IceRiver KS5L LED Indicator Codes (D1-D4) Diagnostic Guide

Photograph or video the live LED block before doing anything else. All four LEDs in frame, 5 seconds of footage to capture cadence. Once you reboot, the diagnostic resets — you can't recover the original state. This single discipline saves more wasted bench hours than any other Tier 1 step on the KS line. Note which LEDs are flashing in sync, which are solid, and which are off.

Hard power-cycle at the breaker for 30 seconds. Not a soft reboot from the web UI — a full mains kill. Clears wedged daemon state and resets the I2C bus. Watch the LEDs come back up. Healthy boot = brief lamp-test on all four, then `D1` active during self-check, then settle to `D2` flashing alone. If the boot pattern recovers and the original fault doesn't return within 30 minutes, you saw a transient — note it and watch for cadence.

Pull the matching numeric code from `Status` → `Miner Log`. Format: `[timestamp] Error: NNN <description>`. The numeric code disambiguates which subcode within the LED's bucket is firing. `D1`+`D3` could be `110` (fan-in low) or `120` (fan-in stalled) or `140` (overspeed) — same LED pattern, different fix paths. Cross-reference the master IceRiver code reference for the full numeric-code table.

Shop-vac the intake filter and grille. Dust on filters = higher inlet temp = `D2`+`D3` overheat. Five minutes of work resolves a meaningful share of LED-flagged thermal cadences. Wipe the grille, vacuum the intake, verify nothing within 15 cm of the front of the chassis is blocking airflow.

Verify ambient with an IR thermometer at the intake grille. Target ≤ 30 °C for KS5/KS5L/KS5M (per IceRiver's official guidance — see the indicator-light reference). Hot Canadian basements in summer hit 38 °C more often than people think — check before you blame the hardware. If ambient is over 30 °C, the LEDs aren't lying — fix the room before you fix the miner.

Verify Ethernet link before assuming a hardware fault when `D1`+`D2` is flashing. The pattern fires identically whether the cable is unplugged, the switch port is dead, the DHCP lease expired, the internet is out, or the pool host is unreachable. Walk the network path from the miner's RJ45 outward before opening the chassis. Ping the miner from your network, ping `8.8.8.8` from a peer device, ping your pool host.

Replace a failing fan if `D1`+`D3` persists after Tier 1. KS5L uses dual `12038` (120 mm × 38 mm) fans, typical part `DF1203812B2UN` or equivalent at 6000 RPM. Power off at the breaker. Open the chassis. Unscrew the fan from its bracket, unplug the connector, swap, screw back, plug in. Reboot, watch the LED block — `D1`+`D3` should clear within 30 seconds of full boot if the fan was the cause.

Verify AC line voltage at the wall under full mining load if `D4` is solid or if the miner is throwing `2xx` bucket codes. Multimeter on AC, probe at the outlet while the miner is hashing at full power. KS5L PSU accepts `180 V`-`285 V` AC. Sag below `190 V` sustained = circuit undersized or shared with too much else. Move to a dedicated `240 V` circuit — single biggest improvement for North American Canadian installs running KS5L.

Try a firmware rollback through the web UI if the LED faults appeared right after an OTA upgrade. Download the previous IceRiver-published version from `iceriver.io/firmware-download/`, flash it through `System` → `Upgrade` over a wired connection. Most post-upgrade `7xx` and `8xx` cadences clear with rollback to last-known-good.

Reseat hashboard ribbon cables if `D2`+`D3` returns under healthy ambient and clean intake. Power off at the breaker. Open the chassis. Disconnect the ribbon and power connectors at each hashboard, visually inspect the contacts for blackening or oxidation, reconnect firmly. Listen for the click. Sensor-side `300`-`302` codes (which surface as `D2`+`D3` decoded) often clear with this alone.

Refresh thermal paste on the chip-side of the hashboards if `D2`+`D3` returns after Tier 1 thermal cleanup. KS5L chips age fast on dried stock paste. Arctic MX-6 or Thermal Grizzly Kryonaut, uniform thin layer, don't glop it on. Reassemble carefully — torque heatsink screws evenly to avoid PCB bow. `350`-`352` cadences under healthy ambient often clear post-refresh.

SD-card reflash recovery if the LED block stays in a non-documented state and the web UI is unreachable (post-`8xx`-failure). Recovery image from `iceriver.io/firmware-download/`, write to microSD with balenaEtcher or `dd`, follow the model-specific procedure (boot button + insert at power-on for most KS units). Watch the LED block during recovery: healthy reflash = lamp-test on all four LEDs, then `D3` and `D4` flash simultaneously while the image writes, then settle to standard boot pattern. Soak-test for an hour after.

Replace the PSU on a KS5L/KS5M if `D4` is solid under healthy ambient and AC voltage. The `BP-H-3640` family runs close to its rated headroom on the flagship KS5 line. Symptoms: repeating `233`-`239` codes, `D4` solid, intermittent reboots under hash load, audible PSU fan ramp at low ambient. Order a known-good replacement, swap (connector job, not soldering — within reach of an intermediate hands-on owner).

Replace a `1004LV100` ASIC chip if `D2`+`D3` persists after paste refresh and a single board reads zero hashrate or returns partial chip count (`9`/`26`/`52`) on a tester fixture. One or more `1004LV100` chips have failed silicon. Hot-air rework: preheat to 150 °C, hot-air at 310-330 °C for ~30 s, reflux, lift, replace, reflow. Practice on a scrap board first — KS5L hashboards aren't cheap.

Stop DIY and ship to D-Central when: (a) LED block stays in an undocumented pattern after a clean SD-card reflash, (b) `D4` solid persists after PSU swap, (c) all-four-off persists after a 30-second mains kill (control board not booting at all), (d) per-chip failure repeats on two different boards (PCB-level fault), or (e) C19 plug or PSU socket shows scorching. Past those points DIY costs more in time + risk than the repair itself.

Ship safely. Anti-static bags for hashboards, double-box with at least 5 cm of foam on every side. Include the original PSU and AC cable so we can reproduce the AC environment. Print the LED-state photo (or a printout of the state description), the numeric code from the log, and the cadence. The more diagnostic context you ship with the unit, the faster the bench turnaround — and the cheaper the repair.