports a state, not a cause: it cannot tell you whether the door was opened by a valid credential, by the handle, by a wedge or by someone who had no business there. Anything downstream that infers a cause from that bit has been built on an assumption nobody wrote down.
Gate D — Fault injection: what you cut and what you watch
stateDiagram-v2
[*] --> Nominal
Nominal --> Degraded : error or unacknowledged frame
Degraded --> Nominal : frame accepted and acknowledged
Degraded --> Absent : link removed or power lost
Absent --> Nominal : link restored and re-established
Absent --> [*]
Fault injection is the part of the plan that gets skipped because the doors are open and the trades are on site, and it is the part that determines what happens at three in the morning. Each fault below is injected deliberately, on a door that is not in service, with the result written down while it is being watched.
| Fault injected | What you should observe | What to write in the log |
|---|---|---|
| Data line opened at the reader end | Defined controller input state, any alarm, any event | The observed state and the recorded event, verbatim |
| Data line opened at the controller end | Same, and whether the two ends behave differently | The difference, if there is one |
| Reader power removed | The reader's behaviour and the door's state | Observed door state and the reason it is acceptable |
| Controller power removed | The door's fail state | The fail state and the approval reference |
| Supply brown-out, then restore | Whether the reader re-negotiates or stays silent, and whether events survive the outage | Events lost across the outage, counted |
| Controller rebooted, server unreachable | Whether credentials remain, whether the door still decides, whether the buffer accepts writes | The answer, and the buffer depth in use |
| Server unreachable, controller running | Whether the door still decides, and where events go | Where the events went and whether they arrived later |
| Credential revoked while the reader is live | How long the stale credential still works | The observed revocation latency |
| Reader replaced with an identical unit | Whether the controller notices, and what the record shows | The record as written after the swap |
CONFIRM-WITH-LOCAL-FIRE-INTERFACE — how a fire-alarm signal reaches the lock and how release is coordinated with the escape side is an interface between two systems, and the rule that release is never the lock's own decision belongs to the approving authority and the local fire-safety regime. Confirm which authority signs the interface off before the plan is signed.
The offline case is the design, not the exception, and it has to be written down before the integration ships. If the honest answer to "what happens when the network, the controller or the server is unavailable" is that the door fails locked, then that is an approval question for the owner of the door and the approving authority — not an engineering choice the integrator can close inside a test report.
Gate E — The acceptance pack
Acceptance is a document, not a walk-around. The pack is what a reader of the file six months later will use to decide whether the behaviour seen on the night was a fault or the design, and it is what an investigator will ask for.
| Pack item | Why it belongs in the pack | Who signs |
|---|---|---|
| Completed test log with pass conditions met | The primary record of what was proved | Integrator and the engineer's representative |
| Assumption ledger, each row marked proved, disproved or not held | Converts the plan into a statement of what the system is | Access-control engineer |
| Configuration record mapping channels or addresses to doors | Reconciles the physical installation to the panel | Commissioning engineer |
| Credential encoding record | Proves the code on the wire is the code intended | Issuing authority |
| Clock reconciliation record | Fixes the timeline the events will be read against | IT lead |
| Event, buffer and retention definitions | Separates the audit log from the event memory | System owner |
| Residual-risk list with a named owner per entry | The place where "not held" becomes somebody's job | Client or system owner |
| Dated datasheets and the supplier's written answers | The evidence base for every protocol and capability statement | Both vendors |
The five answers the plan has to produce
erDiagram
READER_PAYLOAD {
string credential_code
string code_width
string parity_scheme
}
CONTROLLER_EVENT {
string door_id
string decision
string reason_code
datetime event_time
}
SERVER_AUDIT {
string event_id
string source_device
string reader_identity
}
EVENT_MEMORY {
int buffer_depth
string flush_condition
}
READER_PAYLOAD ||--|| CONTROLLER_EVENT : arrives as
CONTROLLER_EVENT ||--|| SERVER_AUDIT : is written as
CONTROLLER_EVENT ||--o{ EVENT_MEMORY : buffers until flushed
The plan is not finished when the tests have run. It is finished when the five standard answers exist in writing, each traceable to a test.
