**8-port WLED distro board — looking for a few people to put it through a hard workout**

[Master Of Packets]

United States Testers only, please ( don't think I could afford the shipping . Unless you would like to pay for shipping.

Long-time lurker, first time posting about something I built.

**It started with boards that kept dying.**

We all know how this goes. Something gets wired wrong at eleven at night, a run gets pinched, a supply gets swapped in a hurry — and a board that had no business dying takes a whole prop dark for the rest of the evening. I wanted hardware that shrugs off the bad-judgment moments instead of turning them into a repair job in the driveway.

So I taught myself PCB design and built the board I wanted to own — one where every failure mode has a specific part standing in front of it, chosen from a datasheet and verified against the copper.

Background, so you know where I'm coming from: day job is senior network architect — infrastructure that isn't allowed to fail. Nights and weekends it's a bench, a soldering iron, and pixels I keep pushing past what the hardware was meant to do. The discipline comes from the day job. The lights are where I get to have fun with it.

**What it is**

The NODE family — three boards, one protected 8-port pixel stage underneath all of them. Three isolated power domains, a fuse on every single output, non-destructive reverse-polarity protection, 5–48 V with no jumpers to change input voltage, 30 A at standard copper, runs stock WLED. Nothing in the power path runs near its limit. What changes between the three is the brain on top and how it gets on the network.

**NODE Mini** — Seeed XIAO ESP32-S3, Wi-Fi. Compact, standard runs, get it on the network and drive pixels.

**NODE Flex** — universal DevKit carrier. Both 0.9″ and 1.0″ row spacings on the board, jumper-mapped to the outputs. Seat whatever ESP32 you already have; swap it when that one goes EOL, so the compatibility list can't go stale. (Olimex and LilyGo use odd row spacing and won't drop straight in — everything else I've tried does.)

**NODE Backbone** — Waveshare ESP32-S3-ETH-POE with a dedicated W5500. Wired 10/100 and PoE for the install that outgrew Wi-Fi. Deterministic frame timing, one cable for power and control to the node.

Design priorities, briefly: protection first — assume it gets abused. Honest specs — real ratings, real derating, trade-offs stated out loud, no spec-sheet trophies that fall apart when someone does the math. Serviceable — fuses you swap by hand, socketed parts, a board meant to be fixed and kept running rather than thrown out when one channel dies. Open — WLED out of the box, docs and board definitions on GitHub.

I'll say plainly that the Bong69 and the QuinLED-Dig-Quad were both inspirations here. I've run them, I like them, and seeing what they did well is a good part of why I thought I could take a swing at my own. This isn't me trying to one-up anybody — it's my take on the same problem, with different priorities in a few places.

**Why I'm posting**

I want people who will actually work it hard and tell me what's wrong. Not looking for a pat on the back — looking for the thing I didn't think of. I've got a handful of boards from the first run and I'll send them out free. No cost, no shipping, no strings, no video required, no NDA.

What I'm hoping for in return is an honest read. Load it up, run a real prop on it, push the channel count, tell me where it flinches. If it turns out my board is worse than what you're already running, I'd rather hear that from you now than after somebody's paid for one (If I decide to try and sell them).

There's no team here to hide behind. If something's wrong with it, it's on me, and I'll make it right.

**What makes someone a good fit**

- Running a show at real scale, or you'll load these channels properly on the bench
- Comfortable in WLED and xLights, and can tell a firmware problem from a hardware one
- You'll actually report back — good, bad, or "this is fine but here's what annoyed me"
- Bonus: you run a Bong69 or a QuinLED-Dig-Quad and can compare directly. I want to know where mine falls short.

**Where things stand**

Stock is thin — this is a small validation run and I'm counting boards on one hand. The test units are free and separate from anything else. I'm looking for people who'll find the problems in this one.

And plainly: this is V1. It works, I've run it, but first-article hardware is first-article hardware. That's the whole reason I want it in the hands of people who'll push it.

I've got full datasheets — power architecture, derating tables against input voltage, and a section on where the Bong69 and QuinLED are the better choice, because in a few real ways they are. Happy to post or link them if there's interest.

Reply here or PM me. And if you want to get into the weeds on the design, I'm happy to — that's the fun part.

Thanks,
John — Wateefy Electronics
github.com/johnvoipguy

 

[domhodgson]

Hey John,

Welcome to the world of controller design!

I've seen your post in a few places and if I can give a bit of feedback (as someone who has wrote a lot about controller testing and marketing)

I'm all for building your own stuff but it's clear you want to release this as a commercial project, but in this text you've not told me WHY I'd use this over any of the other WLED boards, what is different, I see "real ratings, real derating, trade-offs stated out loud, no spec-sheet trophies that fall apart when someone does the math" but let's be honest, that's just ChatGPT marketing copy! You say "it's my take on the same problem, with different priorities in a few places" but what are the priorities? Why would someone use this over the current boards?

