Talk

Smart Speaker Design Prompt (HomePod mini class)

From Wikiprompt, the free prompt encyclopedia

Avid
Contributed byAvidXSource

Nov 11, 2026

Smart Speaker Design Prompt (HomePod mini class) A comprehensive prompt for designing a smart speaker from scratch, covering hardware, software, and mechanical aspects, with detailed specifications and deliverables.

Prompt ContentSave

🌐
# Task Design a smart speaker of the Apple HomePod mini class. One enclosure contains the computer, the microphones, the amplifier, and the speaker driver. The device runs Home Assistant OS (HAOS) and uses Claude as the voice assistant. Deliver a complete hardware, mechanical, and software package that is ready to manufacture. Use an original industrial design. Do not copy Apple's shape, branding, or trade dress. # Voice path 1. The microphone array captures audio continuously. Wake word detection runs on the device. 2. After the wake word, the device sends the speech to a speech-to-text engine. 3. Home Assistant sends the text to Claude through the Anthropic API. Claude can control Home Assistant entities. 4. A text-to-speech engine converts the reply to audio. The device plays the audio through the internal speaker driver. State where each stage runs (on the device or in the cloud) and give the reason. Confirm the method against the current Home Assistant documentation. # Hardware requirements Compute - Select a publicly available SoC or system-on-module (SoM): 64-bit, minimum 4 cores, minimum 4 GB RAM. - Prefer a part that HAOS already supports. Give the reason for the selection and list two rejected alternatives. - Storage: microSD slot. eMMC is optional. Audio output - One full-range speaker driver, 50 mm to 75 mm, in a sealed chamber or with passive radiators. - Class-D amplifier with I2S input, minimum 15 W into the speaker driver. - Target: 85 dB SPL at 1 m, 80 Hz to 18 kHz at -6 dB. Show the chamber calculation from the Thiele-Small parameters of the speaker driver. Audio input - Minimum 4 MEMS microphones in a circular array on the top of the device. - Acoustic echo cancellation with a reference signal from the amplifier output. The device must detect the wake word during music playback. - A hardware mute switch that removes power from the microphones and turns on a red LED. Software must not be able to override the switch. User interface - Capacitive touch zones on the top: volume up, volume down, play/pause, start assistant. - RGB LED ring with these states: idle, listening, thinking, speaking, muted, error. Radios - Wi-Fi on 2.4 GHz and 5 GHz, and Bluetooth 5.0 or later. Use a module that has an FCC ID. - One 802.15.4 radio on the board for Zigbee and Thread. - Keep the antennas away from the speaker driver magnet and from metal parts. Show the antenna keep-out areas. Power and ports - USB-C Power Delivery (PD) sink. Calculate the power budget first, then select the PD contract. - Define the behavior with a weak power supply: limit the amplifier power and show a warning on the LED ring. - One USB-A 2.0 host port. One UART debug header. - No fan. Calculate the SoC temperature at 35 °C ambient in the closed enclosure. # PCB - Use KiCad. Maximum 6 layers. State the stackup and the design rules of the selected fabricator. - Use a main board and a top board (microphones, LEDs, touch) if this helps the mechanical layout. - Isolate the microphones from enclosure vibration. # Enclosure - Target size: maximum 110 mm diameter and 130 mm height. - All parts must print on an FDM printer with a 0.4 mm nozzle and minimum supports. - Seal the acoustic chamber with a gasket. - Use heat-set threaded inserts, screws from the bottom, and rubber feet. - Deliver the parametric source (CadQuery, OpenSCAD, or FreeCAD), STEP, and STL or 3MF for each part. - Put the PCB STEP model in the enclosure model. Report the clearances and each interference. # Software - HAOS publishes .img.xz disk images, not ISO files, and supports a fixed list of boards. If the selected hardware is not on the list, add a board configuration to the HAOS build tree: device tree, kernel configuration, bootloader, and kernel modules for the audio hardware. - Deliver a build script that gets the latest stable HAOS release tag, applies the board support, and makes an .img.xz file for the microSD card. - First boot must give: Wi-Fi onboarding, configured audio devices, a configured voice pipeline, and active LED and touch functions. - The user enters the Anthropic API key during onboarding. Do not put keys or passwords in the image. - Describe the HAOS update method for this board. - Play music through Home Assistant. Decrease the music volume when the device detects the wake word. # Sourcing - Each BOM line must show: manufacturer part number, distributor part number, stock quantity, lifecycle status, and unit price at 1, 100, and 1,000 units. - Each part must be in stock at Digi-Key, Mouser, Arrow, or Newark in the US on the date of the check. Record the date. - Do not use parts with NRND or obsolete status. Give one alternate for each part that is not a generic passive. - Include the speaker driver, passive radiators, gaskets, screws, inserts, and cables in the BOM. # Deliverables /docs requirements, architecture, decisions.md, power budget, thermal calculation, acoustic calculation, verification-report.md /hardware KiCad project, schematic PDF, Gerber and drill files, pick-and-place file, BOM CSV, board STEP /mechanical CAD source, STEP, STL or 3MF, assembly drawing /software HAOS board support, build script, first-boot configuration README assembly instructions and flash instructions # Definition of done - ERC shows 0 errors. DRC shows 0 errors with the fabricator rules. Attach the kicad-cli reports. - Each footprint and pinout agrees with the datasheet. List the datasheet and the page. - 100% of the BOM lines obey the sourcing rules. - The enclosure check shows no interference. - The image build completes, or the report states the exact step that failed. - verification-report.md lists each requirement in this prompt with one status: verified by tool, verified by calculation, or not verified. Give the evidence for each status. # Working rules - Work in this sequence: requirements, architecture and part selection, schematic, layout, enclosure, software, verification. - Do not stop to ask questions. If a requirement is ambiguous, select the option with the lowest risk and record it in decisions.md. - Read the datasheet before you use a part. Do not write pinouts, part numbers, or stock data from memory. - Do not report a check as passed unless you ran the check. - List the items that only a physical prototype can confirm: RF performance, acoustic response, echo cancellation, temperature, and FCC Part 15 tests of the complete device. - If you cannot meet a requirement, state this and propose the nearest alternative. Do not remove a requirement without a record

Sign in to see the full prompt

Continue with:

By logging in, you agree to our Terms of Use and Privacy Policy

Usage

This prompt is designed for use with other. Copy the prompt content above and paste it into your preferred AI tool.

For best results, you may customize the placeholders (indicated by square brackets or capital letters) with your specific requirements.

References

Categories:other| twitter| smart-speaker| hardware-design

Talk