Power Delivery Systems in Mining Hardware

Introduction

Power delivery is the backbone of every ASIC miner. A single Antminer S21 consumes over 3,500 watts — roughly the same as two household ovens running simultaneously. All of that energy must travel from the wall outlet through the PSU, across the hash board, through voltage regulators, and into each individual ASIC chip at precisely the right voltage and current.

When power delivery fails, the consequences range from reduced hashrate and missing chips to catastrophic board damage with burnt traces, blown MOSFETs, and destroyed ASICs. Understanding the complete power path — and knowing how to diagnose failures at every stage — is an essential skill for any hash board repair technician.

This guide covers the full power delivery chain: PSU fundamentals, hash board power architecture, switching regulators, protection circuits, common failures, and diagnostic procedures.

PSU Fundamentals for Mining

Why Mining PSUs Are Different

Consumer PC power supplies typically deliver 500–850W across multiple voltage rails (3.3V, 5V, 12V). Mining PSUs are purpose-built for a single job: delivering massive current on a single 12V rail (or 48V in newer designs) with maximum efficiency.

A mining PSU must:

• Deliver sustained high current (200–300A) without voltage droop
• Operate at high efficiency to minimize electricity waste as heat
• Run 24/7 for years without failure
• Provide stable output under rapidly changing load conditions

The 12V Rail

Nearly all current-generation mining PSUs use a single high-current 12V rail. The 12V output is distributed to hash boards through heavy-gauge cables terminated with multi-pin connectors (typically 6-pin or custom multi-pin headers).

Efficiency Ratings: 80 Plus Certification

PSU efficiency measures how much AC wall power is converted to usable DC output versus wasted as heat. Mining PSUs are rated under the 80 Plus certification program:

At mining scale (hundreds of machines), the difference between Gold and Platinum efficiency across a fleet can mean thousands of dollars per month in electricity costs. Most professional mining operations specify Platinum-rated PSUs as a minimum.

Common Mining PSU Models

Hash Board Power Architecture

The Power Path

Power flows through a hash board in a carefully designed chain, with each stage stepping voltage down and distributing current to smaller groups of components. Here is the complete path from PSU output to ASIC chip core:

PSU 12V Output
    │
    ├─── Heavy-gauge cables (10–12 AWG per conductor)
    │
    ▼
Input Connector (6-pin or multi-pin header)
    │
    ├─── Reverse polarity protection (if present)
    │
    ▼
Bulk Input Capacitors (electrolytic, 1000–2200µF)
    │
    ├─── Filter high-frequency noise from cables
    ├─── Provide local energy storage for transient loads
    │
    ▼
Power Distribution Bus (wide copper pours on PCB)
    │
    ├─── Splits into N voltage domain branches
    │
    ▼
Buck Converter (per voltage domain)
    │
    ├─── Steps 12V down to 0.25–0.40V
    ├─── Controlled by PWM controller IC
    │
    ▼
Domain Output Capacitors (ceramic + polymer, low ESR)
    │
    ├─── Filter switching ripple
    │
    ▼
ASIC Chip Array (6–15 chips per domain)
    │
    ├─── Core voltage: 0.25–0.40V @ 10–20A per chip
    ├─── I/O voltage: 1.8V (separate LDO regulator)
    │
    ▼
Ground return path → back to PSU

Input Stage: Connectors and Bulk Capacitors

The input connector is the first point of contact between the PSU and the hash board. These connectors carry the full board current — typically 60–100A per board. The connector and its solder joints are under immense thermal and mechanical stress.

Bulk capacitors sit immediately after the input connector. Their job is twofold:

1. Energy storage — they act as a local energy reservoir, supplying instantaneous current during load transients (when chips ramp up or shut down)
2. Noise filtering — they absorb high-frequency noise conducted through the power cables

Typical bulk capacitors on mining hash boards are 1000µF to 2200µF electrolytic capacitors rated for 16V or 25V. You will usually find 4–8 of these near the input connector.

Voltage Domains

A hash board does not power all its ASIC chips from a single regulator. Instead, the chips are organized into voltage domains — groups of chips that share a common buck converter and voltage rail.

Each domain has:

• Its own buck converter circuit (high-side and low-side MOSFETs, inductor, controller IC)
• Output capacitors for ripple filtering
• A current sense resistor for overcurrent protection
• Independent enable/disable control from the control board

Why So Many Domains?

The domain architecture exists for three critical reasons:

1. Current distribution — A single regulator delivering 0.3V at 200A would require impossibly large inductors and MOSFETs. Splitting into 12 domains means each regulator handles only 15–20A.