Before you send these out to people with 7A/10A labels on the fuses, have you put 7A / 10A of current through those traces, we've tested a LOT of boards and if they aren't traced properly you will cause fires, I'm sure you probably have done a proper bench test but it's surprising how many commercial boards haven't

I'd add a few more specifics in your marketing as well, how many pixels did your testing allow you to run, You've got a 30A fuse on the input but 8 x 7A / 10A fuses on the output, I can tell you from experience, the more information you give to people the better (and the more questions they will ask) but more information is better than less.

Good luck with the board

Dom

 

[Master Of Packets]

The Test version is only 5-36v, and has 50-60v Caps. Version 1b, is relevent to the table below. in addition 1b also has a DMX port, ADC, MCP for WLED monitoring of fuse status and voltages, fuse status leds. DMX protection circuit against 10-kV ESD (IEC 61000-4-2), 4-kV EFT (IEC 61000-4-4), and 1-kVsurge (IEC 61000-4-5) transients. so that DMX gear that is notorius for shorting and frying controllers is dumped at the RJ45 jack. Reverse polarity protestion for backfeeding from outputs, and many other upgrades.

NODE Platform​

Hardware Reference · 8-Channel Addressable LED Controller Family​

Wateefy Electronics

---

The power section is the product. The module is a consumable.

Document rev: 3.0 · Board rev: V1a · Date: 2026-08-06

---

Scope​

This document covers the hardware common to every board in the NODE family — power architecture, input protection, output stage, environmental data and build process. It applies to NODE Mini, NODE Flex and NODE Backbone alike.

What differs between variants — the MCU module, the logic rail source, board dimensions, GPIO mapping and per-board accepted limitations — is documented in each board's own Variant Specification. Where a variant departs from anything stated here, its supplement says so explicitly in Deltas from the Platform Reference.

Read this document for how the board protects your pixels. Read the supplement for which board to buy.

---

1. Power Architecture​

Three separate domains, each on its own copper region:

Domain	Boundary	What it carries
Input (unprotected)	Screw terminal → ideal-diode stage	Raw supply. Assumed hostile. Transient clamp sits across this node.
Clean (protected)	Ideal-diode output → main fuse	Reverse-blocked, surge-clamped. Bulk capacitance lives here.
Distribution bus (fused)	Main fuse → 8 channel fuses + logic regulator	Everything downstream taps this bus through its own fuse.

2. Input & Protection​

Parameter	Value	Notes
Input voltage range	5 – 48 VDC	Single rail, powers both pixels and onboard logic
Input connector	WECO 158-A-111/02, 2-pole screw terminal	57 A / 10 mm² rated; 30 A design maximum = 53% of rating
Maximum continuous current	30 A	At standard 1 oz outer copper — no copper upgrade required
Main fuse	30 A mini blade, Keystone 3568 holder	Holder rated 30 A / 500 V; user-replaceable
Reverse-polarity protection	LM74700-Q1 ideal-diode controller + CSD18540Q5B	FET held off under reverse input; no current path to load; non-sacrificial and self-recovering
Pass element	CSD18540Q5B NexFET	60 V VDS, 205 A ID, ~2.2 mΩ RDS(on) — 15% of current rating at 30 A design load
Input transient clamp	SMCJ48CA bidirectional TVS, 1500 W	48 V standoff, 53.3 V breakdown; conducts in neither direction at rated input
Bulk capacitance	2 × 1000 µF, 63 V	On the clean rail, downstream of the ideal diode
Ideal-diode charge pump	22 nF, 100 V	VCAP to ANODE, floating — not referenced to ground
Ideal-diode hold-up	47 µF electrolytic, 80 V	Between controller and pass FET, per reference design
Polarity indication	Anti-parallel green/red LED pair	Green = correct, red = reversed; shared 1 kΩ limit; LED reverse stress 2 V against 5 V rating

Voltage headroom summary: 36 V nominal, 48 V absolute maximum. Every component in the power path is rated 63 V or above; every semiconductor is 60 V or above.

