Where ANVIL is going
The beta sizes two amplifier topologies, checks them across process, temperature and supply corners, and proves every number with a simulation. The items below are what we plan to build on that base.
Broader sizing
More circuits and more analyses, with the same per-number traceability the beta has today.
More topologies
PlannedFolded-cascode and telescopic OTAs, fully differential variants with common-mode feedback, and LDO regulators.
Monte Carlo and mismatch
PlannedMismatch analysis and Monte Carlo runs, with the same per-number traceability as today's compliance table.
Noise analysis
PlannedInput-referred noise as a spec, measured and graded like the others.
Layout parasitics
PlannedCheck a sized design with parasitics extracted from its layout, not only at schematic level.
Side-by-side topology comparison
PlannedSize several topologies against one spec in a single run and compare their worst-case compliance in one table.
Your own models and foundry PDKs
PlannedLoad your own model cards and corners, including a foundry PDK, with simulation on infrastructure you control so the models never leave your network.
Design workflows for teams
Sizing is one step in a longer loop from requirements to a reviewed cell. We plan to extend ANVIL along that loop.
Composable workflows
PlannedChain spec extraction, sizing, corner sweeps and report generation into one workflow. Steps pass typed results (spec, netlist, simulation result, report), and workflows are stored as versioned text files you can review like code.
Approval gates
PlannedPoints where a workflow pauses for a designer to confirm extracted specs, approve a simulation budget or pick a topology. A gate pauses only its own branch, and each decision is recorded with its evidence.
Requirement-change impact
PlannedWhen a requirement changes, find the results that depended on the old value, re-check stored simulation data against the new limit without re-simulating, and re-run only what is no longer valid.
Result lineage
PlannedA graph from each requirement through specs, design variants and simulation evidence to the decisions and reports built on them, showing what any result depends on and what depends on it.
No duplicate simulations
PlannedReuse a stored result when the netlist, models and simulator version match a previous run, instead of simulating it again.
Tell us what to build first
If one of these would change how you work, or something you need is missing, write to hello@anvillab.tech.