Voltage Domains and Regulators in Hash Boards

Introduction

Every ASIC mining hash board faces a fundamental engineering challenge: converting 12V from the PSU into the sub-0.4V supply that each chip needs, while delivering hundreds of amps across the entire board. The solution is to divide the board into voltage domains — isolated groups of chips, each powered by its own dedicated voltage regulator circuit.

Understanding voltage domains is essential for ASIC repair. When a hash board reports missing chips, reduced hashrate, or abnormal power consumption, the fault almost always traces back to a specific voltage domain. Knowing how to identify, test, and repair these domains is one of the most valuable skills in mining hardware diagnostics.

This guide covers the electrical theory behind voltage domains, how buck converters work at the component level, model-specific domain layouts, and practical testing procedures you can apply with a basic multimeter.

What Is a Voltage Domain?

A voltage domain is a group of ASIC chips on a hash board that share a single voltage regulator and a common power rail. Every chip in the domain receives the same regulated voltage, drawn from the same buck converter output.

Think of it as a neighborhood on a street: all the houses (chips) on one block share the same electrical transformer (buck converter). If the transformer fails, the entire block loses power. If one house has a short circuit, the transformer's overcurrent protection may trip and shut down the whole block.

Anatomy of a Single Domain

Each voltage domain consists of these key elements:

• Buck converter IC — the controller that drives the switching MOSFETs
• High-side MOSFET — connects the input voltage to the inductor during the "on" phase
• Low-side MOSFET — provides a freewheeling path for inductor current during the "off" phase
• Power inductor — stores energy in its magnetic field, smooths current delivery
• Output capacitors — filter switching ripple, provide local energy storage
• Feedback resistor network — sets and monitors the output voltage
• ASIC chips — the load, typically 5 to 15 chips per domain depending on the model
• Current sense resistor — allows the controller to measure and limit output current
• Input fuse or trace fuse — protects the domain from catastrophic overcurrent events

The buck converter takes 12V from the PSU input bus and steps it down to the chip operating voltage — typically between 0.25V and 0.35V. This extreme step-down ratio is one of the most demanding aspects of hash board power design.

Why Voltage Domains Exist

You might wonder: why not use a single massive regulator to power all the chips? There are three critical reasons.

1. Current Limits

Each ASIC chip draws between 5A and 15A at its operating voltage, depending on the chip generation and clock frequency. A board with 156 chips (like the S21) would require over 1,000A from a single regulator at 0.30V — that is simply not feasible with any practical component.

By dividing the board into 12 domains of ~13 chips each, each regulator only needs to supply roughly 80–100A. While still substantial, this is within the capability of multi-phase buck converters using modern MOSFETs.

2. Trace Resistance and Voltage Drop

PCB copper traces have resistance. Even at very low resistance values (milliohms), when you push hundreds of amps through them, the voltage drop becomes significant. At 0.30V operating voltage, even a 10mV drop represents a 3.3% variation — enough to cause chip instability.

By keeping each domain's power traces short and localized, the designer minimizes resistive losses. Each domain's regulator is physically placed adjacent to the chips it powers, keeping trace lengths to a few centimeters.

3. Fault Isolation

If a single chip develops a short circuit (one of the most common failure modes), it pulls the voltage rail down toward 0V. With domain isolation, only the 7–13 chips in that domain are affected. The remaining 11 domains continue operating normally.

Without domains, a single shorted chip would collapse the voltage for the entire board — 156 chips down instead of 13. Domains act as blast walls, containing failures to a localized area.

Domain Layout by Model

Each Bitmain model has a specific domain layout tailored to its chip's power characteristics. Understanding the layout for the model you are repairing is essential for targeted diagnostics.

Buck Converter Operation in Detail

The voltage regulators on hash boards are synchronous buck converters — the most common DC-DC converter topology for stepping down voltage at high efficiency. Understanding how they work at the component level makes it much easier to diagnose failures.

The Fundamental Problem

The PSU delivers 12V. The chips need 0.30V. That is a 40:1 step-down ratio. Unlike a linear regulator (which would waste 97.5% of the power as heat), a buck converter uses rapid switching to efficiently convert the voltage.

