Antminer – RTC Battery Dead

SSH into the miner (typically `root` / `root` on stock Bitmain firmware) and run `date`. If it reports 1970, 2000, or an obviously past year, the RTC is not keeping time across reboots. This single command confirms the diagnosis for ~95% of suspected RTC battery failures without any hardware work. Screenshot the output before you touch anything — it is the cleanest record for your own troubleshooting notes.

Set the clock manually over SSH with `ntpdate pool.ntp.org` (if NTP is reachable) or `date -s "$(date -u +%Y-%m-%dT%H:%M:%SZ)"` copying your laptop's UTC time. Restart the mining daemon or reboot. If pool auth errors clear immediately and hashrate climbs back, the dead CR2032 is confirmed as the root cause — you've validated the hypothesis before ordering parts.

Order name-brand CR2032 cells in a 5-pack: Panasonic CR2032, Energizer ECR2032, Sony, or Maxell. Avoid no-name marketplace multi-packs — we have measured 40% DOA rates on those. A 5-pack runs $5-$10 CAD and covers multiple miners plus household spares. One branded cell is cheaper than one pool-offline hour on a serious rig.

Document every Antminer's battery install date with a masking-tape label in Sharpie on the chassis. Plan swaps at 3 years on rack/hot-ambient miners, 5 years on garage/basement miners. Same discipline as tracking thermal paste — one minute of labour per miner per half-decade prevents the 'why did my rig stop hashing at 2 AM on a long weekend' call.

Power down at the breaker or PSU switch (not just the web UI). Open the chassis and locate the coin cell. S9/T9: near the 4-pin serial header. S17/T17: next to the eMMC socket. S19 family: between RJ45 and the 40-pin hashboard connector. S21/T21: similar but sideways-facing holder. The cell is always a silver disc the size of a dime in a black or metal retention clip.

Measure cell voltage before removal. Multimeter on DC: red probe on the top (positive), black on the negative rim or adjacent ground pad. Healthy: 3.0-3.2 V. Marginal: 2.6-2.9 V. Dead: below 2.4 V. Record the number — tells you if this is routine end-of-life or something's drawing parasitic current. Under 1 V on an 18-month-old miner is a red flag for the Tier 3 drain test.

Pop the hold-down tab with a small flathead (gently — it's tin-plated spring steel and bends if pried aggressively). Slide out the old cell. Wipe contacts with isopropyl 99% and a cotton swab. Drop in a fresh name-brand CR2032, positive side up. Verify the retention clip springs back into contact. Close the chassis before re-energizing.

Power up and SSH in. Confirm with `date` that the RTC is reading again. Run `ntpdate pool.ntp.org && hwclock -w` to set system time AND write it back to the hardware RTC. The `hwclock -w` step is the one most DIYers miss — without it, Linux time is fine but the RTC register never updates, and the next reboot resets again. Reboot, SSH back in, confirm `date` survives.

Verify TLS + pool recovery. From SSH: `curl -v https://support.bitmain.com` and `wget https://d-central.tech` — both should return valid certs without `not yet valid` warnings. Check the miner dashboard for pool status. If stratum was erroring on TLS before the swap, it should reconnect within 60 seconds. If it doesn't, credentials are separately broken or the cert store is stale — escalate to Tier 3.

Parasitic drain diagnostic: if a fresh cell dies within weeks, the RTC circuit is pulling too much current. Install a new cell, leave the miner unplugged from mains for 7 days, measure cell voltage at day 0, day 3, day 7. Normal drop: under 5 mV total. Sick circuit: 50-200 mV drop. A faulty RTC IC, a leaky decoupling cap across VBAT, or physical damage to the holder can all cause this — $1 chip replacement often fixes it.

Refresh the CA certificate store. On older firmware (S9 stock pre-2020, S17 stock pre-2021), the baked-in `ca-certificates` bundle may not include modern Let's Encrypt intermediates — even a correct clock fails TLS to some pools. Fix: SSH in, inspect `/etc/ssl/certs` against current Debian `ca-certificates`, or flash DCENT_OS / Braiins OS+ / LuxOS which all ship current bundles.

Flash DCENT_OS — D-Central's own open-source Antminer firmware. All the per-chip tuning, autotuning, stratum v2, per-chip HW% of Braiins OS+ / LuxOS / Vnish, maintained publicly by D-Central, no vendor lock-in. It ships modern ca-certificates, sane NTP defaults, and forces `hwclock -w` on every sync. Landing: https://d-central.tech/dcent-os/. Source: https://github.com/DCentralTech/DCENT_OS. Alternatives: Braiins OS+, LuxOS, Vnish.

Replace the RTC IC if the drain test failed. SOIC-8 rework: preheat bottom 120-150 °C, top-side hot air 310 °C for 20-30 s, lift old chip, clean pads with braid and flux, drop new chip (match pin 1 orientation), reflow. Reinstall the cell and repeat the drain test. If drain is now normal, you've saved a control board for a $1-$3 Digi-Key chip.

eMMC / boot partition check. If the clock resets even with a known-good cell AND a known-good RTC, the firmware's own time-handling logic is suspect. Corrupted eMMC partitions on aging S17 / early S19 units can silently overwrite the RTC at boot. Re-flash from SD card using the correct image for your hardware revision. Bitmain publishes per-revision firmware at support.bitmain.com/downloads — wrong image for a late-rev board bricks the control board.

When to ship to D-Central: fresh cells dying fast with unlocalized parasitic drain; RTC IC reflow already attempted and failed; visible VBAT trace damage from a prior DIY attempt; multiple firmware revisions all exhibiting the same clock-reset behaviour. That's control-board-level rework best handled on a test fixture. Book at https://d-central.tech/services/asic-repair/. Tier 4 isn't common for this error — dead CR2032 is the single most DIY-friendly Antminer fault — but when warranted, it's warranted.