Derating against input voltage. Utilization of each component's own rating. 5 V, 12 V and 24 V are the mainstream pixel voltages and are shown in bold:

Component	Rating	5 V	12 V	24 V	36 V
Bulk input capacitors	63 V	8 %	19 %	38 %	57 %
Channel output caps	80 V	6 %	15 %	30 %	45 %
Ideal-diode hold-up	80 V	6 %	15 %	30 %	45 %
Buck input electrolytic	80 V	6 %	15 %	30 %	45 %
HV-plane ceramics	100 V	5 %	12 %	24 %	36 %
Ideal-diode controller	65 V	8 %	18 %	37 %	55 %
Pass MOSFET VDS	60 V	8 %	20 %	40 %	60 %
Buck regulator VIN	80 V	6 %	15 %	30 %	45 %

At the 36 V nominal rating no part on the plane exceeds 60 % of its own limit. At the 48 V absolute maximum nothing exceeds 80 %. On the 5 V, 12 V and 24 V strips that carry the overwhelming majority of installations, the entire plane runs under 40 %.

3. Outputs​

Parameter	Value	Notes
Channels	8	Independent, individually fused — no shared fuses
Output connector	Phoenix MKDS-series 3-pole, per channel	V+ / DATA / GND; 12 A rated → 63% at the 7.5 A fitted fuse, 83% at the 10 A maximum
Per-channel fuse	7.5 A mini blade fitted, 10 A maximum — Keystone 3568 holder	25% of holder rating at 7.5 A, 33% at 10 A; user-replaceable
Level translation	SN74HCT541 octal buffer, DIP-20 socket	3.3 V drive vs 2.0 V VIH = 1.3 V margin
Data series resistance	330 Ω, axial through-hole, per channel	Limits fault current back into the buffer
Data ESD protection	PESD5V0F1BL, per channel, connector side	0.4–0.55 pF junction capacitance; RC ≈ 181 ps — no measurable effect on WS281x bit timing
Data drive current	≈ 4 mA per channel	67% of ±6 mA buffer rating
Buffer supply current	≈ 45 mA total	64% of 70 mA device maximum
Per-channel decoupling	47 µF electrolytic 80 V + 100 nF ceramic 100 V	Local to each output connector
Channel activity indication	One LED per channel, 470 Ω limited	≈ 5 mA, 25% of LED rating

4. Environmental & Compliance​

Parameter	Value
Operating temperature	−20 °C to +60 °C (design target; not qualification tested)
Substrate rating	UL-recognized, JLC-1 marking
RoHS	Compliant
Workmanship	IPC Class 2
Electrical test	Flying-probe, 100% of boards
Traceability	2D barcode serial number, per board

Mounting & Enclosure​

No enclosure is included, and none is planned. This is the norm in this segment — bare board, mounted in whatever the installation already uses: a CG1500, a DIN rail, a weatherproof box, a plywood backer.

What NODE adds instead is conformal coating, applied after assembly with connectors, sockets, fuse clips, and switches masked. Competing boards in this class ship uncoated. A coated board in a vented enclosure tolerates condensation and humidity in a way a bare board does not, which matters for the seasonal outdoor installs this was built for.

Four corner mounting holes are provided. Hole spacing and a downloadable mount STL: to be published at first article.

---

Build & Assurance​

• Fabrication — 4-layer, 1.6 mm, high-Tg substrate, ENIG finish, epoxy-filled and capped vias, IPC Class 2 workmanship
• Assembly — professional SMT assembly, flying-probe electrical test on 100% of boards
• Traceability — 2D barcode serial on every board; UL-recognized substrate marking
• Coating — conformal coating applied after assembly, with connectors, sockets, fuse clips, and switches masked
• Design review — every net verified against manufacturer datasheets; schematics independently reviewed by two automated analysis passes with human verification of every finding

 

[AAH]

The first few issues that I see without actually touching a board or seeing details of the schematic or seeing the pcb layout.
The clearance around mounting holes needs to be much bigger. The resistor at the top left will end up with a screw head making contact. If your holes are 3mm then someone is going to drill them out to 1/8". When someone drills to 1/8" it's a sure bet that someone else will go to 3/16" as that's the size of the fasteners they want to use. I've seen the 2.5mm holes for mounting Pi's drilled to both 3mm and 1/8". I've had 3mm holes in my own boards drilled to 1/8" MANY times. I think I now use 3.5mm as that at least covers 3mm and 1/8".
The value of your output resistors is too high. This is going to limit the distance to the first pixel to probably 5m or potentially significantly less.
The the IC details and jumper details on the silkscreen. The board may be made to suit 2 different output buffers but as there is a jumper needed that information should be on the silkscreen.
Presumably the board is meant to be ran via an ESP32 module. That should be indicated on the silkscreen as should it's orientation.
The terminal markings for V- and V+ on the input and 0V, Data, +V for the outputs may be on the silkscreen but obscured by perspective.
Dual wipe tin plated IC sockets are terrible. Machined pin gold plated IC sockets are far more reliable at the expense of a little more care needed when inserted the IC's.