Circuit Components

A buck converter consists of five essential elements:

       12V Input
          │
     ┌────┴────┐
     │ High-Side│
     │  MOSFET  │ ← Controlled by PWM signal
     └────┬────┘
          │ (Switch Node)
          ├──────[L]──────┬──── Output (0.30V)
          │   Inductor     │
     ┌────┴────┐      ┌───┴───┐
     │ Low-Side │      │ Output │
     │  MOSFET  │      │  Caps  │
     └────┬────┘      └───┬───┘
          │                │
         GND             GND

Switching Cycle Explained

The buck converter operates in two alternating phases, switching at frequencies typically between 300kHz and 1MHz:

The 2.5% Duty Cycle Challenge

A 2.5% duty cycle at 500kHz means the high-side MOSFET is on for only 50 nanoseconds per cycle. This creates several engineering challenges:

• Timing precision — the controller must generate extremely precise, short pulses. Any jitter in the timing directly translates to output voltage ripple.
• Gate driver speed — the MOSFET must turn on and off within nanoseconds. Slow transitions waste power and generate heat.
• Minimum on-time — every MOSFET and controller has a minimum on-time. If the required on-time is shorter than this minimum, the converter cannot regulate properly. This is why some hash boards use lower switching frequencies or multi-phase designs.
• Ripple voltage — at such low duty cycles, the ripple voltage as a percentage of the output becomes significant. An 0.01V ripple on a 0.30V output is a 3.3% variation.

Feedback Loop and Voltage Regulation

The buck converter does not blindly switch at a fixed duty cycle. It uses a closed-loop feedback system:

1. Output voltage sensing — a resistor divider (two precision resistors) scales the output voltage to a level the controller IC can read (typically 0.6V–0.8V reference).
2. Error amplifier — inside the controller IC, the sensed voltage is compared to an internal reference. If the output is too low, the error signal increases. If too high, it decreases.
3. PWM modulator — the error signal adjusts the duty cycle. Higher error = longer on-time = more energy delivered = voltage rises. Lower error = shorter on-time = less energy = voltage drops.
4. Compensation network — capacitors and resistors around the error amplifier set the loop response speed. Too fast and it oscillates. Too slow and it cannot respond to load transients (like a chip suddenly starting a new hashing job).

This feedback loop runs continuously, adjusting the duty cycle thousands of times per second to maintain the target voltage despite changes in load current and input voltage.

Current Limiting and Protection

Hash board buck converters include several protection features:

• Overcurrent protection (OCP) — a sense resistor in the current path allows the controller to measure output current. If it exceeds the limit (typically 120–150% of rated current), the converter shuts down or reduces duty cycle.
• Overvoltage protection (OVP) — if the output voltage exceeds a threshold (typically 120% of nominal), the controller shuts down to protect the chips.
• Undervoltage lockout (UVLO) — if the input voltage drops below a minimum threshold, the converter disables to prevent erratic operation.
• Thermal shutdown — if the controller IC or MOSFETs overheat, the converter shuts down.

Voltage Tuning by Firmware

Modern mining firmware does not run all domains at a fixed voltage. Instead, it dynamically tunes the voltage for each domain to optimize the balance between hashrate, power consumption, and stability.

How Auto-Tuning Works

The control board firmware adjusts domain voltages through a digital interface (typically I2C or SPI to the buck controller ICs, or through the ASIC chip chain). The tuning process works as follows:

Voltage vs. Performance Tradeoffs

The relationship between domain voltage and performance is not linear:

• Too high — chips work correctly but waste power as heat. Efficiency drops. Thermal throttling may occur.
• Optimal — chips hash at full speed with minimal errors and lowest power consumption.
• Too low — chips begin producing hardware errors. Some nonces are computed incorrectly. Effective hashrate drops even though the chips appear to be working.
• Way too low — chips stop responding entirely. The firmware reports them as "missing."

The optimal voltage is typically within a window of only 20–30mV. This tight tolerance is why voltage regulator accuracy matters so much on hash boards.

Testing Voltage Domains with a Multimeter

A multimeter is the most fundamental tool for voltage domain diagnostics. You can identify dead domains, shorted chips, and unstable regulators with basic DC voltage measurements.

Where to Probe

Each domain has specific test points where you can measure the output voltage:

Expected Readings by Model

Here is what you should see when measuring healthy domains:

Interpreting Abnormal Readings

What your multimeter reading tells you:

0V (no voltage at all) The domain's buck converter is not switching. Possible causes:

• Dead controller IC (no gate drive signals)
• Blown input fuse or trace fuse
• Open inductor (inductor winding broken, no current path)
• Controller in shutdown mode due to a protection fault (overcurrent, overvoltage, thermal)
• No enable signal from the ASIC daisy chain or control board

Voltage significantly higher than expected (e.g., 0.45V+ on an S21) The feedback loop is not regulating properly. Possible causes:

• Open feedback resistor (controller sees 0V feedback, drives duty cycle to maximum)
• Shorted chip pulling down the voltage sense point in an unusual way
• Failed controller IC with stuck-high PWM output
• This is a dangerous condition — overvoltage can permanently damage ASIC chips

