2026-10-01 11:15:07 +10:00
4 changed files with 436 additions and 0 deletions
--- a/knowledge/manuals/IM36K10E20-01EN.md
+++ b/knowledge/manuals/IM36K10E20-01EN.md
@ -0,0 +1,70 @@
 ---
 manual: IM 36K10E20-01EN
 title: OPC Interface
 edition: 2nd Edition, Sep. 2023
 version: R1.03
 status: from docs; OPC UA point and certificate forms compared with quick load on the test system 2026-10-01 (user configured a point in the Engineering Module, exported and diffed)
 ---
 # OPC Interface (R1.03)
 Engineering Module procedures for OPC Classic, OPC UA client and OPC UA PubSub. Concepts,
 setup files and the OPC servers: `IM36K10L45-01EN.md`. Everything here is under
 *Configuration → I/O processing → Equipment → (OPC DA client | OPC A&E client | OPC UA Client |
 OPC UA PubSub)*. CI Server must be running. Section numbers refer to the manual.
 ## 1. OPC Classic (§1)
 - DA: line (equipment node, process name — the manual says it must start with `OPC` for DA and
  A&E lines, `OPX` for OPC UA), station (OPC server node and ProgID, status item), group
  (deadband, requested update rate, status item), point (external relation, group, OPC tag,
  browse). Then connect items to points (Tutorial "Define points").
 - A&E: line, station (server, subscription filters), subscribed attributes; to show events in
  the current alarm overview, map event sources to items. External names must be unique per
  station.
 ## 2. OPC UA client (§2)
 Order: line → station → subscription → points (→ methods) → alarm components / event type
 filters / event nodes / template settings (only when alarms are wanted; no alarms are received
 until event nodes are configured).
 | Object | Key fields (dataset field) |
 |---|---|
 | Line | name ≤ 31, equipment node (*None* = this node), **process name starts with `OPX`, ≤ 11 characters** (`EQUIPMENT_MAN`) |
 | Station | name ≤ 19; *Host(:port)* + *Discovery* fill session/server name and endpoint, or type the endpoint (`SERVER_01_ENDPOINT`); redundant servers + *Active server item*; *On scan*; *General communication status item* (`STATUS_ITEM`); security policy and mode; anonymous or user/password; requested session time-out; alarm tab (request active alarms, store all alarms as SOE, severity → priority mapping) |
 | Subscription | publishing interval, lifetime and keep-alive counts, max notifications, priority, publish enabled, status item |
 | Point | external relation Input / Output / Input + Output; subscription; *Namespace*, *Node id type*, *Node Id* (or *Browse path*); read-only *Last value*, *Data type*, *Access level*, *Value rank*; array element; options (own sampling interval, monitoring mode, queue size); conversion (offset, step) |
 | Method | namespace, node ID or browse path, process areas |
 **Observed (2026-10-01)**: browsing in the point form stops at the item itself
 (`Objects/<section>/…/<item>`) and stores the item **object** (`{"nsid":"<n>"}`), which has no
 value: *Last value*/*Data type* stay empty and the point is never monitored (revised sampling
 interval 0, item status 13). Type the node ID of the value instead:
 `{"0":"Value","itnm":"<item>"}` (namespace `urn:Yokogawa:CI_SERVER_OPC_UA_SERVER`, type
 STRING). The form also fills `OPC_DISPLAY_NAME` and `OPC_TAG_INFO` (browse bookkeeping;
 not needed for quick load). ISS-033.
 ## 3. Certificates (§2.2)
 *OPC UA Client → Certificates*: **Trusted**, **Rejected** (server certificates arrive here),
 **Own certificate**. Trust = open the rejected certificate, tick *Trusted*, Apply; reject =
 untick. Restart the Engineering Module to refresh the folders. *Modify* the own certificate
 (name, organisation, country, domain or IP, RSA key strength, signature algorithm, validity).
 **CI Server OPC UA client ↔ CI Server OPC UA server (§2.2.6)**: trust the server certificate
 on the client; copy the client's own certificate
 (`%TLS_ALLUSERS_PATH%\pki\opxuac\pkiclient\own\certs`) to the server's
 `…\pki\opxuas\pkiserver\trusted\certs`; the manual also asks to tick *Use CI Server entity names
 for nodeids* in `opxuas.sup` and restart the server. On the test system the loopback worked
 **without** that option, using `itnm` node IDs (ISS-032).
 ## 4. OPC UA PubSub (§3)
 UDP publish/subscribe of Data Access values from the CI Server OPC UA server address space
 (OPC visible items). Objects: line (needs its own copy of `ServerConfig.xml` named for the
 instance), published datasets, station, writer groups + dataset writers (publisher), reader
 groups + dataset readers (subscriber), then *Upload dataset configurations to runtime*.
