Migrating from neuralSPOT

NSX is a modular replacement for the parts of legacy neuralSPOT that matter for current Ambiq bare-metal work: bring-up, build, profiling and device integration. It is not a one-to-one repackaging, and it does not try to be.

The approach is to preserve capabilities rather than package boundaries. Large, opinionated legacy modules get split into smaller NSX modules where that improves reuse, and anything app-specific stays optional rather than joining the baseline.

What NSX covers today

The migrated core is the baseline you need to create an app, configure and build it reproducibly, flash and inspect it on hardware, bring up common board peripherals, and run profiling and instrumentation. That is why the first wave centered on nsx-core, nsx-perf, nsx-pmu-armv8m, nsx-power, nsx-i2c, nsx-spi, nsx-uart, nsx-soc-hal, nsx-cmsis-startup, nsx-ambiq-hal and nsx-ambiq-bsp.

It does not yet carry every legacy application stack.

Refactoring direction

Several legacy modules carried assumptions about examples, transport choices, power behavior or application structure. NSX keeps the useful parts while making the modules less opinionated:

1. Mixed-purpose utility bundles are split into focused modules.
2. Explicit metadata and dependency closure replace implicit coupling.
3. Board and SDK specifics stay in wrapper modules.
4. Optional connectivity, sensor and ML stacks stay out of the baseline unless they are broadly useful.
5. Small, stable APIs are preferred, so the same surface serves smoke tests, examples and product code.

You can see the result in the current module set: legacy PMU and perf helpers came out of ns-utils into nsx-pmu-armv8m and nsx-perf; power gained a first-class home in nsx-power; mixed peripheral helpers were retired into focused surfaces; and SDK consumption is split across provider, HAL, BSP, startup and SoC modules instead of being buried in app code.

Migration matrix

Status meanings:

• Migrated: a first-class NSX replacement exists.
• Split: the capability exists but is spread across several NSX modules.
• Partial: some of the capability exists; the replacement is incomplete.
• Future: no NSX replacement yet.
• Dropped: a decision was made not to migrate it.

Legacy module	Status	NSX home	Notes
ns-core	Migrated	nsx-core	Core runtime and bring-up surface.
ns-harness	Split	nsx-core, nsx-perf, nsx-pmu-armv8m	Print, delay and interrupt shims moved into nsx-core; profiling into nsx-perf and nsx-pmu-armv8m. TFLM DebugLog registration is superseded by heliaRT’s own implementation. The remaining gap is packaging only: the MicroProfilerInterface glue (NsxPmuProfiler) still lives in the kws_infer example rather than in a reusable module.
ns-i2c	Migrated	nsx-i2c	Bus wrapper and register-driver helpers. The sample device drivers (MPU6050, MAX86150) moved to nsx-sensors.
ns-spi	Migrated	nsx-spi	SPI wrapper support.
ns-uart	Migrated	nsx-uart	UART wrapper support.
ns-features	Future	none	The quaternion and Euler feature-extraction helpers have not been ported.
ns-peripherals	Migrated	nsx-power, nsx-psram, board button definitions in nsx-gpio	The mixed legacy bucket was retired into focused surfaces rather than kept under one name.
ns-utils	Split	nsx-core, nsx-perf, nsx-pmu-armv8m, nsx-power	The portable helpers (nsx_printf, nsx_delay_us, interrupt enable and disable) live directly in nsx-core.
ns-ble	Migrated	nsx-ble with nsx-cordio	An early baseline: one service and one connection, no out-of-band pairing. The ble_webble example exercises it across the CI build matrix.
ns-usb	Migrated	nsx-usb	USB CDC serial on TinyUSB.
ns-imu	Partial	nsx-sensors	The ICM-45605 driver was ported at TDK basic-driver scope. The rest of the generic ns-imu wrapper was deliberately not carried forward.
ns-audio	Migrated	nsx-audio	PDM capture with DMA-backed sampling and callback delivery.
ns-physiokit	Migrated	nsx-physiokit	Biosignal primitives for ECG, PPG, respiration and HRV, built on helia-dsp.
ns-tileio	Migrated	nsx-tileio-ble, nsx-tileio-usb	Split by transport, with Python host tools.
ns-camera	Future	none	Narrow, app-specific demo hardware (Arducam Mega over SPI, Apollo4 only). A direct port is probably not worth it.
ns-ipc	Dropped	none	Not migrated.
ns-model	Dropped	none	Not migrated.
ns-nnsp	Dropped	none	Not migrated.
ns-rpc	Dropped	none	Not migrated. nsx-nanopb is available as a protobuf framing building block if a transport wrapper is revisited.

Two names in that table have no catalog page

nsx-gpio and nsx-psram are vendored inside the nsx-ambiq-sdk monorepo rather than registered as standalone entries, so they do not appear in the module catalog. They are still real modules you can depend on.

Deciding whether to migrate something

Before porting a legacy module, ask:

1. Does it help lightweight bare-metal NSX apps?
2. Is it generic enough to serve more than one app or demo?
3. Can it be expressed with a smaller, less opinionated API surface?
4. Is it better as an optional module than as a baseline dependency?
5. Does it fit the NSX module model cleanly?

A module that answers yes to most of these is a good candidate.

Where NSX already replaces the legacy core

NSX already covers board bring-up, smoke tests, the build, flash and view workflow, common peripheral access, and profiling and PMU instrumentation. Chasing broad legacy parity for its own sake is not the goal; simplifying the old module boundaries as they move is.