A high voltage divider steps down kilovolt-level potentials (e.g., 10kV DC from electrostatic precipitators, CRT flybacks, or capacitor banks) to safe, measurable voltages (e.g., 0–5V) for microcontroller ADCs. Unlike standard low-voltage dividers, high voltage dividers must manage extreme voltage gradients, resistive voltage coefficients (VCR), and surface leakage currents that otherwise destroy measurement accuracy or cause catastrophic arcing.
The direct answer to scaling 10kV down to 5V is a specialized series-parallel resistor network using high-ohmic metal glaze or thick-film resistors, paired with a precision low-side shunt and a PCB guard ring to eliminate leakage errors. Below is the exact topology, failure behavior, and component-level design walkthrough.
Topology, Node Labels, and Extreme Behavior
The standard topology for DC and low-frequency AC high voltage measurement is the resistive divider. The circuit consists of an upper resistor string (R1) and a lower shunt resistor (R2).
- $V_{HV}$ (High Voltage Input): The kilovolt source node.
- $V_{sense}$ (Divider Output): The scaled-down voltage node routed to the ADC or op-amp buffer.
- $V_{guard}$ (Guard Ring): A copper trace driven at the same potential as $V_{sense}$ to eliminate surface leakage currents across the PCB dielectric.
- $GND$ (System Ground): The 0V reference for both the HV supply and the measurement system.
At these voltages, standard series/parallel assumptions break down due to parasitic effects. The table below details exactly what happens when elements fail or degrade at the extremes.
| Fault Condition | $V_{sense}$ Result | Physical Consequence & Risk | Recovery / Design Mitigation |
|---|---|---|---|
| R1 String Opens | Drops to 0V (or floats to ADC leakage voltage) | No HV path. Safe for LV side, but loss of measurement. | Use series string for R1; a single open fails safe. Add a high-value pull-down (10MΩ) at $V_{sense}$ to prevent floating ADC pins. |
| R1 String Shorts | Spikes to $V_{HV}$ (10kV) | Catastrophic. Destroys ADC, microcontroller, and poses lethal shock hazard. | Physically impossible if R1 is built from 20+ discrete series resistors. Never use a single HV resistor for R1. |
| R2 Shunt Opens | Spikes to $V_{HV}$ (10kV) | Lethal hazard. Arcing across R2 pads will vaporize PCB copper. | Place a gas discharge tube (GDT) or 5.1V Zener diode in parallel with R2 to clamp $V_{sense}$ if R2 fails open. |
| R2 Shunt Shorts | Drops to 0V | R1 string dissipates maximum power. If R1 is under-rated, it will overheat and catch fire. | Ensure R1 string total power rating is ≥ 2x the calculated $I^2R$ dissipation at max $V_{HV}$. |
| Surface Leakage (Dirty PCB) | Non-linear reading, drifts with humidity | Microamp leakage across flux residue bypasses R2, introducing 5-20% measurement error. | Implement a $V_{guard}$ copper ring around $V_{sense}$. Clean board with isopropyl alcohol and apply silicone conformal coating. |
Why Resistive Over Capacitive or Inductive Alternatives?
When measuring high voltages, engineers often debate between resistive dividers, capacitive dividers, and potential transformers (PTs). The choice depends entirely on the signal type (DC vs. AC) and the required bandwidth.
For mixed DC/AC or pure DC systems (like battery storage monitoring or DC-DC converters), the resistive topology is the only viable choice. Capacitive dividers block DC entirely, and PTs only operate on AC. Furthermore, as detailed in standard DC circuit theory, resistive dividers provide a predictable, frequency-independent scaling factor, provided parasitic capacitance is managed.
| Criteria | Resistive Divider | Capacitive Divider | Potential Transformer (PT) |
|---|---|---|---|
| Signal Compatibility | DC, AC, and Pulsed DC | AC and Pulsed DC only (Blocks DC) | AC only (50/60Hz line frequency) |
| Continuous Power Dissipation | Moderate (requires thermal management) | Near Zero (reactive impedance) | Low (core and copper losses only) |
| Physical Footprint | Compact (can be PCB mounted) | Moderate (requires large HV capacitors) | Massive (heavy iron core, oil-filled) |
| Cost (10kV class) | Low ($15 - $50 in components) | High ($200+ for HV rated caps) | Very High ($500+ for metering grade) |
Design Walkthrough: 10kV DC to 5V ADC Interface
Let us design a divider to measure up to 10,000V DC and scale it to a 5.0V maximum output for a 12-bit ADC. The required ratio is 2000:1.
