Decision

The Accordion A1, currently acting as the programmable PSU feeding the B12 CRPS Test Adapter and B6_Production_test_board, shall be removed and replaced by the Accordion A2 already inside the test fixture. This is the primary driver of the R0.0 → R1.0 hardware BOM revision.

Rationale

• Accordion A2 is already physically present inside the fixture, making it the natural PSU replacement.
• Removing A1 reduces cable complexity and consolidates PSU management onto a single platform.
• A2 requires a BOM upgrade to include the PSU module (currently absent) before it can fulfill this role.

Alternatives considered

• Keep A1 alongside A2: Rejected — adds unnecessary complexity and defeats the consolidation goal.
• Use external bench PSU: Rejected — not compatible with automated test sequences and Synplex/OperatorUI workflow.

Consequences

• ~22 TestStand/Python test cases that reference A1 PSU control must be re-written for the A2 API (see REQ-003).
• Accordion A2 firmware upgrade required in conjunction with hardware BOM change.
• Full regression test required before Synplex deployment (see REQ-004).

Decision

The B12 CB test station upgrade uses two separate repositories, one per DUT.

The CB0 repo is set as the project's primary repositoryUrl. CB1 is tracked here.

Rationale

DUTs are distinct boards with separate test suites. Keeping repos separate maintains clean separation of test scope per board.

Consequences

• Test cases in PLEASE tagged [CB0] or [CB1] to identify which repo/DUT they belong to.
• A1→A2 PSU porting work (REQ-003) spans both repos — ~22 tests total across both.

Finding

After reading the Python source, all A1 PSU control is in a single shared file that is copied (Imported/) into every test folder:

H200_12F_CB_common_testStand_preparations.py

The porting work (REQ-003) is primarily a change to this one file, not 22 separate test scripts.

A1 PSU API calls (current)

All go through esharp.accordion.Accordion class:

from esharp.accordion import Accordion

accdn_handle = Accordion(agent)
accdn_handle.open()
accdn_handle.reset(True)
accdn_handle.loadModule("CRPS")

# PSU channels (numeric index):
accdn_handle.setPsu(0, True/False, 12.0, 3)    # Standby voltage — 12V/3A
accdn_handle.setPsu(1, True/False, 12.0, 2)    # Testboard voltage — 12V/2A
accdn_handle.setPsu(2, True/False, 12.0, 75)   # CRPS voltage — 12V/75A

A2 PSU API (target)

The A2 path is implemented via the dual-mode accordion_a2.py facade (set_rail("standby"/"testboard"/"crps", enable)), which writes M2Top PMBus resources on A2 and falls back to setPsu(ch,…) on A1. standby/testboard channel strings confirmed on live hw — see E-010 / E-012.

Migration scope (Python — DONE & committed)

1. common_TestStand_load_CB_accordion() — replaced Accordion(agent) with AccordionA2 attach pattern ✅
2. common_standby_voltage_enable/disable() — remapped to A2 set_rail("standby") ✅
3. common_testboard_voltage_enable/disable() — remapped to A2 set_rail("testboard") ✅
4. common_CRPS_voltage_enable/disable() — kept deferred on A2 (E-010, CRPS bring-up WIP) ⏳
5. Updated file re-copied into all Imported/ dirs; committed/pushed (CB0 225dc2e, CB1 331946a) ✅

Native TestStand (.seq) — MANUAL, gated, MS-owned (NOT automatable)

PwrRailsSetup (CB0/CB1) contains additional A1 PSU calls inside the TestStand sequences. The .seq files are binary — they cannot be read, parsed, or edited programmatically (no Python/text to diff). This rework must be done manually by MS within the project's gate process; it is outside the scope of any automated/AI edit. Tracking only — no code action available here.

Decision

The logical rail mapping in accordion_a2.py (_RAILS) is confirmed and kept as-is:

Rationale

A1 side (from code, git HEAD pre-migration):

• setPsu(1) → B6_PROD_TB is explicit: the test_board_powerEnable flag in every test top-file carries the comment "For B12, BOA_PROD_TB must be powered on…" and feeds en_CBTB → common_testboard_voltage_enable() → setPsu(1, True, 12.0, 2).
• setPsu(0) → TA/AWO: docstring "Enables standby voltage to the CB", enabled by default on every attach (always-on semantics); the AWO enables live on the TA GPIO expander (0.15.ESH10000462.CB0_PSM_AWO_EN, TA_PMBUS_AWO_EN) and the Bsp tests check PSM_AWO_PG after standby comes up.

Physical mapping (confirmed by Mikael Svensson, 2026-06-10): A1 ch0 (standby) is the left-hand channel output on the A1. On the new M2Top PSU in the A2, the corresponding channel is CH1 — located on the right-hand side looking at the two channel outputs. The sides are mirrored between A1 and M2Top: identify channels by name, never by position.