| Question | Where the answer comes from | What a finished answer looks like |
|---|---|---|
| What does each side send, and what happens when it does not arrive? | T-01 to T-08 and Gate D | A named degraded mode for the path: the controller's last-known state for that input, the door's behaviour, and where the event lands |
| Who owns the clock? | T-11 | One named clock as authoritative for credential events, a measured drift figure, and the tolerance the system tolerates |
| What is written down, where does it live, and who can read it? | T-12 | The audit log and the event memory named as separate things, with location, retention and reader roles for each |
| What happens when the network, the controller or the server is unavailable? | Gate D rows and T-13 | The door's behaviour stated, with the approval reference, and the buffering that keeps events recoverable |
| Which side owns the fix? | The escalation table below | A boundary drawn in advance rather than negotiated during an outage |
CONFIRM-WITH-LOCAL-PRIVACY — who presented which credential at which door and at what time is personal data in most jurisdictions, and the retention period and the reader roles for that record are set by the data controller and the local regime. Confirm the retention setting, the access list and the lawful basis with the local data-protection contact before the record is put into service.
Who gets the call when a test fails
The plan cannot tell you in advance whether the fault will be a reader fault, a controller fault, a cable fault or a configuration fault. It can tell you which test isolates the fault fastest, and that is what keeps an outage from becoming a four-party meeting. The contractual allocation of the fix is a separate question — Who Owns Wiegand vs OSDP: Supplier, Integrator or Client settles that — but the operational sequence below has to exist whether or not the contract does.
| Symptom at the door | First test that isolates it | Whose scope owns it | Evidence that closes it |
|---|---|---|---|
| One credential denied, others accepted | T-02, then T-03 | Issuing authority, then the encoder | Encoding record corrected and re-tested |
| Every credential denied at one door only | T-04, T-05 | Commissioning engineer | Mapping record reconciled to the door |
| Access granted for the wrong credential | T-01, T-02 | Reader and panel vendors together | Format agreement confirmed in writing |
| Events missing from the record | T-08, T-12, then the Gate D server row | Panel vendor, then the system owner | Buffer depth and flush behaviour documented |
| Times in the record disagree with real time | T-11 | IT lead and panel vendor together | Clock ownership recorded and drift corrected |
| Door state wrong on power loss | T-13 | Door owner and approving authority | Approval reference on file |
| Door locks and does not release on the escape side | T-15 and the exit device note | Door owner and approving authority | Interface confirmed against the local regime |
Where the wider credential ecosystem sits relative to this path
The consumer-facing ecosystem names — Matter, Thread, Zigbee, Z-Wave, Apple Home Key, HomeKit and UWB — are provisioning and radio-layer terms, not reader-to-controller protocol terms. Confirm on the dated product page for the specific SKU whether a credential provisioned that way can be presented to this controller at all, and what the controller is given when it is. What the controller ultimately sees on the wire is still a code in a format both ends must agree on, and that agreement is the only part of the question this plan covers.
Turning the plan into a demand a supplier can answer in writing
A test plan that arrives after the purchase order cannot be enforced. The equivalent of the plan has to be in the specification and the request for quotation, in language that a supplier can answer with a document rather than with confidence.
| Demand to place | The wording that closes it |
|---|---|
| Declare the code format, width and parity on the reader side | A table in the response, confirmed for the exact model |
| Declare the controller's behaviour on an unrecognised format | A sentence, not a silence |
| Name the specification version and the implemented feature set | A conformance statement with a date |
| State whether a protected session is available for the quoted pairing | An explicit yes or no, with the licence named |
| State the event list and the buffer depth | A table, and a flush condition |
| State the power-loss behaviour of the door assembly | A sentence tied to the fail state |
| State the reader's mounting and cabling requirements | A drawing, with maximum run length and termination stated |
Our own catalogue is a standalone lock class rather than an enterprise credential reader, and nothing here claims that any product in it implements an addressed reader protocol; the confirm-on-the-dated-product-page rule applies to every model-specific statement in this guide. Where a project pairs one of our locks with a third-party panel, the panel-side facts are taken from that vendor's dated datasheet and the lock-side facts from our dated product page, and the test plan is where the two are made to meet. That pairing, rather than either brochure, is the deliverable.