Voltage significantly lower than expected (e.g., 0.15V on an S21) The domain is loaded beyond the regulator's capacity. Possible causes:

• One or more shorted ASIC chips pulling excessive current
• Shorted output capacitor
• Partially functional regulator (one phase of a multi-phase converter dead)

Fluctuating voltage (reading jumps around) The regulator is unstable. Possible causes:

• Failed compensation network (oscillating feedback loop)
• Intermittent connection (cracked solder joint on MOSFET, inductor, or capacitor)
• Marginal load — the converter is at the edge of its current capability
• Defective output capacitors (ESR too high due to aging or thermal damage)

Voltage on half the domains, zero on the other half The board may have two input power sections (some models split the 12V input). Check the input connectors and fuses for both halves.

Identifying Failed Domains from Symptoms

You do not always need a multimeter to suspect a domain failure. The miner's software provides clues:

Missing Chips in Groups

If the miner's status page shows a block of consecutive missing chips, that strongly suggests a failed voltage domain. For example:

• S21: chips 0–12 missing (Domain 1 dead) while chips 13–155 work
• S19: chips 42–48 missing (Domain 7 dead) while all others work

The key indicator is that the missing chips are consecutive and the count roughly matches the chips-per-domain for that model. Random scattered missing chips suggest a communication chain problem, not a domain failure.

Partial Hashrate Loss

A healthy 3-board S21 miner produces approximately 200 TH/s (around 67 TH/s per board). If one board drops to about 56 TH/s (roughly 83% of normal), that suggests 2 of 12 domains have failed — a 17% loss in line with losing 2 out of 12 domains.

Quick math for domain failure estimation:

Hashrate loss per domain = Board hashrate / Number of domains
S21: ~67 TH/s / 12 = ~5.6 TH/s per domain
S19: ~33 TH/s / 12 = ~2.8 TH/s per domain

Abnormal Power Consumption

A board with a dead domain draws less power than expected. If two of three boards draw 1,400W each but the third draws only 1,200W, the low-power board likely has one or two dead domains. The remaining domains still hash and draw power, but the total is reduced proportionally.

Overcurrent or Overheat Errors

If the miner logs show overcurrent warnings or a specific chain (board) repeatedly triggers thermal protection, a domain-level short circuit may be dragging down performance and generating excess heat in the regulator components.

Domain-Specific Repair Strategies

Once you have identified the failed domain, here are the most common repair scenarios and their solutions.

Failed Buck Converter (Controller IC or MOSFETs)

Symptoms: 0V output on the domain, no switching activity visible on oscilloscope, input fuse intact.

Diagnosis:

1. Check for 12V at the buck converter's input pins — confirms power is reaching the converter.
2. Check the enable pin of the controller IC — should be logic high during normal operation.
3. If input power and enable are present but no output, the controller IC or gate driver is likely dead.

Repair:

• Replace the buck converter controller IC. These are typically small QFN or TSSOP packages. You will need the exact part number — read it from the IC marking or from a known-good board.
• If the MOSFETs are external (not integrated into the controller), check them individually with a diode-mode test. A shorted MOSFET reads near 0 ohms between drain and source.
• Replace any shorted MOSFETs. Use exact replacement parts — the on-resistance and gate charge specifications are critical at these current levels.

Shorted ASIC Chip Pulling Domain Down

Symptoms: Domain voltage significantly lower than expected (but not 0V), regulator components running hot, possible overcurrent errors in firmware logs.

Diagnosis:

1. Measure the domain output voltage — it will be noticeably lower than the other domains.
2. With power off, measure resistance across the domain output capacitors. A healthy domain might read 0.5–2 ohms. A shorted domain will read near 0 ohms.
3. To identify the specific shorted chip, use a thermal camera or apply low voltage (1–2V current-limited to 1A) and feel for the hot chip with your finger (board must be at room temperature first).

Repair:

• Remove the shorted chip using a hot air rework station (350–380°C for lead-free solder).
• If you have a replacement chip, install it. If not, bridge the communication chain connections (CI/CO and RI/RO) across the empty pad to maintain the daisy chain. The domain will operate with one fewer chip but remain functional.
• After repair, re-measure the domain voltage. It should return to the expected range.

Open Inductor

Symptoms: 0V output on the domain, controller IC may be switching (visible on oscilloscope at the switch node), but no current reaches the output.