Alternatives considered

• Swap to CH1=testboard / CH2=standby: rejected — physical inspection of the channel outputs confirms CH1 is the standby feed despite its mirrored position.

Consequences

• accordion_a2.py _RAILS table stays unchanged; the "UNCONFIRMED" comment is replaced with this confirmation (propagated to all 24 Imported/ copies in CB0+CB1 repos).
• The standby and testboard rails are now fully portable/ready on A2. Remaining PSU blocker is CRPS only (E-010 / E-002).
• CM upgrade instructions (REQ-007) should warn about the mirrored channel positions when cabling.

Decision

Black plastic standoff sleeve prototype designed by Felix G was reviewed and accepted by MS on 2026-06-17. Felix's team is producing 5 pcs for the R1.0 upgrade fleet.

What the sleeve does

Lifts the main fixture plate (black lid frame) slightly so it sits level on the grey base enclosure. Sleeves are mounted on the aluminum extrusion rails at the sides of the lid frame, secured with threaded metal inserts and bolts.

Prototype photos (2026-06-17)

Six photos captured at acceptance:

1. Overview — lid raised (front/side): Full enclosure showing the gap between black top frame and grey base where sleeves locate. Red E-stop button visible on front panel.
2. Close-up — two sleeve blocks on corner rail (lid open): Shows the black plastic blocks with brass/threaded inserts seated on the extrusion rail. Two blocks visible at the corner, one above the other.
3. Inside view — sleeve block on rail with Perspex label: Clear acrylic/Perspex material visible (ISO 9001 certification label on material sheet). Accordion (red) and PCB/cable harness visible in background.
4. Side rail — single sleeve block with bolt: One sleeve block on the aluminum extrusion from the side; bolt head visible on top. Station interior with Accordion and braided cable harness.
5. Corner detail — sleeve integrated, closed position: Sleeve block visible as raised block between black lid frame and grey base; Lasbricka M4 (EPN1000184) fastener bins in foreground on bench.
6. Full station assembled — lid level: HARMONIX SPACE front panel visible. Fixture lid sits level on the enclosure with sleeve solution in place. Station in final assembled state on workbench.

Material

Acrylic/Perspex (black). Threaded metal inserts for bolted attachment to extrusion rail.

Production status

5 pcs ordered; Felix G's team responsible for delivery. No ESH number assigned to sleeve part yet — to be determined if formal BOM entry needed for CM upgrade instructions (REQ-007).

Decision

The formal R1.0 HW BOM review (MS-001) is deferred to 2026-08-31 and will be done together with Martin J.

Known BOM delta items (R0.0 → R1.0)

Why deferred

BOM formalization not on critical path for the 2026-06-25 software/deployment deadline. Martin J is the right co-author for the hardware BOM entries.

Situation

The R1.0 hardware revision includes a fresh firmware flash for the Accordion A2. This was confirmed by MS 2026-06-08.

The existing AccordionTsQ2/accordion.py wrapper (already present in every Imported/ folder in both CB0 and CB1 repos) was generated against an older A2 firmware. It cannot be assumed compatible with the new firmware.

Consequence

The A1→A2 PSU porting work (REQ-003, E-008) is blocked until:

1. New A2 hardware with fresh firmware is available in the test station
2. The new firmware API is documented or a new Python wrapper is generated
3. PSU channel string names are confirmed for the new firmware
4. 75A CRPS channel capability is verified on the new PSU module

Risk to schedule

Deadline is 2026-06-25 — 17 days from today. If the new A2 hardware+firmware does not arrive soon, there is insufficient time to port, regression test, and deploy before the deadline.

What is needed immediately

• New A2 firmware API spec or autogenerated Python wrapper (accordion.py)
• Confirmed PSU channel name mapping: standby (12V/3A), testboard (12V/2A), CRPS (12V/75A)
• Confirmation that A2 PSU module can source 75A for CRPS load
• ETA for new A2 hardware delivery to test station

Current state of code

All six PSU control functions in H200_12F_CB_common_testStand_preparations.py use esharp.accordion.Accordion (A1 class, numeric indices). The AccordionTsQ2/accordion.py file exists in every Imported/ folder but is the pre-upgrade A2 API — both are potentially wrong for R1.0.

Situation

The CRPS 12V/75A supply is an external, USB-controlled server-style PSU — never internal to A1. A1 loaded it via loadModule("CRPS"); the handler talks to it over a USB serial bridge via a Microchip MCP2221A USB HID bridge. In the R1.0 upgrade it is re-cabled to the A2.