 Limits per line: 4 stations, 8 writer groups, 8 reader groups, 32 dataset writers, 32 dataset
 readers, 32 published datasets. Status mapping as for the server (NORMAL → Good; NOT_INIT,
 UPD_OFF, OFFSCAN, OFFLINE → Uncertain; BAD_VALUE → Bad). Not practiced (needs a PubSub peer).
--- a/knowledge/manuals/IM36K10L30-01EN.md
+++ b/knowledge/manuals/IM36K10L30-01EN.md
@ -0,0 +1,133 @@
 ---
 manual: IM 36K10L30-01EN
 title: BUS Service Engineering
 edition: 3rd Edition, Sep. 2023
 version: R1.03
 status: from docs; installed tools, setup files (dur.sup, durm_1_udp.sup, time_table.sup, fmfprt.sup, durutl.sup) and durm_start.cmd checked read-only 2026-10-01; durwfp verified; umhman, umhtestprog, umhmsg fail on the test system (ISS-030)
 ---
 # BUS Service Engineering (R1.03)
 BUS Service is CI Server's messaging layer ("DUR", Dynamic User Region): every CI Server
 process attaches to a shared memory area and gets a **queue** (mailbox) under its DUR name;
 processes talk by messages. It also does error-text management (UMH), node-to-node links
 (MDUR), time handling, printer management and process control utilities. Section numbers refer
 to the manual.
 ## 1. Processes (§1.2)
 | Process | Role |
 |---|---|
 | `durini` | Initialises DUR at start, then exits |
 | `durtim` | Runs every second: system time, summer/winter time, queue time-outs, checks that processes still exist (`DUR-W-REM_PROC, Process <x> exited unexpectedly`), clock queue |
 | `durprt` | Writes intercepted messages to `tls\lst\DURPRT.out` (debugging; leave running) |
 | `umh` / `umhlog` | Build error texts from codes (`tls\dat\messages.umh`) and write them to the message log (`umhlog.log`) |
 | Optional | `dur` (DUR display utility, interactive), `durbat` (start processes on request, also remote), `durclq` (finer clock-queue resolution: `durclq 50` = 50 ms), `durdbg`, `durutl` (logical addresses, node time sync), `durecs` (external clock), `dursim`, `durstp`, `durwfp`, `umhcr`, `umhlogtoo`, `umhman`, `umhmsg` |
 Process names: fixed (e.g. `UMH`), `PID<n>` when none given, or set with `-n <name>` (most
 CI Server executables, e.g. `opxuas -n opxuas`).
 ## 2. Messages and nodes (§2–§4)
 - Requests are asynchronous; CI Server makes them synchronous with a *reply required* flag. The
  **server** converts data between different computer architectures.
 - **MDUR** (§3): one `durm_<line>` process pair per network line (`durm_udp` + `durm_udp_slv`
  on Windows), started by `tls\com\durm_start.cmd`, configured in
  `tls\sup\durm_<line>_udp.sup`: `line`, `quality` (lower = preferred line), `io_timeout`
  (default 8000 ms, 8 checks), `portnr` (same on all nodes of one line; > 17000 advised),
  `max_snd`, `cluster`, `compress`, `connect_delay`, `max_poll_int`, `our_host_name`,
  `remote_set` (up to four nodes sharing one node number = redundant set). Redundant nodes
  only over TCP/IP. `durmdbg -q <durm> -f <file>` traces a line; always stop with `-s`.
 - `dur.sup`: node number, `host_node`, `rout_type` (ENDPOINT or ROUTER for multi-node
  routing), sizes (`pool_kbytes`, `dur_max_proc`).
 - **Logical addresses** (§4): node −1 in an address = look up a table (`durlnm.dat`), distributed
  by `durutl` from a master node listed with `lnm_node = …` in `durutl.sup`.
 **Installed (single node)**: `dur.sup` `host_node 0`, `pool_kbytes 250000`, `dur_max_proc 250`,
 `rout_type 2 ENDPOINT`; `durm_1_udp.sup` `line 1`, `quality 500`, `io_timeout 8000`,
 `portnr 17001`, plus keywords not in the manual: `sndportnr 17101`, `ip_adr_method IPv4`,
 `ENABLE_DURMS N, 0` (host-host encryption, §13); `durm_1` is running although there is no
 second node; `durutl.sup` has no active lines.