What this plan deliberately leaves
A test plan that tries to cover the whole door system proves nothing, because the failure modes of one path hide the failures of another and every case gets ambiguous. The boundary here is the reader-to-controller path and the two endpoints named on the first page. Four adjacent subjects sit outside it and each needs its own instrument.
Credential issuing and encoding are an input, not a component. The encoding procedure is what makes T-03 possible; the plan does not test the desk, the encoder or the issuing workflow, and it should not, because a mismatch between a desk tool and a door is found in minutes if the encoding record exists and is compared.
The door hardware is a separate path. The lock, the strike, the door position contact and the exit device arrive on the same cable in some installations, which is exactly why the path boundary has to be written down — otherwise the panel's view of a door-open bit gets confused with the panel's view of a credential event. The controller-side exposure of those signals has its own treatment in Door Controller Interfaces: The Test Plan a Buyer Should Demand.
Network segmentation and server-side behaviour are a third subject and a fourth. Whether the panel's traffic is reachable, and what the application does when its own dependencies are unavailable, are decided at the network and platform layer, not at the reader.
Retention and lawful basis are a data question rather than a wire question, and they are handed to the data controller under the local confirmation line rather than answered here.
Write the exclusions down at the front of the plan, with a named owner for each. The plan then stops being a document about everything and becomes a document that can be executed on a door in an afternoon — which is the only kind that gets executed.
What to carry into the handover file
The pack from Gate E is a subset of a larger handover, and the rest of that handover has its own discipline — Documenting Wiegand vs OSDP for Handover and Audit covers the record set the client's own auditor will expect, while Door Position Contacts: The Test Plan a Buyer Should Demand and Request to Exit Devices: The Test Plan a Buyer Should Demand run the same gate structure over the two adjacent signals that arrive on the same cable. The published commissioning material — Smart Lock Commissioning Checklist and Smart Lock Testing After Installation — is where a gate-based plan usually meets a line-based one, and the reconciliation between the two is worth doing before the first door.
It is worth pairing this with smart lock supplier due diligence checklist. That is the same question wiegand versus osdp test plan has to answer here.
A one-page plan you can send this week
Everything above reduces to fifteen cases, and this is the sheet to hand to a supplier or to run on a bench.
| ID | Case | Pass condition | Evidence to keep |
|---|---|---|---|
| T-01 | Repeated reads of one credential | Identical payload every time | Payload extract |
| T-02 | Credential presented in the other code width | Rejected, with a stated reason | Event extract |
| T-03 | Payload reconciled against the credential record | Code matches the intended credential | Reconciliation list |
| T-04 | Marginal or worn credential | Clean rejection, no alternative payload | Event extract |
| T-05 | Mapping of panel channel or address to the physical door | One-to-one and signed | Mapping record |
| T-06 | Unconfigured device placed on the bus | Not silently adopted | Event and panel state |
| T-07 | Damaged or interrupted frame | Visible as an error on the path | Log extract |
| T-08 | Events raised by the reader | Each claimed event appears in the record | Event list |
| T-09 | Indicator request from the panel | Answered, with a defined fallback | Behaviour note |
| T-10 | Credential revoked while the reader is live | Revocation takes effect, with observed latency | Timing note |
| T-11 | Clock comparison across reader, panel and server | One owner named, drift measured | Clock record |
| T-12 | Record content for one event | Time, source, door, credential, decision, reason present | Raw record |
| T-13 | Power removed from the assembly | Door takes the approved fail state | Approval reference |
| T-14 | Server unreachable | Door still decides, events buffer and are recovered | Buffer note |
| T-15 | Exit device operated | Escape side releases; the signal is reported, not commanded | Wiring note |
Where the buyer ends up after the last gate
A reader-to-controller path is not finished when a card opens a door. It is finished when the assumptions both sides were relying on have been written down, tested, and either proven or explicitly recorded as unheld — and when the offline case has been watched rather than imagined. If you want that plan applied to a specific door population and a specific pairing of hardware, send the door schedule and the panel model through the request a quote route and we will work through the Gate A questions with you before any model is nominated. The sample route, order samples, lets the bench cases in Gate B be run on the real article rather than on a datasheet description of it.

