# Record 02 — the first centerfed block print did not mate

**Status:** rescued into the repository 2026-09-11 from an untracked working note.

---

## What was attempted

A print of the **3-pump centerfed block** (`block_3pumps_91px_22um_0/`),
intended to mate with the SolidWorks reference manifold — the external fixture
that supplies fluid and pneumatics to the chip.

## What happened — OBSERVED

**It did not mate.** Verbatim from the source:

> Initial print of the 3-pump centerfed block **did not align with the SW
> reference manifold** — interface mismatch.

This is a fit failure between a printed part and physical mating hardware. The
part printed; it would not connect.

## Why — INFERRED

> **Root cause:** +Y face shared fluid manifolds did not match the SW manifold's
> port layout.

The chip presented shared fluid manifolds on its +Y face. The reference manifold
expected a different port layout. Recorded as a root cause rather than a
hypothesis in the source, but it is a diagnosis after the fact, not an
instrumented measurement, so it is marked INFERRED here.

Worth preserving alongside it, because it is the part most easily lost:

> Three wrong diagnostic attempts were made, caught, and reverted before the
> correct fix.

## What was concluded and done — OBSERVED (the design change is in the files)

> **Fix:** replaced the +Y face shared manifolds with **6 independent +Z face
> cylindrical cups** (D=2.5 mm × depth 3 mm). One inlet + one outlet per pump →
> each pump is fluidically independent and can run a different fluid.

That geometry is now the project's fluid interface and **is** in a tracked file:

```json
"cup_diameter_mm": 2.5,
"cup_depth_mm": 3.0,
"bulk_top_z_mm": 11.5,
"cup_floor_z_mm": 8.5,
"substrate_cup_mode": "minimal_active_cups"
```
`openmfda_flow/flow/platforms/h.r.3.3/physical_interface.json` — `fluid_interface`

Six cups on a 2-column × 3-row grid: x ∈ {2.0, 17.0} mm, y ∈ {4.0, 9.0, 14.0} mm.
Inlets on the left column, outlets on the right.

So the *fix* survived into the repository. **The failure that caused it did
not** — until this document. Anyone reading `physical_interface.json` today sees
six independent +Z cups and no reason why, and would have no way to know that
shared +Y manifolds were tried first and did not fit.

Two further consequences recorded in the same source, both of which did survive
into code:

> - Internal channel Z is **derived from cell exports (1.21 mm)**, not hardcoded.
> - **Edge clearance validator** added: HARD floor 0.50 mm, PREF floor 0.75 mm.

## Note on the pitch correction

The cell name `block_3pumps_91px_22um_0` carries `22um`, and the pitch has since
been corrected — the physical machine is a Phrozen Sonic Mighty 12K at
19 × 24 µm. See [`index.md`](index.md).

**This record is unaffected in substance.** The cups are specified in
millimetres and the mismatch was a port-*layout* failure, not a scale failure.
The cell name retains `22um` as a historical label.

## Dates

| what | when | how known |
|---|---|---|
| the print and the mismatch | **before 2026-05-11** | the source lists it as already resolved in a state summary dated 2026-05-11 |
| source file last written | 2026-05-11 | file mtime |

## Source

Recovered from an untracked working note in the lab's working copy of the
design software, `openmfda_flow` — section "B. Centerfed block chip
(`block_3pumps_91px_22um_0/`)".

That note was a summary of the project's state, not a lab record. It is the
least durable place any of these findings was held: a routine edit to it would
have deleted a physical print result with no trace.