 ## 3. Messages and the message log (§5, §6)
 - Codes look like `FAC-S-ACRONYM` (S = S/I/W/E/F). Text comes from `messages.umh`; dynamic
  values are inserted by name; several codes can form one history entry.
 - `umhlog` writes to `tls\log\umhlog.log`; `-d <node>` also logs centrally on another node,
  `-q <printer queue>` prints too.
 - Tools: `umhtestprog` (tests the message path), `umhmsg <hex code> [<var> <value>]…`
  (issue a message from a script), `umhman` (look up a hex code, interactive),
  `umhlogtoo` (copy of all messages to a terminal). **On the test system** `umhman` crashes
  with redirected input, `umhtestprog` crashes, and `umhmsg` ends with "license expired"
  (ISS-030).
 - A message shown as `UMH Message Text Error: Message number <hex> (…)` means the text could
  not be formatted (e.g. a `%` in a PRINT text, ISS-025).
 ## 4. Time (§7, §8, §11)
 - `time_table.sup`: `AUTOMATIC` (installed) = BUS Service takes summer/winter times from
  Windows at start (scans ±15 years); otherwise `LST_OFFSET` (minutes, **positive = west**),
  `LST_NAME`, `LCT_NAME`, `CHANGE` lines; month/day abbreviations. Reload via the DUR display
  utility. Installed file still has the Central European example values (−60, MET/MEST); they
  are ignored while `AUTOMATIC` is set.
 - Node time sync with `durutl` on the master: `tms_params = <interval s>, <limit1>, <limit2>,
  <limit3>, <step2>, <step3>` and one `tms_node = <n>` per slave; beyond limit3 the slave is
  set at once and an error is logged. External clock (serial Wharton/Trimble/Arbiter) with
  `durecs` and `durecs.sup` (`ecs_params`). On Windows normally use Windows time
  (`system-check`, `w32tm`).
 ## 5. Process control utilities (§10)
 | Command | Use |
 |---|---|
 | `durwfp -p <name> [-t <s>] [-e]` | Exit 0 if the process is attached (waits up to `-t`, default 60; `-t 0` = check now); `-e` waits for it to go away. Verified: `umh`, `opcman` → 0, unknown name → 1 |
 | `durstp -m -p <name>` | Send a stop message (clean stop). `-k` kills, `-m -k` = stop, then kill after `-t` s; `-f` kills even with the common area locked (dangerous); `*` wildcard (`-p alm\*`). Never `durstp -k -p \*` |
 | `durbat` | Starts processes on request from other processes (also remote) |
 `durstp` stops CI Server processes: use only with the user's approval.
 ## 6. Printer management (§12)
 `FMFPRT` (printer manager, one per node) and one `FMFP_<device>` process per device (first 8
 characters of the device name must be unique). **Destinations** (logical, e.g.
 `SCREEN_CAPTURE`, `ALARMS`, `MESSAGES`, `DEFAULT`) map to **devices** (instances of an output
 type) with an optional backup device. Configured in `fmfprt.sup`; devices and destinations in
 datasets `PRINTER_DEV_DF`, `PRINTER_DEST_DF`, `PRINTER_TYPE_DF`, `PRINTER_OPER_DEST_DF`
 (job/device lists `PRINTER_JOB_OVV`, `PRINTER_DEV_OVV`).
 **Installed `fmfprt.sup`**: `QUEUE_LIFE 1800` s (stalled job expires), `MAX_JOBS_SHOWN 10`,
 `DEVICE_POLL 5` s, `MAX_BUSY 60` polls; output types `HARDCOPY` (`fmfphc`, Windows printing,
 Courier New 8 pt, 72 lines, landscape), `FILE` (`fmfpfl`, file name
 `TLS_%time:DDMMYY%_%seq:03%.lst` in `tls\lst`), `APPLICATION` (`fmfpap`, program set to
 Internet Explorer); fixed destinations `ALARMS`→`ALDPRT`, `AUDIT_TRAIL`→`ADTLOG`,
 `MESSAGES`→`UMHLOG`, `PRINTER_1`→`RPTMAN`, `SCREEN_CAPTURE`→`CWSSHOT`. `NODE_SHARE = <node|-1>`
 shares printers between nodes. Logging processes do not print by default (enable per process:
 `aldprt.sup LOGPRT`, `adtsys.sup PRT_DEV`, `umhlog -q`).
