New “Parallel Resistance Method” Proposed to Replace Unstable EV Insulation Testing Standard
In the high-stakes arena of electric vehicle (EV) safety, a quiet but potentially game-changing shift is underway—one that could redefine how automakers and certification labs across the globe verify a critical, yet often overlooked, safety parameter: insulation resistance.
At first glance, insulation resistance sounds like the kind of engineering detail reserved for technical manuals and compliance checklists. But its implications are far more visceral: it’s the invisible barrier that prevents lethal current from leaping from a high-voltage battery pack—often carrying over 800 volts—into the metal chassis, and ultimately, into a person touching the car’s door handle during a fault condition. When that barrier degrades due to aging, moisture ingress, or mechanical damage, the risk of electric shock, fire, or even fatality rises sharply. That’s why global safety standards, from China’s GB 18384 to the UN’s ECE R100 and ISO 6469, mandate strict minimum thresholds—typically 500 ohms per volt of system voltage—for this parameter.
But here’s the catch: the method prescribed by these standards to measure insulation resistance may itself be fundamentally flawed.
A recent study published in Standard Science, a peer-reviewed journal focused on metrology and regulatory frameworks, has exposed serious instability in the widely adopted “dual-voltmeter method” (often called the “two-meter method”) enshrined in China’s current GB 18384-2020 standard—and by extension, echoed in related standards like GB 38031 (battery safety) and GB 38032 (electric bus safety). The paper, authored by Gao Feng, Li Liangyu, and Yu Qun from the National Passenger Car Quality Inspection and Test Center, along with Li Rui from FAW TOYOTA Research & Development Co., Ltd., doesn’t just point out the problem—it offers a robust, field-tested alternative: the “Parallel Resistance Method.”
This isn’t just a theoretical exercise. For automakers, battery suppliers, and test labs, the implications are immediate and operational. Unstable measurement results mean inconsistent pass/fail decisions, costly retests, production line stoppages, and—most critically—potential safety gaps masked by unreliable data.
So what’s wrong with the current method? And why has it taken until now to raise the alarm?
To understand the flaw, one must first appreciate how the dual-voltmeter method works—and why it seemed so elegant on paper.
The core idea is deceptively simple: an EV’s high-voltage system (the REESS—Rechargeable Energy Storage System, i.e., the main battery) sits electrically isolated from the vehicle’s metal frame (the “electric platform”) by two insulating paths: one from the positive terminal to chassis, and another from the negative terminal to chassis. These are modeled as two resistors, Ri1 and Ri2, with the lower of the two defining the system’s overall insulation resistance, Ri.
The standard method instructs engineers to connect two high-impedance digital voltmeters—one between the positive terminal and chassis, the other between the negative terminal and chassis. They record the two voltages: U1 (higher) and U1’ (lower). Then, a known “test resistor” (R0), typically 1 megohm, is connected across the terminal showing the higher voltage, and the two voltages are measured again: U2 and U2’.
From these four voltage readings—and crucially, the known internal resistances of the two voltmeters (r1 and r2, usually around 10 megohms)—a set of equations derived from Kirchhoff’s Current Law is solved to back-calculate Ri1 and Ri2.
On paper, with ideal components, it works beautifully. In practice, however, the math is astonishingly brittle.
The study’s authors ran repeated tests on a production-intent prototype using an automated measurement system—eliminating human error in reading dials or pressing buttons. The results were alarming. Across two measurement cycles spaced forty minutes apart (with full disconnection in between), the calculated insulation resistance swung wildly: from ~192 kilohms in the first run to a staggering ~1.41 megohms in the second. Worse still, during intermediate calculations, the system reported a negative resistance for Ri1—a physical impossibility, serving as a red flag that the underlying computation had become numerically unstable.
Why does this happen? The paper identifies three intertwined culprits.
First, and most insidious, is the time-domain behavior of real-world EV systems. Modern high-voltage architectures aren’t simple DC circuits. They contain parasitic capacitances—especially so-called “Y-capacitors” used in EMI filters—that form RC networks with the insulation resistances. When the voltmeters and the test resistor are connected, they upset the existing DC equilibrium. The circuit doesn’t snap to a new steady state; it relaxes toward it, governed by time constants (τ = R × C) that can be several seconds or even minutes long due to the extremely high resistances involved (megohms) and non-trivial capacitances (nanofarads to microfarads). During this settling period, the two measured voltages don’t move in lockstep. Instead, they exhibit small, coupled oscillations—like two pendulums weakly linked by a spring—where a tiny rise in U1 coincides with a tiny dip in U1’, and vice versa. A measurement taken a fraction of a second too early or too late captures a different point on this decaying waveform, feeding dramatically different numbers into the calculation.
Second, the assumption that a voltmeter’s internal resistance is a fixed, known constant is often invalid. The authors note that even a meter rated for 10 MΩ can exhibit shifts of nearly 50 kΩ depending on the actual voltage it’s measuring. While 50 kΩ sounds small next to 10 MΩ (a 0.5% change), in the finely balanced algebra of the dual-voltmeter equations, this minor perturbation gets amplified into major output swings. Think of it as trying to balance a pencil perfectly on its tip: a breath of air is enough to knock it over.