2. Fault isolation — A shorted chip takes down only its domain, not the entire board.

3. Voltage accuracy — At 0.3V, even 10mV of error is a 3.3% deviation. Shorter power paths (fewer chips per domain) reduce resistive voltage drop across traces, keeping each chip closer to the target voltage.

Buck Converters: The Workhorse of Mining Power

What Is a Buck Converter?

A buck converter (also called a step-down switching regulator) efficiently converts a higher input voltage to a lower output voltage. In mining hardware, buck converters step 12V down to the 0.25–0.40V range required by ASIC chips.

Unlike a linear regulator that burns excess voltage as heat, a buck converter uses high-frequency switching to achieve efficiencies of 85–95%, even with extreme step-down ratios (12V to 0.3V is a 40:1 ratio).

How a Buck Converter Works

A buck converter has four essential components:

1. High-side MOSFET (Q1) — connects the input voltage to the inductor when ON
2. Low-side MOSFET (Q2) — provides a current path through the inductor when Q1 is OFF
3. Inductor (L) — stores energy in its magnetic field and smooths current
4. Output capacitors (C) — filter voltage ripple to produce a steady DC output

The PWM controller IC drives Q1 and Q2 alternately at a switching frequency of 300kHz–1MHz:

Duty Cycle

The duty cycle determines the output voltage. It is the fraction of each switching period that Q1 is ON:

Duty Cycle (D) = Vout / Vin

For a typical mining domain:
D = 0.30V / 12V = 0.025 = 2.5%

This means Q1 is ON for only 2.5% of each switching cycle. At a 500kHz switching frequency, each period is 2 microseconds, so Q1 is ON for only 50 nanoseconds per cycle. This extreme duty cycle (called a "very low duty cycle") is one of the key design challenges in mining power delivery.

Multi-Phase Buck Converters

Some hash board designs use multi-phase buck converters, where two or more sets of MOSFETs and inductors operate in parallel with staggered timing. Benefits include:

• Reduced ripple — the staggered phases partially cancel each other's ripple
• Thermal distribution — power dissipation is spread across more components
• Faster transient response — more phases can react to load changes more quickly
• Higher effective switching frequency — N phases at 500kHz behave like a single phase at N x 500kHz for ripple purposes

LDO vs. Switching Regulators

Hash boards use two types of voltage regulators, each chosen for different roles based on their characteristics.

Switching Regulators (Buck Converters)

Used for: ASIC core voltage (0.25–0.40V, high current)

LDO (Low-Dropout Linear Regulators)

Used for: I/O voltage (1.8V), PLL voltage (0.8V–1.0V), reference voltages

Where Each Type Appears on a Hash Board

12V Input ──┬── Buck Converter ──→ 0.30V Core Voltage (per domain, high current)
            │
            ├── Buck Converter ──→ 3.3V Logic Rail
            │        │
            │        └── LDO ──→ 1.8V I/O Voltage (per chip or per group)
            │
            └── LDO ──→ 0.8V PLL Reference Voltage

The 1.8V I/O rail is particularly important. It powers the UART/SPI communication interface on each ASIC chip. If the 1.8V rail fails or becomes noisy, chips will fail to respond to commands even though their cores may be functioning. This is a common misdiagnosis — technicians replace "dead" chips that are actually fine, when the real fault is a failed 1.8V LDO.

Current Sensing and Protection Circuits

Mining hash boards incorporate multiple layers of protection to prevent catastrophic damage from electrical faults.

Overcurrent Protection (OCP)

Each voltage domain includes a current sense resistor (typically 0.5–2 milliohms) in series with the power path. The buck converter controller monitors the voltage drop across this resistor to measure current:

I = V_sense / R_sense

Example: 15mV across a 1mΩ resistor = 15A

When current exceeds the OCP threshold (typically 120–150% of rated current), the controller shuts down the domain within microseconds. OCP protects against:

• Shorted ASIC chips
• Shorted output capacitors
• PCB trace shorts (solder bridges, contamination)

Overvoltage Protection (OVP)

OVP monitors the output voltage and shuts down the regulator if the output exceeds a safe threshold (typically 110–120% of nominal). This protects ASIC chips from being destroyed by excessive voltage.

A common OVP trigger scenario: the high-side MOSFET (Q1) fails shorted (drain-to-source short). This connects the 12V input directly to the output rail, instantly destroying every ASIC chip in that domain. Good OVP circuits include a crowbar SCR that shorts the output to ground to clamp the voltage before it reaches destructive levels.