 The CI Service account must have the printers; `DEFAULT` uses that account's default printer.
 Clean-up after a crash: stop CI Server, empty `tls\lst\printing`, delete
 `tls\sav\fmf_prt_last_id.sav`, `fmfprt.sav` and `tls\dat\printer_job_ovv.*`, start.
 ## 7. MDUR encryption (§13)
 Host-host encryption (AES-256, certificates) between CI Cores: every node must use it; no
 enterprise configurations. Keys: `gen_key is_ca=1` once (CA key), `gen_key is_ca=0` on each
 node → `%TLS_ALLUSERS_PATH%\pki\mel\dur\key`; certificates with the interactive `cert_write`
 (CA certificate to `…\pki\mel\dur\crt\ca\ca.crt` on every node, node certificate to
 `…\crt\own`; create the folders first; log `tls\log\cert_write_log.log`). Then tick *Enable
 HostHost encryption* in each `durm_x_udp.sup` (installed keyword `ENABLE_DURMS`). Keep the CA
 key off the servers. Not practiced (single node; `tls\pki\mel` does not exist yet).
 ## 8. Adaptive queues (§14)
 Each process has a fixed queue (the *adaptive minimum*); for processes listed on the
 *Adaptive queues* tab of `dur.sup` (`ADAPTIVE_QUEUE_SETS = <process>, <percent>`), DUR may
 grow the queue temporarily by up to that percentage of the total queue area, avoiding overflows
 during bursts. Add node queues for multi-node systems and new processes (extra executors,
 DSS servers, equipment managers). In the performance monitor *Minimum KB*/*Maximum KB*/*Current
 KB* and a high-queue of 100 % show that a queue needed more space → raise its size. Total area
 exhausted → `DUR_F_NO_SPACE`.
 **Installed**: `ALM` 10, `ALH` 10, `WFLGDI` 20, `WFLCDI`/`WFLIDI`/`WFLADI`/`ITHEIX`/`ITHSER` 2,
 `DSSSDA*`/`DSSEVT*`/`OPCEXE` 1 %. Note: `OPCEXE` does not match the installed executor names
 `OPCEXE1`…`4` (no wildcard) — see ISS-026.
--- a/knowledge/manuals/IM36K10L45-01EN.md
+++ b/knowledge/manuals/IM36K10L45-01EN.md
@ -0,0 +1,182 @@
 ---
 manual: IM 36K10L45-01EN
 title: Data Access Engineering
 edition: 3rd Edition, Sep. 2023
 version: R1.03
 status: from docs; installed setup files (odbsrv.sup, opxsys.sup, opxuas.sup, ServerConfig.xml), access_start.cmd, PKI folders, processes and ports checked 2026-10-01; OPC UA client → own OPC UA server loopback verified with CI-Eng-Sample (station CIENG_UA); ODBC not working on the test system (ISS-008); OPC DA/A&E from docs
 ---
 # Data Access Engineering (R1.03)
 ACCESS Service connects CI Server data to other software: **ODBC** (SQL on datasets), **OPC
 Classic** (DA and A&E, client and server, DCOM) and **OPC UA** (client and server). Section
 numbers refer to the manual. Engineering Module forms: `IM36K10E20-01EN.md`. Hands-on:
 skill `data-access`.
 ## 1. ODBC (§1–§3)
 - Multi-tier: ODBC application + driver manager + **SimbaClient** driver on the client;
  **SimbaServer** + ACCESS Service ODBC engine (`odbsrv`) on the CI Server. Read and write of
  existing dataset records with SQL (API level Core, SQL level minimum); datasets cannot be
  created or deleted over ODBC. Needs an ODBC licence.
 - `odbsrv.sup`: `PORT` (1583), `USESSL` (default on; certificate
  `%TLS_ROOT_PATH%\tls\pki\odbsrv\SampleServerCertificate.pem`, key `SampleServerKey.pem`),
  `REMOVE_PADDING_SPACES`, trace (`TRACE`, `VERBOSE`, `TRACE_FILE`), hidden data
  (`TABLE_SUPPRESSION`, `FIELD_SUPPRESSION`, `FIELD_CONCEAL`, all on).