Current state (2026-06-10) — NOT working yet

Per MS: the CRPS is still being brought up on A2. What appears on the live unit right now is an in-progress attempt, not a verified/working path — do not treat CRPS as done.

• GET /api/Modules/loaded shows a CRPS module (CrpsHandler.dll, /dev/ttyUSB0, enabled:true). This is the team's WIP bring-up of the external PSU, not a confirmed-functional control path.

Fix 1 — ARM64 handler port (Daniel Rhodin, WIP)

Root cause / owner

The blocker is an architecture port: the CRPS handler was 32-bit on the A1 platform; the A2 (AGENT Q2 Base) runs 64-bit ARM, so the handler / its native dependency needed fixing for ARM64. Daniel Rhodin is working the CRPS fixes — pending.

What is still needed

• Daniel: finish the 32→64-bit ARM port of the CRPS handler so the external PSU is controllable on A2 — in progress
• Identify the actual control surface for the bulk supply's enable / voltage / current — separate from the TA sideband signals
• Confirm 12 V / 75 A capability of the unit
• Port common_CRPS_voltage_enable/disable() and the CRPS path in common_TestStand_load_CB_accordion() once the above are known

Fix 2 — udev rule for MCP2221A HID access (bake into A2 OS image)

Background

The CRPS external USB PSU connects to the A2 via the MCP2221A USB HID bridge. On Linux, /dev/hidraw* nodes default to root:root 0600, which blocks the accordion firmware service (non-root) from opening the device.

The fix (documented as E-001 on ESH10000158 R6) is a udev rule:

/etc/udev/rules.d/99-mcp2221.rules

SUBSYSTEM=="hidraw", ATTRS{idVendor}=="04d8", ATTRS{idProduct}=="00dd", MODE="0666"

This was applied manually on one A2 unit to unblock integration testing. For all future A2 deployments this rule must be present out of the box.

Request to firmware lead (Daniel Rhodin)

Include /etc/udev/rules.d/99-mcp2221.rules (content above) in the A2 base OS image / provisioning so every shipped or re-imaged A2 has CRPS HID access without manual intervention.

Alternatively, group-based approach:

SUBSYSTEM=="hidraw", ATTRS{idVendor}=="04d8", ATTRS{idProduct}=="00dd", GROUP="accordion", MODE="0660"

(requires the accordion service user to be a member of the accordion group)

What is needed

• Daniel: include udev rule in A2 OS image / provisioning

Correction note

The resources 0.15.ESH10000462.CRPS_12V_ENn / CRPS_12V_PG / CRPS_1_PRESENTn / CRPS_ISHARE / CRPS_CR_BUSn belong to the B12 CB TA module (ESH10000462, slot 0.15) — these are the DUT-side CRPS interface / sideband signals exercised by SidebandsTest continuity checks (CrpsTAGpioHandler, expander 0x20). They are NOT control of the external bulk CRPS PSU.

⚠️ An earlier revision of this notice incorrectly concluded "CRPS enable = these TA GPIOs / CRPS substantially resolved." That was wrong (conflated the TA's DUT-interface signals with the bulk-supply control). Reverted.

M2Top standby/testboard — confirmed against live hw (unchanged, unblocked)

• 0.16.ESH10000533.0x4F.CH1.* — standby (12V/3A) → B12_CRPS_TA / CB AWO domain
• 0.16.ESH10000533.0x4F.CH2.* — testboard (12V/2A) → B6_PROD_TB

All three resources the shim writes per channel (.OUTPUT_VOLTAGE, .IOUT_OC_FAULT_LIMIT, .OUTPUT_ENABLE) exist on the live unit. CH1/CH2 assignment confirmed 2026-06-10 — see decision E-012 (positions mirrored vs A1; identify by name, not position).

Impact on porting (REQ-003)

• common_standby_voltage_enable/disable() and common_testboard_voltage_enable/disable() — ready, validated against live hw.
• common_CRPS_voltage_enable/disable() + the load_crps()/CRPS path — still blocked on Daniel's ARM64 handler port + udev rule. Shim correctly keeps these deferred (NotImplementedError / skip on A2).

Merged into E-010.

Status (2026-06-10)

R1.0 BOM has been created. SharpMES is updated to R1. The BOM is otherwise complete with one item outstanding:

Pending: ESH number for the Accordion A2 top cover — a variant specific to this tailored A2 (M2PSU + B12 TA configuration). This part does not yet have an ESH number in the system.

Owner: Felix G — waiting on ESH number assignment.

Action

• Felix G: provide ESH number for the A2 top cover (M2PSU + B12 TA variant) — ESH10000691
• MS: add the top cover to the R1.0 BOM — resolved 2026-06-22

Impact on MS-001

MS-001 (R1.0 HW BOM approved) — top cover now has ESH number; BOM can be finalised.