Standard carbon or thick-film resistors exhibit a negative VCR, meaning their resistance drops as the applied voltage gradient increases. At 10kV, a standard resistor's value can shift by 5%, ruining your calibration. Always specify metal glaze or specialized thick-film HV resistors with a VCR of < -50 ppm/V.
1. Selecting R2 (The Low-Side Shunt)
We need a high-precision, ultra-stable shunt. Choose a 1.000 MΩ Vishay Z-Foil resistor (e.g., VPR221Z series). Z-foil provides a 0.1% tolerance and a near-zero temperature coefficient (0.2 ppm/°C), ensuring the bottom half of the divider does not drift with ambient temperature changes.
2. Selecting R1 (The High-Side String)
To achieve a 2000:1 ratio with a 1 MΩ R2, R1 must be 1999 MΩ (effectively 2000 MΩ or 2 GΩ for practical purposes).
Critical Error to Avoid: Do not buy a single 2 GΩ resistor. A single resistor body cannot withstand the 10kV gradient; it will arc internally or across its surface. Furthermore, standard 1/4W resistors have a maximum working voltage of ~250V, regardless of their ohmic value.
The Fix: Build R1 using a series string of 20 individual 100 MΩ resistors. Using specialized high-voltage resistors like the Ohmite MOX series or equivalent metal glaze types, each resistor sees a maximum of 500V (10,000V / 20). This keeps the voltage gradient well within the safe operating area of the resistive element and prevents internal arcing.
3. Power and Thermal Verification
- Total Resistance: 2000 MΩ + 1 MΩ = 2001 MΩ
- Current Draw: $I = 10,000V / 2,001,000,000Ω = 4.99 μA$
- Power in R1 String: $P = I^2R = (4.99 μA)^2 \times 2000 MΩ = 49.9 mW$
- Power per R1 Resistor: $49.9 mW / 20 = 2.5 mW$
Since 2.5 mW is vastly below the 500 mW (1/2W) rating of each resistor, thermal drift from self-heating is eliminated. However, the physical spacing between the 20 resistors must be maintained to prevent corona discharge in the air gaps.
4. ADC Drive and Filtering
Microcontroller ADCs use a sample-and-hold capacitor that draws a sudden spike of current when the sampling switch closes. This spike will cause $V_{sense}$ to droop if the divider impedance is too high. Place a 10 nF C0G (NP0) ceramic capacitor directly in parallel with R2. C0G dielectrics are mandatory here; X7R or Y5V capacitors exhibit severe capacitance loss under DC bias and will introduce non-linear distortion into your measurement.
Step-by-Step Low-Voltage Breadboard Validation
Before potting your R1 string in silicone or epoxy, you must verify that no solder joints are cold and that the ratio is mathematically sound. We do this by injecting a safe 10.000V DC signal and measuring the microvolt output.
- Assemble the R1 String: Solder the 20x 100 MΩ resistors in series on a perforated phenolic board (FR4 is acceptable if cleaned thoroughly, but phenolic has higher surface insulation resistance). Leave 5mm of physical clearance between each resistor body.
- Attach R2 and C1: Solder the 1 MΩ Z-foil resistor and 10 nF C0G capacitor in parallel at the bottom of the R1 string. Connect the ground lead.
- Inject Safe Test Voltage: Connect a calibrated bench power supply set to exactly 10.000V DC to the $V_{HV}$ input node. Connect the bench supply ground to the divider $GND$.
- Measure the Scaled Output: Using a 6.5-digit bench multimeter (e.g., Keysight 34461A), measure the voltage across R2. Expected Reading: $10.000V \times (1 MΩ / 2001 MΩ) = 4.9975 mV$.
- Verify Guard Ring Continuity: If your PCB includes a guard ring, inject the 4.99 mV signal into the guard trace via a unity-gain op-amp buffer. Verify with the multimeter that the voltage difference between $V_{sense}$ and $V_{guard}$ is exactly 0.000 mV. Any deviation indicates a short or high-resistance joint in the guard trace.
Once the 10V validation confirms the expected ~5mV output (proving a 2000:1 ratio), the divider is ready for high-voltage integration. For the final deployment, coat the entire R1 string and $V_{sense}$ node in a high-dielectric strength RTV silicone (like Dow Corning 3140) to suppress corona discharge and block ambient humidity from altering the surface leakage paths.