 - Client: install the VC++ 2019 redistributable and `SimbaClientDriverxx.zip` from the medium
  (`install.bat` as administrator); define a DSN with driver `CiSimbaClientDriver` (64-bit) or
  `CiSimbaClientDriver32`, server IP and port 1583. Before R1.03, uninstall the old Simba parts.
 - SQL wildcards `%`/`_` vs. MS Office `*`/`?`: escape `_` and `%` in names with `\`.
 - Common errors (§2.3): IM003 driver missing; IM002 DSN unknown; 08S01 SimbaServer not
  running / port mismatch / BUS Service not reachable.
 **Installed (test system)**: `odbsrv.sup` `PORT 1583`, `USESSL YES`, but the folder
 `%TLS_ROOT_PATH%\tls\pki\odbsrv` does not exist and no sample certificate is shipped, so
 `odbsrv` stops at every start (`CppServerAPI::Run() failed: InitializeSslCtx():
 SSL_CTX_use_certificate_chain_file() failed`, then `DUR-W-REM_PROC, Process ODBSRV exited
 unexpectedly`); port 1583 does not listen. Fix: put a certificate and key there, or set
 `USESSL = NO` (admin; not applied, ISS-008). `access_start.cmd` writes the ODBC registry
 defaults once (when `%TLS_DAT_PATH%\access.new` exists).
 ## 2. OPC Classic basics (§4)
 DCOM-based, Windows only. CI Server has DA client (`opxdac`), DA server (`opxdas12`, DA 1.0A
 and 2.0, custom + automation interfaces), A&E client (`opxaec`), A&E server (`opxaes`).
 - Quality: OPC DA quality goes to the **lower** 16 bits of the CI Server quality word, OPC A&E
  quality to the **upper** 16 bits. Item to OPC: CI Server quality low byte → OPC high byte;
  OPC low byte = offline → LAST_KNOWN, blocked → LOCAL_OVERRIDE, not initialised →
  NOT_CONNECTED, else GOOD.
 - Types: `VT_UI1/UI2/INT/I1/I2/I4` → Integer, `VT_UI4/UINT/R4/R8` → Real, `VT_BSTR` → String,
  `VT_BOOL` → Boolean.
 - Items, sections and objects appear to OPC servers (DA, A&E **and UA**) only with *OPC
  visible* (and every parent section visible).
 ## 3. OPC DA client (§5)
 - One client process per DA line (name given in the line), started/stopped with CI Server;
  stop one with `durstp -msk -t10 -p <client>`. Uses a save file in `%TLS_SAV_PATH%` named
  after the process; delete it (client stopped) if it gets inconsistent.
 - Server check every `OPXDAC_SRV_ALIVE` ms (default 30000); station status item: 1 running,
  2 failed, 3 no config, 4 suspended, 5 test. Not connected → mapped items OFFLINE. Group
  status item: 1 active / 0 not active (items offline).
 - Setup (`<client>.sup`, else `opxsys.sup`): `OPXDAC_SAFE_CON` (allow connecting to a CI Server
  DA server — **never to the own node or a DURM-connected node: deadlock**), `OPXDAC_PING`
  (ICMP check before DCOM, fast failure), `…_PING_TIMEOUT` (3 s), `…_DUR_QSZ`, `…_HBEAT`
  (100 ms), `…_ITMBUF`, `…_FLUSH_GRPQUE` (500 ms), `…_BROWSE_CACHE`, `…_DISABLE_BROWSING`,
  `…_CHECK_MAPPING`, `…_ASYNC_WRITE`, `…_USE_CANONICAL_DATA_TYPE` (default asks VT_R8),
  `…_IGNORE_BAD_QC_UPDATES`, `…_CLEAR_SAV_FILE`, `…_DISABLE_STATUS_UPDATE`.
 - Advice: one OPC server per client process (a DCOM time-out blocks the whole process).
  Quick-load templates: export `LINE_DF`, `STATION_DF`, `OPC_GROUP_DF`, `ITEM_DF`,
  `POINT_DF` of one station made in the Engineering Module.
 - **DCOM tunnelling**: run the client on a PC in the OPC server's domain (CI Server with BUS,
  DATABASE, EQUIPMENT, ACCESS licence) connected to the host by MDUR (`durm_<n>_udp.sup`,
  port ≥ 20000, `dur.sup` host node = CI Core), *use DCOM tunnelling to host* on the client's
  setup file; on the host add the line with *Equipment node* = tunneller node. Use `intmon`
  node monitoring with *Set slave items offline when node unavailable*.