Undervoltage Protection (UVP)

UVP shuts down the domain if the output voltage falls below a safe minimum (typically 70–80% of nominal). Low voltage causes ASIC chips to malfunction, produce incorrect hashes, and draw excessive current as internal logic states become undefined. UVP prevents this cascade.

Thermal Protection

Many buck converter controller ICs include an internal temperature sensor. If the controller die temperature exceeds its thermal shutdown threshold (typically 150°C), it disables all outputs until the temperature falls to a safe level.

Protection Circuit Summary

Common Power-Related Failures

Power delivery failures account for roughly 40–50% of all hash board repairs. Understanding the failure modes helps you diagnose issues quickly and accurately.

1. Blown MOSFETs

Symptoms: Domain completely dead (no voltage output), visible burn marks on MOSFET package, short circuit between drain and source pins.

Root causes:

• Sustained overcurrent from a shorted ASIC chip
• Gate drive failure (controller IC not switching properly)
• Voltage spikes from hot-plugging hash boards (never hot-plug)
• Manufacturing defect (counterfeit or underrated components)

Diagnosis: Measure resistance between drain and source with a multimeter in diode mode. A healthy MOSFET shows high resistance (megaohms) in both directions. A blown MOSFET shows near-zero resistance (short circuit) in one or both directions.

2. Shorted Capacitors

Symptoms: Domain will not start (OCP trips immediately), PSU enters protection mode (clicking or cycling on/off), visible bulging or cracked capacitor.

Root causes:

• Voltage overshoot during startup
• Electrolytic capacitor aging (electrolyte dries out after 3–5 years at high temperature)
• Physical damage from shipping or handling
• Ceramic capacitor cracking from PCB flex

Diagnosis: Disconnect power. Measure capacitance and ESR with an LCR meter. A shorted capacitor reads near-zero resistance. A dried-out electrolytic shows high ESR (> 100mΩ) and reduced capacitance.

3. Failed Voltage Regulators (Controller IC)

Symptoms: Domain output voltage is wrong (too high, too low, or oscillating), domain output is 0V but MOSFETs test good, chip errors only in one domain.

Root causes:

• Feedback resistor drift (changes output voltage setpoint)
• Controller IC internal failure (ESD damage, thermal stress)
• Compensation network component failure (causes oscillation)

Diagnosis: Measure the output voltage with a multimeter. Compare to the expected voltage for that miner model. Scope the output to check for oscillation (requires oscilloscope).

4. Connector Burnout

Symptoms: Intermittent power to board, blackened or melted connector pins, board works when pressed into connector but fails when released, reduced hashrate that fluctuates.

Root causes:

• Loose connector fit (increased contact resistance → heating → oxidation → more resistance → more heating)
• Corroded pins from humidity or chemical exposure
• Connector rated below actual current draw
• Repeated insertion/removal cycles wearing contact springs

Diagnosis: Visual inspection is often sufficient. Measure voltage at the board input connector while the board is under load — compare to PSU output voltage. A drop greater than 0.3V indicates excessive connector or cable resistance.

5. Failed Current Sense Resistor

Symptoms: OCP trips at low load (false overcurrent), or OCP never trips (no protection — dangerous), domain operates at wrong power level.

Root causes:

• Resistor cracked from thermal cycling
• Solder joint failure (hairline cracks)
• Resistor value drifted from sustained overcurrent

Diagnosis: Measure the sense resistor value with a precision multimeter (you need milliohm resolution). Compare to the schematic value. Even small drift (0.5mΩ to 0.8mΩ) can cause significant OCP threshold errors.

6. PCB Trace Damage

Symptoms: Domain intermittently fails, hashrate drops under high ambient temperature (trace expands and crack opens), visible dark spots on PCB surface.

Root causes:

• Traces undersized for actual current (design or manufacturing defect)
• Electromigration (copper atoms physically move under sustained high current density)
• Corrosion from environmental exposure
• Physical damage (dropped board, flexed PCB)

Diagnosis: Visual inspection under magnification. Thermal camera shows hot spots on damaged traces. Measure resistance across the trace and compare to a known-good board.