Third, there’s the cumulative effect of instrumentation tolerances. A 0.1% error in the test resistor, a 0.05% drift in the voltmeter’s voltage reading, a 0.2% variation in its internal resistance—each is within the spec of high-end lab equipment. Yet when fed into a calculation that involves division by very small differences (e.g., U1 – U1’), these tiny errors compound nonlinearly, yielding massive uncertainty in the final Ri value.
This lack of robustness—the inability of the method to produce consistent results despite minor, unavoidable real-world variations—is the paper’s central critique. A safety test method must be not just accurate in theory, but repeatable and reproducible in the messy reality of a test bay, a production line, or a service garage.
Faced with this instability, the authors propose a paradigm shift: stop fighting the voltmeters. Instead, neutralize their influence.
Enter the “Parallel Resistance Method.”
The core insight is disarmingly straightforward: if the voltmeters’ high internal resistance is the source of instability, why not make their resistance irrelevant? The solution is to deliberately add a pair of “shunt resistors”—Rx—across both the positive-to-chassis and negative-to-chassis paths before any measurements begin. These Rx resistors aren’t arbitrary; they are carefully chosen to be at least two orders of magnitude smaller than the voltmeter’s internal resistance. So, if the meter is 10 MΩ, Rx might be 100 kΩ or lower.
Why does this help? Ohm’s Law provides the answer. When a 100 kΩ resistor is placed in parallel with a 10 MΩ resistor, the combined resistance is approximately 99 kΩ—effectively just the 100 kΩ, since the meter’s contribution is negligible. By adding identical Rx resistors on both sides, you create a new, dominant parallel path for current. The previously critical Ri values (which could be in the megohm range) are now effectively “swamped” in the calculations, while the known, stable Rx values take center stage. The system’s behavior becomes far less sensitive to the tiny parasitic effects that plagued the original method.
The measurement sequence is similar but crucially different. First, with Rx in place, you measure the two voltages (U1, U1’). Then, you add the test resistor (R0) to the higher-voltage side and measure again (U2, U2’). The equations used to compute the equivalent insulation resistance (Ri ∥ Rx) are simpler and, critically, no longer contain the voltmeter’s internal resistance as a variable. Once you have Ri ∥ Rx, extracting the true Ri is a trivial application of the parallel-resistance formula.
Perhaps the most compelling practical advantage? You only need one voltmeter. Since the meter’s internal resistance no longer matters, you can use the same meter to take all four readings sequentially—swapping it between the positive and negative terminals. This slashes equipment costs, simplifies fixturing, and eliminates the need to match or precisely characterize two separate meters.
The paper’s authors note that this method isn’t entirely new—it was previously explored in their 2021 work in Automotive Practical Technology—but its formal proposal as a replacement for the standard method, backed by empirical evidence of the incumbent’s failure, gives it new weight.
Automakers and suppliers are already taking notice. Engineers at FAW TOYOTA, a co-authoring institution, are reportedly evaluating the method for internal validation protocols. Test labs that serve multiple OEMs are intrigued by the promise of faster, more reliable results—reducing the time per test cycle from several minutes (while waiting for voltages to settle) to under a minute.
Of course, any deviation from a mandated standard is a serious proposition. GB 18384-2020 is not just a guideline; it’s cited in China’s mandatory type-approval process for all new EVs. Changing it would require a formal standard revision process, involving industry consensus, regulatory review, and likely international harmonization efforts, given the global nature of automotive supply chains.
But the paper makes a compelling pragmatic argument: even if the standard isn’t revised tomorrow, the knowledge of this instability should change how the test is performed today. The authors urge test engineers to treat dual-voltmeter results with extreme skepticism—especially if the numbers seem out of character or if repeated measurements disagree. They recommend performing multiple trials, allowing ample settling time (far longer than most automated scripts currently use), and cross-validating critical results with alternative methods, such as dedicated insulation resistance testers (meggers) that use different principles.
This story is bigger than a single test method. It’s a reminder that in the rush to electrify transportation, the foundational processes that ensure safety can’t be taken for granted. Every new megawatt-hour of battery capacity shipped brings with it an obligation to verify—reliably, repeatably, and robustly—that the vehicle won’t become a hazard the moment something goes wrong.
The parallel resistance method may not be the final word. Other approaches, such as high-frequency injection techniques or active balancing circuits, are also under investigation. But what Gao Feng and his colleagues have done is vital: they’ve sounded an alarm not with fearmongering, but with data, physics, and a constructive alternative. In the meticulous world of automotive safety, that kind of rigorous, solution-oriented critique is how progress is truly made.
As EVs evolve, gaining higher voltages, more complex architectures, and faster charging speeds, the demand for bulletproof diagnostic and verification methods will only intensify. The insulation resistance test might operate quietly in the background, but getting it right is anything but trivial. It’s the difference between a safety net that holds—and one that unravels under pressure.
Gao Feng, Li Liangyu, Yu Qun (National Passenger Car Quality Inspection and Test Center); Li Rui (FAW TOYOTA Research & Development Co., Ltd.). Standard Science, 2023, Issue 2 (Part 2), pp. 49–51. DOI: 10.3969/j.issn.1002-5944.2023.04.007