 ## 4. OPC DA server (§6)
 - COM server `opxdas12.exe`, ProgID `Yokogawa.CIOpcDas.<version>`, DUR name `OPXDAS`; started by
  COM on the first client connect; stop with `durstp -p opxdas` (clients get a shutdown).
 - Namespace = sections/items/sub-items (hierarchical or flat), filters (type, access, VB LIKE
  pattern). *Cache read* is fast and correct; *device read* blocks — use only after queue
  overflow. Timestamp = time written into the server cache.
 - `opxsys.sup` *Data Access Server* tab: `OPXDAS_DUR_QSZ` (100 kB), `…_HBEAT`, `…_ITMBUF`,
  `OPXDAS_USER` (log on as a CI Server user to apply its authorisation; empty = full access,
  fastest), `OPXDAS_DCOM_TUN`, `OPXDAS_IGNORE_FTQ`. Installed also: `OPXDAS_SET_LOG_LEVEL`
  (not described).
 - Remote access needs DCOM configuration (`dcomcnfg`: authentication Connect, impersonation
  Identify, access/launch permissions for an OPC account; identity of the server = that account),
  `secpol` (*Create permanent shared objects*, *Network access: Classic*), firewall rules for
  `opxdas12.exe`, `opxdac.exe`, `opcenum.exe`, `mmc.exe` and DCOM — not needed where the IT
  Security Tool is applied. Same account/password on both sides in a workgroup.
 - Troubleshooting: `OCSRegistryCheck.exe` (GUI; installed), `opxdas12 /RegServer`,
  `opccommonsetup` / `opcenum /RegServer` / `opcenum /Service` (set the service to automatic),
  `regsvr32 opcproxy.dll` and `opccomn_ps.dll` in `%WINDIR%\SysWOW64`. Test locally first.
 ## 5. OPC A&E client and server (§7, §8)
 - Client `opxaec` (default name `OPXAEC`, `-n` for more), setup `<name>.sup` or `opxsys.sup`
  (`OPXAEC_SAFE_CON` — same deadlock rule, `…_DUR_QSZ`, `…_SRV_ALIVE`, `…_PING_TIMEOUT`,
  `…_TRACE`); maps condition events to items; one server per client advised.
 - Server `opxaes`, ProgID `Yokogawa.CIOpcAes.<version>`, DUR name `OPXAES`; publishes items
  with alarm monitoring as conditions; `OPXAES_ACK_ALLOWED` (acknowledge from clients),
  `…_ALM_REC_DELAY` (30 s), `…_BROWSE_PREP_INI`, and `OPXAES_EC_ATTR_*` switches for the
  CI Server-specific event attributes (old/new value, quality, status, alarm state, item
  description, units, node, group, number, sub-number). Installed: all on.
 ## 6. OPC UA client (§9)
 - OPC UA TCP only; security by X.509 certificates (application authentication), user
  anonymous or name/password; modes None/Sign/SignAndEncrypt; policies Basic256Sha256,
  Aes128-Sha256-RsaOaep, Aes256-Sha256-RsaPss recommended (None, Basic128Rsa15, Basic256 not).
 - Data types supported: Boolean, (S)Byte, (U)Int16/32, Float, Double, String and others as in
  Table 9.4.2-2; 1-D and 2-D arrays by element. OPC UA status → CI Server quality: 0 good,
  1 bad, 2 uncertain.
 - A&C: external alarms from ConditionType and derived types (ALARM/NORMAL events), other
  events INFORMATIONAL; filters by event type, component, event node; severity → priority
  table (901–1000 → 1 … 1–100 → 15), per station; don't manage
  NonExclusiveLimitAlarmType. Only `NodeDeleted` model changes are handled.
 - Redundancy: no client redundancy (use HAC); servers: transparent, or non-transparent cold
  backup with two endpoints (*Active server item* 0/1/2).
 - Engineering objects: line (starts a client process; **name must start with `OPX`**, max 11
  characters), station (= session), subscription, point (= monitored item), alarm component /
  event type filter / event node / template settings, method.
 - Certificates: the client process creates `<host>_<PROCESS>.der` in
  `%TLS_ALLUSERS_PATH%\pki\opxuac\pkiclient\own\certs` (also listed in dataset
  `OPCUAC_PKI_CERTIFICATES_DF`); server certificates arrive in `…\pkiclient\rejected` and must be
  trusted (Engineering Module, or move to `…\pkiclient\trusted\certs`); folders `deleted`,
  `rejected` added by CI Server. The server must trust the client's certificate too.
