Skip to content

Digital board v3r4: 5 V I/O, pin 21, pull strength (SDK + MCP) - #123

Merged
wardviaene merged 4 commits into
mainfrom
digital-v3r4
Nov 19, 2026
Merged

wardviaene merged 4 commits into
mainfrom
digital-v3r4

Conversation

@wardviaene

Copy link
Copy Markdown
Member

Host side of the digital board v3r4 firmware stack: embeddedci-com/benchpod-firmware#175, embeddedci-com/benchpod-firmware#176, embeddedci-com/benchpod-firmware#178. Merge after those (the bundled feature map and error catalog come from #178).

SDK

  • set_la_voltage(5.0) when supports("la_5v"): sends mv: 5000. Every other pod raises ValueError before anything is sent, so an older board sees no new traffic. Other bad values name 5.0 only where the pod has it.
  • LaVoltage.pin21: new Pin21 (source, setting, measured, state, up, ok) from the reply's vref_out; None on other boards.
  • Capabilities.vref_out from the hw facts. Wiring accepts la_mv: 5000.
  • PullState.strength (10k, 2.2k, both, none). la_pins() passes the reply's pull_dirs to each entry (v3r4 entries leave dirs out to fit the cloud reply).
  • No strength or mode parameter (5v-support 12.4: the user keeps on/off, the pod picks the strength from the pin's function). The firmware takes "strength" for a raw command() if ever needed.
  • Feature map synced (revision 7: la_5v 6, la_pull_strength 7) and error catalog (BP0028 5 V did not come up, BP0029 pin 21 is down).

MCP

  • set_la_voltage: 5.0 allowed; voltage optional (omitted = read); returns pin21 and tells the agent what high means. connect(la_voltage=5.0), --la-voltage 5.0.
  • capabilities.vref_out; pull channels have strength; board_rev description lists v3r2/v3r4.
  • Tool and API surface snapshots refreshed (only additions; set_la_voltage.voltage became optional).

Tests: new test_digital_v3r4.py (v3r4 pod and an older pod: capabilities, 5 V, BP0028, refusal before sending, strength, pull_dirs, profile). MCP e2e set_la_voltage check no longer compares the whole dict (the result now has pin21). All unit suites pass; test_bundled_copy_matches_the_firmware passes against the firmware branch (BENCHPOD_FIRMWARE_DIR) and skips in CI.

Not testable without a v3r4 board: everything that needs the real pin 21 and expanders. On today's boards the regression run (hwe2e and SDK e2e on benchpod-30a9f1) is reported on the firmware PRs.

🤖 Generated with Claude Code

wardviaene and others added 4 commits November 19, 2026 11:36
SDK: set_la_voltage(5.0) on a pod with the la_5v feature (the bank stays at
3.3 V, J1 pin 21 goes to 5 V); other pods get ValueError before anything is
sent. Wiring la_mv takes 5000. LaVoltage.pin21 (new Pin21) reports pin 21's
source, setting, measured voltage and state on v3r4. Capabilities.vref_out is
the hardware fact. PullState.strength reports what the pod picked; la_pins()
takes the reply's pull_dirs. No new parameters: the pod picks the strength.

MCP: set_la_voltage takes 5.0 and an omitted voltage (read only) and returns
pin21; connect and --la-voltage take 5.0; capabilities has vref_out; pull
channels have strength.

Feature map (revision 7) and error catalog (BP0028, BP0029) synced from
benchpod-firmware#178.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…entries on v3r4

- la_pins(la=...) lists only those channels. Firmware after 3.10.12 sends only them, older firmware
  sends all 14 and the SDK filters. A full listing answered with "reply too large" is read in two
  parts (LA1-7, LA8-14).
- PullState.held (and MCP pull_status/set_pull "held"): why a pull that was set is off now on
  digital v3r4 (pin21_down, dut_off, fault). enabled/state are the live state (firmware #181).
- la_dir() and gpio entries on v3r4 had pull_dirs ("up",) only: their replies leave dirs out and
  have no pull_dirs. They now get ("up","down") when the pod has la_pull_strength.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A full la_pins listing that older firmware answers with "reply too large" (it has no la subset,
so the SDK cannot read it in parts) now says that firmware after 3.10.12 reads it in parts, how
to update, and that bp.la_pins(la=[...]) asks for fewer channels. Like the status hint.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The SDK README, the set_la_voltage docstring and the MCP tool point to
embeddedci.com/docs/benchpod-targets/5v-rs485#5v-io instead of explaining 5 V I/O again. The
README no longer says the pod has no 5 V I/O (only boards before v3r4). The MCP instructions and
la-voltage-not-set hints name 5.0 on a digital v3r4 pod. The docstring named the lease code
BP0025 for a 5 V that did not come up; it is BP0028 (the test fake had it too).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@wardviaene
wardviaene enabled auto-merge (squash) November 19, 2026 17:38
@wardviaene
wardviaene merged commit 369ebd9 into main Nov 19, 2026
20 checks passed
@wardviaene wardviaene mentioned this pull request Nov 19, 2026
@wardviaene
wardviaene deleted the digital-v3r4 branch November 20, 2026 02:47
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant