WiFi Voice AI Hub Prototype Development

flux

I tried the non-repeated recovery paths: deleted/rebuilt the entire 5V copper, then deleted all routed copper for all nets and rerouted the board from scratch. The issue persists: router says nothing remains to route, but DRC still reports 1 5V airwire; D356 behavior therefore remains a Flux export/checker blocker, not something I can honestly certify as manufacturing-safe.
Open items:
  • I cannot provide written manufacturing-safe confirmation.
  • Flux support needs to fix/confirm the IPC-D-356 N/C export and remaining false/stale 5V airwire.
  • I documented the blocker here: manufacturing-export-blocker-d356-5v-airwire.
Suggested next steps:
  1. Send Flux support this project link and the blocker doc.
  2. Ask for written confirmation/fix of IPC-D-356 net serialization.
  3. Do not fabricate until DRC airwires are zero or Flux provides explicit false-positive confirmation.
A

asdfghjikolp

Can you tell me what happens if you make a small schematic change to the 5V net and re-sync. does the airwire or N/C behavior change at all?

flux

I’ll make a minimal reversible source-net change on 5V, let the layout/export state refresh, then check whether the 5V airwire and IPC-D-356 behavior change.

flux

I made a small reversible schematic change: temporarily disconnected C7:P1 from 5V, then reconnected it to ^5V.
Result:
  • 5V schematic net restored correctly to 20 pins.
  • Airwire behavior did not change: DRC still reports 1 airwire on 5V (Trace a5b4).
  • Router still says nothing remains to route, so the contradiction persists.
  • I generated a fresh D356 export: wifi-voice-ai-hub.d356, but my sandbox could not fetch/parse the CDN file due DNS failure, so I can’t truthfully confirm whether the N/C contents changed.
Conclusion: the schematic re-sync did not clear the Flux-side 5V airwire issue. If your fresh D356 still marks real pads as N/C, that strongly confirms this is an export/checker platform bug, not stale schematic connectivity.
Escalate and verify
Try fresh clone
Audit schematic portals
Generate evidence package
Check manufacturing alternatives
Sign up to send

Files

Pins

Symbol

Layout

3D View

Beginner Plan — Turning WiFi Voice AI Hub Into a Working Prototype
Plain-English Goal
You already have a PCB design for a USB-C powered ESP32-S3 voice assistant hub. The next job is to manufacture one or more boards, assemble them, load firmware, test each subsystem, and iterate if anything fails.
What You Need To Buy / Prepare
  • Assembled PCB from a PCB assembly vendor using this Flux design.
  • USB-C 5 V wall adapter rated at least 2 A.
  • USB-C cable that supports data, not only charging.
  • External 4–8 Ω speaker, ideally 3 W or higher.
  • Computer for programming the ESP32-S3.
  • Multimeter.
  • Optional but helpful: USB power meter, current-limited bench supply, oscilloscope, hot-air/rework help.
Step 1 — Final Design Review Before Ordering
  1. Open the PCB view and visually inspect:
    • USB-C connector at the board edge.
    • ESP32 antenna at/near the edge with no copper or parts blocking the antenna area.
    • Speaker connector accessible at the edge.
    • Buttons accessible.
  2. Confirm the PCB fab can support the fine routing used in this revision:
    • Minimum trace width used: about 0.08–0.10 mm.
    • If your vendor cannot support this, revise the board for larger footprints or looser routing rules before ordering.
  3. Export manufacturing files:
    • Gerbers.
    • Drill files.
    • BOM.
    • Pick-and-place / centroid file.
  4. Upload those to the assembly vendor and resolve any part-stock substitutions before paying.
Step 2 — Order the Prototype
Recommended first order:
  • Quantity: 2–5 assembled boards.
  • Assembly side: top side.
  • PCB: 4-layer.
  • Solder mask: any color; green is safest/cheapest.
  • Surface finish: ENIG preferred, HASL acceptable if vendor supports the footprints.
Do not order hundreds yet. First revision boards often need fixes.
Step 3 — Inspect Boards When They Arrive
Before plugging in USB:
  1. Check for obvious solder bridges, missing parts, rotated parts, or damaged connectors.
  2. Verify the ESP32 module, USB-C connector, microphone, regulator, and amplifier are installed correctly.
  3. Use a multimeter in resistance mode:
    • Check 5V to GND is not shorted.
    • Check 3V3 to GND is not shorted.
If either rail looks nearly shorted, do not power it. Inspect/rework first.
Step 4 — First Power Test
  1. Plug into a current-limited 5 V source if available, or use a USB power meter.
  2. Check that nothing gets hot.
  3. Measure:
    • USB/input rail near 5 V.
    • 3V3 rail near 3.3 V.
  4. If current is very high or voltage collapses, unplug immediately.
Step 5 — Programming / USB Test
  1. Connect the board to a computer using a data-capable USB-C cable.
  2. Confirm the ESP32-S3 appears as a USB serial/programming device.
  3. Test BOOT and RESET buttons:
    • Hold BOOT while resetting if needed for flashing.
  4. Flash a minimal ESP-IDF “hello world” or USB serial test.
Step 6 — Firmware Build-Up Order
Build firmware in small steps, not all at once:
  1. USB serial logging works.
  2. Wi-Fi connects to your network.
  3. Microphone captures audio samples.
  4. Speaker plays a simple sine wave or WAV sample at low volume.
  5. Rolling audio buffer works in PSRAM.
  6. Cloud upload works for a short test recording.
  7. Cloud response/TTS playback works.
  8. Wake-trigger/cloud pipeline works end-to-end.
Step 7 — Audio Loudness Test
  1. Start with low digital volume.
  2. Use a 4 Ω or 8 Ω speaker rated for at least the amplifier power.
  3. Verify sound is clean and not distorted.
  4. Check USB current while playing loud audio.
  5. If the sound is not loud enough, options are:
    • Use a more efficient speaker/enclosure.
    • Use a larger 5 V supply if current-limited.
    • Move to a higher-power amplifier in the next PCB revision.
Step 8 — Enclosure / Acoustic Work
The speaker and enclosure matter as much as the amplifier.
  • A bare tiny speaker on a desk sounds weak.
  • A proper speaker enclosure can make it much louder.
  • Keep the microphone away from the speaker path to reduce feedback/echo.
  • Add acoustic openings for mic and speaker.
Step 9 — Real-World Testing
Test for:
  • Wi-Fi range.
  • USB power stability.
  • Speaker volume and distortion.
  • Microphone pickup quality.
  • Echo/feedback during playback.
  • Heat after 30–60 minutes.
  • Cloud latency.
  • Recovery after unplug/replug.
Step 10 — Revision Decision
After testing, make a Rev B list:
  • Any soldering/manufacturing issues.
  • Any parts unavailable or expensive.
  • Louder amplifier needed?
  • Better regulator needed?
  • Better mic placement?
  • Need enclosure mounting holes?
  • Need status LED?
  • Need mute button?
Only order a larger batch after Rev A works reliably.
Expected Loudness
This design uses a MAX98357A mono class-D amplifier powered from 5 V. In ideal conditions it is roughly in the small smart-speaker / loud desktop gadget class, not the large Bluetooth speaker class.
Approximate output depends heavily on the speaker:
  • With an 8 Ω speaker: roughly around 1–2 W usable output.
  • With a 4 Ω speaker: roughly around 2–3 W usable output if the USB supply can provide enough current.
  • With a typical 85 dB/W/m speaker: about 88–90 dB at 1 meter.
  • With a more efficient 90 dB/W/m speaker: about 93–95 dB at 1 meter.
Practical takeaway: it should be clearly audible in a room for voice responses, but “room-filling” depends on the speaker/enclosure. For louder music-like playback or a large room, plan a future revision with a higher-power amplifier and better speaker enclosure.

  • Plain-English Goal

  • What You Need To Buy / Prepare

  • Step 1 — Final Design Review Before Ordering

  • Step 2 — Order the Prototype

  • Step 3 — Inspect Boards When They Arrive

  • Step 4 — First Power Test

  • Step 5 — Programming / USB Test

  • Step 6 — Firmware Build-Up Order

  • Step 7 — Audio Loudness Test

  • Step 8 — Enclosure / Acoustic Work

  • Step 9 — Real-World Testing

  • Step 10 — Revision Decision

  • Expected Loudness