Panel is completely black? "HUB75E" was never the thing to check — new compatibility guide + call for panel reports

[engmung]

If your LED panel is completely black — correct wiring, correct firmware, serial log reporting success — the build probably isn't the problem. The panel's driver ICs are.

Patternflow scans the panel directly from the ESP32-S3. No sending card, no receiving card, nothing in between. That works for one whole class of panel and cannot work for another, and the two look identical from the outside: same HUB75E ribbon, same 128×64, same P2.5, same 320×160 mm.

HUB75 / HUB75E is a connector and a set of signal names, not a protocol guarantee. Like a USB-C plug — same shape, many different things behind it.

The catch that makes this hard

The obvious advice is "check the driver IC before you buy." We audited our own BOM-linked listing — the panel we know works — to see whether that's even followable:

Where you'd look	What's there
Title / spec table	Resolution, pitch, SMD 2121, 5 V/30 W, HUB-75, 1/32 scan, dimensions. No driver IC
Marketplace AI summary	Restates the above. No driver IC
Attached PDF "user manual"	EU compliance boilerplate. No driver IC
Buyer reviews	🎯 ESP32 confirmations, seller-contact reports, photos of the board

So the practical method isn't hunting for a part number. It's reading the reviews.

What to actually do

Before you buy — search the reviews for ESP32 / Arduino / Raspberry Pi / HUB75 / WLED. On our BOM listing that surfaces an auto-extracted "compatible with Raspberry Pi" tag, a buyer driving it from an ESP32, and a buyer who questioned the seller first and then had an ESP32-S3 displaying images. Somebody has usually already run the exact test you care about. Review photos often show the back of the panel too.

When it arrives — flash stock and see. PANEL_PROFILE defaults to PANEL_STANDARD, which sends no init sequence at all and drives most panels. Only if it comes up completely dark do you change it. Don't pick a setting from the part number: FM6124 needs no init and FM6126 does, one digit apart.

Full guide, with the chip tiers, a copy-paste question for the seller, a symptom→cause table, and how to tell the LED driver from the buffer and the row driver on a board:

📖 docs/panel-compatibility.md

Shipped in #258.

Verified panels so far

Panels actually run on Patternflow, not inferred from datasheets — all on PANEL_STANDARD, no init:

LED driver IC	Marking	Notes
FM6124	FM6124EJ / A012527	The BOM-linked 128×64 P2.5 320×160 panel
ICN2037 (Chipone)	ICN2037BP / D19337NA, D01170MA	Two panels, same chip, different lots
DP5125D	DP5125D / 3G2608	Row driver on the same board: SM5166P

Two of these — ICN2037 and DP5125D — have no entry in the ESP32-HUB75-MatrixPanel-DMA driver list at all, and work anyway. Absence from a supported list is not evidence a panel won't work. The hard stop is the S-PWM architecture, not a missing part number.

Help make this list real

Running a panel other than the BOM-linked one? Please comment here — working or not. A one-line "ICN2038S, lit up on stock" is genuinely useful.

Ideal report: driver IC part number and its lot-code line, silkscreen model string (e.g. P2.5(2121)-32S-320X160-V6.1), resolution + scan rate, which PANEL_PROFILE you used, and a close-up of the back of the PCB. Those go straight into the table above.

(Changed the panel resolution rather than the panel — a 64×64, a chain of two? That's #224, a different question.)

Two chips are currently flagged as disputed and first-hand evidence would settle them:

• RUL6024 — the library README calls it S-PWM, the DMD_STM32 chart calls it FM6126-compatible.
• DP3364 / DP3364S — circulates in buyer's guides as S-PWM but appears in none of the sources we checked.

Credit

Every panel I happened to own worked, so this was never on my radar. @SimonePDA ran into the failure, raised it, went and researched it properly, and proposed what to do about it — which opened up a part of Patternflow I'd been blind to. A panel that lights up is where every build starts. Thank you.

[engmung]

Sourcing note: classic-chip panels may be narrowing to P2.5

A field report passed on privately, recorded here unverified — same standing as the disputed chips above.

The claim, in two parts:

1. Classic shift-register chips are now concentrated almost entirely in P2.5 indoor panels.
2. At other pitches the market has largely moved to S-PWM, far enough that sourcing a classic-chip panel at a different pitch can mean buying secondhand.

Why it matters if it holds

The guide currently frames driver ICs as something you check per listing. If this is right, pitch itself is a coarse pre-filter — and the BOM-linked P2.5 panel is not merely a known-good pick, it sits in what may be the last pitch where new classic stock is routine.

It would also make #224 harder in practice than in principle. Changing PANEL_RES_W/H is three lines of config, but that assumes a compatible panel at your chosen pitch is actually purchasable — which is a supply question, not a firmware one.

Please confirm or contradict

This is one report and it deserves corroboration before it goes into docs/panel-compatibility.md:

• Bought a new (not used) classic-chip panel at a pitch other than P2.5 recently? Which pitch, which driver IC, roughly when, and where from?
• Hit the opposite — a P2.5 that turned out to be S-PWM?
• Regional differences are interesting too. Availability may not look the same from every market.

Either answer is useful. A contradiction is arguably more useful, since it would mean the current per-listing advice is still the right shape.