Diagnosis:

1. With power off, measure continuity through the inductor. A healthy inductor reads near 0 ohms (typically 1–5 milliohms). An open inductor reads infinite resistance (OL on the meter).
2. Visually inspect the inductor for cracks, burn marks, or discoloration.

Repair:

• Replace the inductor with an exact match. The inductance value (typically 100nH–470nH), current rating, and DC resistance are all critical. Using a wrong inductor value will cause the converter to oscillate or produce the wrong voltage.
• Inductor failures are relatively rare. If one fails, inspect the domain for a root cause — an overcurrent event from a shorted chip may have burned out the inductor.

Blown Input Fuse or Trace Fuse

Symptoms: 0V output on the domain, no 12V present at the buck converter's input pins.

Diagnosis:

1. Trace the 12V input path from the board connector to the domain's buck converter.
2. Look for a small chip fuse (often a 0402 or 0603 component) or a narrow PCB trace designed to act as a fuse.
3. Measure continuity through the fuse. An intact fuse reads near 0 ohms. A blown fuse reads infinite (OL).

Repair:

• Replace the fuse with the same rating. If you cannot determine the original rating, check the schematic or use the value from a working domain.
• A blown fuse indicates that an overcurrent event occurred. Before powering on after replacing the fuse, check the domain for shorted chips or shorted MOSFETs. If the root cause is not fixed, the new fuse will blow immediately.

Failed Output Capacitors

Symptoms: High ripple voltage on the domain (visible as voltage fluctuation on a multimeter, more precisely diagnosed with an oscilloscope), intermittent chip errors in the domain, or chips working at reduced performance.

Diagnosis:

1. Visually inspect capacitors for cracks, burn marks, or desoldered connections.
2. With power off and capacitors discharged, measure capacitance if your meter supports it. Ceramic capacitors can lose capacitance due to DC bias effects and cracking.
3. Use an oscilloscope to measure ripple voltage at the output. Healthy domains should show less than 20mV peak-to-peak ripple. Failing capacitors cause excessive ripple (50mV+).

Repair:

• Replace visibly damaged capacitors. Use ceramic capacitors of the same value (check board documentation or measure from a healthy domain).
• If multiple capacitors appear damaged, the domain may have experienced an overvoltage event. Check the buck converter's feedback network before powering on.

Advanced: Multi-Phase Buck Converters

Some higher-end hash boards and newer generations use multi-phase buck converters where two or more switching stages share the load for a single domain. Each phase handles a portion of the total current, reducing stress on individual components and improving transient response.

In a multi-phase design, the phases switch at staggered intervals (e.g., 180 degrees apart for 2-phase, 120 degrees for 3-phase). This effectively multiplies the ripple frequency and reduces ripple amplitude, resulting in a cleaner output voltage.

If one phase of a multi-phase converter fails, the domain may still operate but at reduced current capacity. The remaining phase(s) absorb the full load, running hotter and potentially triggering overcurrent protection under heavy hashing loads. Symptoms include intermittent domain dropouts under load and MOSFET overheating on the surviving phase(s).

Key Takeaways

• A voltage domain is a group of ASIC chips sharing one buck converter and power rail, typically 7–13 chips per domain on Bitmain boards.
• Hash boards use domains because of current limits, trace resistance, and fault isolation — you cannot practically power 150+ chips from a single regulator.
• Buck converters on hash boards operate at extremely low duty cycles (2–3%) to step 12V down to ~0.3V, which demands precise timing and aggressive filtering.
• The feedback loop (resistor divider, error amplifier, PWM modulator) continuously adjusts the duty cycle to maintain the target voltage.
• Firmware dynamically tunes each domain's voltage to optimize the hashrate-to-power ratio, with optimal voltages varying by 20–30mV between domains.
• When testing with a multimeter: 0V = dead regulator, high voltage = feedback fault, low voltage = shorted chip, fluctuating = instability.
• Failed domains cause consecutive missing chips in the miner's status page and proportional hashrate loss.
• Common domain repairs: replace failed MOSFETs/controller ICs, remove shorted chips, replace open inductors, replace blown fuses.

Apply This Knowledge

Now that you understand voltage domain theory and testing, apply these skills:

• How Hash Boards Work — revisit the overall board architecture with your new understanding of power delivery
• Power Delivery Systems — go deeper into PSU-to-board power distribution and connector specifications
• Essential Repair Tools — make sure you have the right multimeter and probes for domain testing
• Your First Hash Board Repair — a step-by-step walkthrough that applies domain testing in a real repair scenario
• Auto-Tuning Explained — understand how firmware optimizes domain voltages in real time
• Reading Schematics — learn to trace voltage domain circuits on actual hash board schematics