 - Setup (`<process>.sup`, else `opxuac.sup`): `OPXC_DUR_QUEUE_SIZE` (10000 kB), heartbeats,
  buffers, `OPXC_COM_STAT_CHK_INTERVAL` (10 s), `OPXC_STATION_CON_TRY_INTERVAL` (30 s; observed
  retries every ~35 s), `OPXC_TIMESTAMP_OPTION` (SERVER/SOURCE), `OPXC_LOG_LEVEL`, service and
  session time-outs, `OPXC_SERVICE_COMPLIANCE`. Log file `%TLS_LOG_PATH%\<PROCESS>.log`.
 - Quick load (§9.10–9.11): `OPCUAC_LINE_DF`, `…_STATION_DF`, `…_SUBSCRIPTION_DF`, `…_POINT_DF`
  (+ alarm and method datasets). Dataset `OPCUAC_TAG_DF` browses a connected server
  (filter on `OPC_STATION_NAME` and `OPC_TAG_PATH`).
 - Installed: `opxuac_mon` (a monitor instance) is started by `access_start.cmd`.
 ## 7. OPC UA server (§10)
 - Process `OPXUAS` (`opxuas -n opxuas`, started by `access_start.cmd`; stop `durstp -m -p
  opxuas`), Unified Automation C++ SDK, port **34493**. Log `%TLS_ALLUSERS_PATH%\log\OPXUAS.log`
  and `OpxUas_Sdk*.log`.
 - Visible data: sections, objects, items with *OPC visible*; items also need **Read** (read
  values, browse into the item) and **Write** (write values, alarm methods). Parent sections must
  be visible. Without Read the item is listed but browsing into it fails
  (`UncertainNotAllNodesAvailable`).
 - Users: anonymous = like `OPC_UA_SERVER_READ`; authorisation actions `OPC_UA_SERVER_READ`
  (browse, read, history) and `OPC_UA_SERVER_MODIFY` (+ write, methods).
 - Address space: `Objects` → sections (folders) → objects → items (`CIItemNumericType`,
  `…BooleanType`, `…StringType`, `…NumericSubType`) with child nodes such as `Value`,
  `Status`, `Blocked`, `Offline`, `HMIInfo`, and alarm condition nodes. Types are in
  `http://www.yokogawa.com/CI_Server/` (ns 3 in the manual). **CI Server entities (sections,
  items) are in namespace `urn:Yokogawa:CI_SERVER_OPC_UA_SERVER`** (verified by browsing;
  the manual does not name it, ISS-032).
 - NodeIds are JSON strings: `{"nsid":"<namespace id>"}` for the entity,
  `{"0":"Value","nsid":"<n>"}` for its value; or by name: `{"0":"Value","itnm":"<item>"}`
  (case sensitive; verified). The `nsid` changes when the item is recreated — prefer `itnm`.
  `OPXUAS_USE_NAME_NODEIDS` makes the server use names in its own IDs.
 - Status mapping: NORMAL → Good; NOT_INIT, UPD_OFF, OFFSCAN, OFFLINE, UNREACH →
  BadNoCommunication; BAD_VALUE → Bad. Types: Boolean, Int32, Double, String. Priority →
  severity: 1 → 1000, 2 → 900 … 15 → 50, 0 → 1 (changed in R1.03).
 - History: items in a history group show HistoryRead; max 5000 values per node per call.
 - Alarm methods: Acknowledge, Confirm (= acknowledge), Disable/Enable (force to normal);
  external alarms Acknowledge/Confirm passed to the source; `CIManualReset` for CENTUM.
 - `opxuas.sup`: General (`FTFW_*` queue, heartbeat, log), OPC UA Server tab
  (`OPXUAS_OPTIONAL_NODES`, `…_MODEL_CHANGE_EVENTS`, `…_USE_NAME_NODEIDS`,
  `…_USE_DYNAMIC_SCANRATE`, `…_SOURCE_TIME_STAMPS`, `…_USE_RAW_STATUS`,
  `…_VARIABLE_CACHE_SIZE` 100000, `…_NUM_VALUES_PER_NODE` 5000), Alarm settings (`EVENTS` =
  ALARM, ACKN, DEACT, INFO), Logging. Installed keyword names: `OPXUAS_DYNAMIC_SCANRATE`
  (manual `…_USE_DYNAMIC_SCANRATE`) and extra `OPXUAS_SET_HIGH_PRIO_15` (ISS-034).