Measuring and Diagnosing Power Issues

Essential Measurements

A standard digital multimeter (DMM) is sufficient for most power delivery diagnostics. Here are the critical measurements and what they tell you:

Interpreting Measurements

Advanced Diagnostics (Oscilloscope)

Some power faults are invisible to a multimeter because they involve high-frequency behavior:

• Switching waveform analysis — verify the buck converter is switching at the correct frequency and duty cycle
• Ripple measurement — excessive output ripple (> 50mV peak-to-peak) indicates failing output capacitors or a compensation network problem
• Startup sequencing — some boards require domains to power up in a specific order; incorrect sequencing causes lockup
• Transient response — how the regulator responds to sudden load changes; ringing or oscillation indicates instability

Power Path Diagram: Complete System View

Understanding the full power path from the wall to the ASIC chip helps contextualize every component's role:

┌──────────────────────────────────────────────────────────┐
│                    AC MAINS (240V)                        │
└──────────────────────┬───────────────────────────────────┘
                       │
                       ▼
┌──────────────────────────────────────────────────────────┐
│                    PSU (APW12)                            │
│  AC Input → EMI Filter → PFC → LLC Resonant → 12V DC    │
│  Efficiency: 94%    Output: 12V @ 290A (3,480W)          │
└──────────────────────┬───────────────────────────────────┘
                       │  3x cable bundles (one per board)
            ┌──────────┼──────────┐
            ▼          ▼          ▼
     ┌──────────┐┌──────────┐┌──────────┐
     │ Board #1 ││ Board #2 ││ Board #3 │
     │  ~97A    ││  ~97A    ││  ~97A    │
     └────┬─────┘└──────────┘└──────────┘
          │
          ▼
┌──────────────────────────────────────────────────────────┐
│                 HASH BOARD INTERNALS                      │
│                                                          │
│  12V ─→ [Bulk Caps] ─→ Power Distribution Bus            │
│                          │                               │
│          ┌───────────────┼───────────────┐               │
│          ▼               ▼               ▼               │
│     ┌─────────┐    ┌─────────┐    ┌─────────┐           │
│     │Domain 1 │    │Domain 2 │    │Domain 12│           │
│     │Buck Conv│    │Buck Conv│ .. │Buck Conv│           │
│     │12V→0.3V │    │12V→0.3V │    │12V→0.3V │           │
│     └────┬────┘    └────┬────┘    └────┬────┘           │
│          ▼               ▼               ▼               │
│     ┌─────────┐    ┌─────────┐    ┌─────────┐           │
│     │13 ASICs │    │13 ASICs │    │13 ASICs │           │
│     │~0.3V ea │    │~0.3V ea │    │~0.3V ea │           │
│     │~8A each │    │~8A each │    │~8A each │           │
│     └─────────┘    └─────────┘    └─────────┘           │
│                                                          │
│  3.3V Rail ─→ LDO ─→ 1.8V I/O (communication bus)       │
│  3.3V Rail ─→ LDO ─→ 0.8V PLL (clock generation)        │
└──────────────────────────────────────────────────────────┘

Key Takeaways

1. Mining PSUs deliver extreme current (200–300A at 12V). Respect the energy — always disconnect before working on boards.

2. Hash boards use voltage domains to divide power delivery into manageable, fault-isolated segments. Each domain has its own buck converter, protection circuits, and group of ASIC chips.

3. Buck converters operate at extreme duty cycles (2–3%) to step 12V down to ~0.3V. Understanding the switching cycle helps diagnose regulator faults.

4. LDOs handle low-current, noise-sensitive rails (1.8V I/O, PLL reference). A failed 1.8V LDO causes communication failure that mimics dead chips.

5. Protection circuits (OCP, OVP, UVP) are your board's immune system. When they trigger, find the root cause — do not bypass them.

6. The most destructive failure is a shorted high-side MOSFET, which puts 12V on a 0.3V rail. Always check MOSFETs before replacing dead ASICs in a domain.

7. Connector burnout is progressive and preventable. Regular inspection of power connectors is essential preventive maintenance.

8. A multimeter can diagnose 80% of power faults. Systematic voltage measurements across the power path quickly narrow down the fault location.

Apply This Knowledge

Now that you understand the power delivery chain, put this knowledge into practice:

• How Hash Boards Work — see how power delivery fits into the complete hash board architecture
• Voltage Domains & Regulators — deep dive into domain-level regulation and tuning
• Essential Repair Tools — the measurement equipment you need for power diagnostics
• Your First Hash Board Repair — hands-on walkthrough using power measurement techniques
• Antminer S21 Hashboard Repair — model-specific repair guide with power path reference
• Antminer S19 Pro Hashboard Repair — S19-specific voltage domain layout and common failures