tasks/sky130A/OpenFASOC/cases/current_mirror/problem.md

SKY130 1:1 Interdigitated Current Mirror Layout Task

Objective

Implement CMIRROR from the supplied transistor-level circuit, preserving topology, fixed device parameters, four-terminal body connections and the ordered interface. Submit a candidate GDS for electrical evaluation of its distributed RC extraction.

Optimization objective

Maximize the declared quality score within the solve budget while satisfying all physical and necessary functional requirements. The first passing DRC, LVS and required electrical checks establish a feasible baseline. Continue improving the scored metrics from that baseline; a passing check or task_success = true does not mean that the quality objectives have been achieved.

Inputs and Interface

Ordered ports are VREF VCOPY VSS VB. Declared inputs include problem.md, materials/circuit.spice and materials/testbench.spice. The runtime provides models and process resources; reference layouts, author code and qualification evidence are maintainer-only materials.

Operating Conditions

Supply 1.8 V; reference currents 10, 30, 60 uA. Sweep output voltage from 0.6 to 1.8 V in 50 mV steps. Observe maximum relative current error and bias voltage. All electrical experiments use fixed TT models at 27 °C; delivered testbenches define interpolation, edges, loads, sweeps, observation times and power intervals.

Physical Requirements

Submit a nonempty GDS no larger than 10485760 bytes, with top cell CMIRROR. Functional geometry must fit within the 60 × 60 um envelope on the declared layer set. Pass complete Magic DRC without waivers and independent Netgen LVS checking named ports, models, dimensions, multiplicities and body/contact connections. After physical and geometric checks pass, extract the candidate's distributed interconnect resistance, capacitance and devices, then run every declared experiment.

Electrical Requirements and Scoring

Artifact, geometry, DRC, LVS and declared functional requirements must pass. Parasitic extraction is also mandatory: it must produce a complete, electrically consistent netlist usable by the required post-layout simulations. A completed candidate rejection scores zero; infrastructure errors do not establish candidate failure.

For layout scoring, cap every positively weighted raw quality at 1 before aggregation; the score ranges from 0 to 100. Reaching every weighted target scores 100, and no minimum score is required for validity.

Absolute current-ratio error must be nonnegative; bias must be within the declared 0.3–1.2 V operating range. The current-transfer quality target is 3% worst absolute relative error across every declared current and the full output-voltage sweep, with a 0.5 percentage-point normalization scale. It is a challenging continuous quality goal, not a rejection threshold. Power and absolute errors must be nonnegative; all required measurements must be complete, finite and in correct units.

Metric	Functional requirement or measurement domain	Aggregation	Score weight
functional_area	No fixed performance bound	max	50%
ratio_error	0 … +∞ 1	max	50%
bias_voltage	0.3 … 1.2 V	max	0%

Retain independent source-netlist observations under exactly the same conditions. For each candidate worst-sweep error x, quality is 2*(t+s)/(t+x+2*s), with target t=0.03 and scale s=0.005; take the worst quality across the three currents. Reports distinguish the measured source errors from the declared target. This deterministic metric covers compliance and routing effects; it does not measure statistical matching yield. Zero-weight observations still undergo measurement and functional checks.

The area quality target is 90 um2. The eight fixed 3-by-2 um gate rectangles occupy 48 um2, leaving a 42 um2 integration allocation for diffusion, spacing, contacts and routing. This is a quality budget, not a process minimum or an acceptance bound. Full quality requires both this compact footprint and the 3% worst-sweep current-error target; spare area performance cannot compensate for electrical degradation.

Use layout scoring. First compute each metric's raw quality q from the worst paired condition, with raw area quality q_area = area_target / candidate_functional_area. Credit c_i = min(1, q_i) for every positively weighted metric, then compute S = 100 * product(c_i ** w_i). The weights sum to 1 and the score is bounded by 100. An improvement beyond one target cannot compensate for missing another target. Physical, extraction, or functional rejection scores zero; other missing or invalid measurements leave the score unknown, and zero-weight observations must also be valid. Reaching every positively weighted target scores 100; a feasible reference may score far below 100. Weights express quality tradeoffs, not acceptance bounds.

Tools and Submission

Soft solve budget: 1 hour.

Read /protocol/task.json for input/output paths, constraints and the resources object; /protocol/harness.json describes the active check and submission interface.

Use the task information, tools and SKY130 resources provided by the runtime protocol. Write only the final GDS to output/final.gds and explicitly submit through the declared interface. Source simulation serves characterization; accepted results come from complete physical and RC electrical evaluation of the candidate GDS. Conclusions cover only the declared conditions, excluding PVT, mismatch, reliability and product signoff.

Solve workflow

1. Read the declared circuit and materials, /protocol/task.json for paths, constraints and reviewed resources, and /protocol/harness.json for capabilities and helper commands. Task inputs under /task and resources under /resources are read-only; construct the layout in /workspace with the declared tools. With layout MCP, paths and commands refer to the remote EDA workspace. Create your own implementation with real devices, interconnect and pins, using your own tool loop without delegating to other agents; reference layouts and author code are not solver inputs.
2. Inspect resources.external before using external tools. Each command starts a fresh shell: when tool modules are declared, source /usr/share/Modules/init/bash and load the selected tool's modules; export its declared tool_environment settings in that same command. Repeat as needed and use batch or headless operation. Local manuals, command references, and release notes may be available within the declared EDA installation and PDK resource directories. Locate these resources through /protocol/task.json and consult the documentation for the installed tool version when diagnosing unfamiliar syntax or tool failures.
3. Generate an initial candidate and submit a snapshot early. When process-feedback is available, check the current GDS with layout MCP check or python -I /protocol/process_check.py before the final submission. This runs the final evaluator's complete plan and returns engineering diagnostics and candidate measurements, without scores or normalized quality factors. Checks have no count limit and consume the same wall-clock budget. Inspect the physical and functional results and all scored metrics; successful tool execution alone does not establish requirement satisfaction. Read paginated diagnostics with report when using MCP. Local exploratory checks do not replace this complete check.
4. Use the remaining time for measured improvements to the declared quality objectives. Identify the limiting scored metrics, make targeted layout changes, regenerate and check the candidate, and compare measurements with the previous candidate and the stated scoring targets and trade-offs. Preserve the best valid candidate and revert changes that worsen the overall objective. Repeat while time and useful improvements remain; passing validity checks alone is not a stopping condition.
5. Stop optimization when the budget expires, all declared quality objectives are achieved, or measured trials support convergence. If stopping early, explain the changes tried, the metric comparisons and the remaining limitations. Track remaining time through session status. At soft-budget expiry, finish or cancel already-started work and submit immediately; do not start another command, check or optimization round. If an MCP execution is running, use poll with its execution_id and the returned next_offset as offset until completion, or cancel it and wait for it to stop.
6. Finish or cancel outstanding work, restore your best candidate to the declared output path, and explicitly submit it before ending the session. With layout MCP submit, remove the /workspace/ prefix from output.path; with the container interface, run python -I /protocol/submit.py. Wait for the receipt: only the last accepted snapshot is judged, and writing a file alone is not submission. Acceptance confirms delivery; the framework independently evaluates validity and quality after the session.