 - `%TLS_ROOT_PATH%\tls\exe\ServerConfig.xml`: endpoint `opc.tcp://[NodeName]:34493`; trust list
  `…\pki\opxuas\pkiserver\trusted\certs`, rejected `…\pkiserver\rejected` (created at the first
  rejection), own certificate `ciserveruaserver.der`. **Installed**: only Basic256Sha256,
  Aes128_Sha256_RsaOaep, Aes256_Sha256_RsaPss with Sign / SignAndEncrypt (no None endpoint);
  anonymous, user/password **and certificate** user tokens enabled (manual: certificate off,
  ISS-034); `AutomaticallyTrustAllClientCertificates` false.
 - First connection of any client: trust the server certificate on the client, then move the
  client certificate from the server's `rejected` to `trusted\certs`. UaExpert is the
  suggested test client.
 - Discovery: LDS on 4840 only when several servers are on a host; GDS for central certificate
  management (concepts only).
--- a/knowledge/manuals/IM36K10L55-01EN.md
+++ b/knowledge/manuals/IM36K10L55-01EN.md
@ -0,0 +1,51 @@
 ---
 manual: IM 36K10L55-01EN
 title: Exchange Tool Engineering
 edition: 3rd Edition, Sep. 2023
 version: R1.03
 status: from docs only (needs CENTUM VP / an RGS node or CCC Prodigy data; not available on the test system)
 ---
 # Exchange Tool Engineering (R1.03)
 Two tools convert another system's engineering data into CI Server engineering data
 (a "publication" = generated quick load files plus `load.cmd`). Section numbers refer to the
 manual; Appendix 1 lists the supported CENTUM VP function blocks.
 ## 1. CI Exchange for CENTUM VP (§2)
 - Runs on an **RGS** (Remote Gateway Station: CI Server PC with a Vnet/IP card and driver from
  the medium, `VirtualProduct_CIServer_packageRxxx.zip`, option *installVirtProd*) that talks
  to CENTUM VP. *Single server*: the RGS is also the CI Core. *Non-single*: an upper CI Core
  gets the data from the RGS (MDUR).
 - Solutions: **RGS** (manual or automatic mode; automatic = the tool is embedded in the RGS,
  ENG HIS equalisation enabled, the RGS registered in the CENTUM VP project; publications
  generated after each CENTUM VP download) and **RGS with EAL** (external alarms; manual mode
  only).
 - Manual workflow: start the tool → create a project (*Single Server* ticked or not) → import
  CENTUM VP engineering data from the ENG HIS (and MTL file) → *Publish* (name pattern, folder,
  optional SQL tag filter such as `%PVI%`) → copy the publication folder to
  `C:\Program Files (x86)\Yokogawa\CIServer\Utility` → run its `load.cmd` on the RGS and/or CI
  Core (in an IT-security environment run `cmd.exe` *as different user* first). Automatic
  publications land in `%TLS_ALLUSERS_PATH%\lst\rgs\pub\`.
 - Output: items `<section>.<TAG>.<PARAM>` (e.g. `<Section>.FIC001.PV`), points
  `<STATION>:[<item>]`, plus lines/stations, displays and symbols (displays only on the CI
  Core in non-single configurations).
 - Precautions (§2.15): refresh open LIST displays after a CENTUM VP online download; CI Server
  does not allow alarm limits on the wrong side of SV; short data gaps after Vnet scan changes
  or RDC switch-over; trends may lag after large downloads — reload displays, roll over the
  history group, tune scan types.
 - Installed: launcher `%TLS_ROOT_PATH%\tls\com\jwuiutl_ciexchange.cmd` (no RGS on the test
  system).
 ## 2. CI Exchange for Prodigy (§3)
 - For CCC Prodigy turbo-machinery controllers (antisurge, performance, speed). Input: the
  Modbus register list (`.csv`, from MakerPro) per controller; the station name and IP are
  entered in the tool.
 - Install: copy `Windows\EngineeringTools\CIExchange_Prodigy.zip` from the medium, extract to
  `C:\Program Files (x86)\Yokogawa\CIServer\Utility` (create the folder), run
  `CIExchange_Prodigy.cmd`.
 - Workflow: create project → import source configuration → generate publication → load it.
  Output: items `<section>.<Prodigy variable path>`, Modbus points `<STATION>:[<item>]`, and
  faceplates (title and engineering unit can be customised, §3.12.1).