With the spread of battery-electric propulsion, the cabin-heating systems of modern electric cars are increasingly built around positive temperature coefficient (PTC) thermistor heaters. These devices offer fast warm-up, self-limiting behaviour, low surface temperature and good energy efficiency, so they have become a preferred solution for the heating, ventilation and air-conditioning pack of many electric cars. However, PTC heaters are also power-electronic converters. They contain high-voltage power stages, insulated-gate bipolar transistor (IGBT) bridges, DC-DC modules, communication interfaces and digital controllers inside one compact enclosure. The switching actions of these circuits generate significant electromagnetic interference (EMI), and the close geometric spacing of the power stages, filter inductors and control wiring leads to pronounced near-field coupling. Unchecked, such coupling can disturb the sensitive sensors, battery-management units and infotainment systems that are distributed through an electric car. Consequently, an understanding of the coupling mechanism and a systematic suppression workflow are essential for making PTC heaters compliant with automotive electromagnetic-compatibility (EMC) requirements.
EMI appears through a complete interference chain: a disturbance source, a coupling path, and a victim receiver. The source is usually the change of voltage and current with time; the path can be a conducted route or a radiated route; the victim is any circuit whose operating margin is insufficient. For an electric car, many electronic modules share the same chassis ground and the same high-voltage bus, so a PTC heater that generates wideband noise can pollute the common power network and radiate energy directly to nearby wiring. An effective solution therefore has to combine source-level treatments, propagation-path attenuation, and receiver hardening. In my study, I addressed the problem from the perspective of power integrity (PI) and signal integrity (SI) while paying particular attention to the near-field coupling of the PTC elements. I performed simulations before verification tests, in order to find an efficient route from noise identification to suppression.
EMC regulations for electric cars are strict. Standards such as CISPR 25 and ISO 11452 define the test methods and limits that vehicle components must satisfy. Testing a physical prototype inside a semi-anechoic chamber can expose serious problems; however, repeated prototype modifications are expensive and delay the development cycle. Therefore, simulation-based analysis at an early stage is very valuable for electric cars. Full-wave electromagnetic tools, circuit-level signal-integrity solvers and power-integrity analyzers let the designer predict the problematic frequency bands before a printed-circuit board (PCB) is manufactured. In this research, I used a commercial simulation platform to build a high-frequency IGBT model, to characterize the switching behaviour that produces EMI, to analyze the transmission-line reflections and the power-distribution-network impedance, and to evaluate the effectiveness of common-mode and differential-mode filters. The final step of the investigation was a near-field coupling experiment performed on a real PTC heater, where I validated two countermeasures: a shielding cover and a decoupling capacitor network.
1. Background and research context
The traction system of an electric car uses high-voltage batteries, motor inverters and DC-DC converters. These subsystems switch large currents at high speed, and the voltage slew rate (dv/dt) and current slew rate (di/dt) become prominent sources of disturbance. Many investigations reported in the EMC literature agree that the rapid commutation of power semiconductor devices is one of the primary causes of both conducted EMI and radiated EMI. The situation in a PTC heater is similar because the heater is basically a switched-mode power converter. An IGBT bridge chops the high-voltage DC input; the resulting pulse-width-modulated (PWM) waveform contains rich harmonic components whose amplitudes depend on the voltage level, switching frequency, rise time and parasitic ringing. The harmonic energy flows into the battery harness, into the control signal lines and into the air as radiation. Since the PTC heater is often installed close to other electronic modules in the front compartment of an electric car, the near-field coupling path cannot be neglected.
There are generally two classes of EMI inside PTC heaters: conducted interference and radiated interference. Conducted interference propagates along the high-voltage cables, the low-voltage supply lines, the LIN bus and the ground return paths. Radiated interference is generated by the loop antennas that are formed by PCB traces and cables. In my work, I particularly treated the near-field region, which is the zone where the distance between the source and the victim is smaller than λ/(2π). In this region, the electric and magnetic fields behave differently from those in the far field. A source with high dv/dt acts like an electric dipole, while a source with high di/dt acts like a magnetic dipole. The wave impedance is no longer constant. Thus, knowing whether the field is predominantly electric or magnetic is important for designing effective shielding and filtering countermeasures.
The EMC study of PTC heaters has to integrate work on standards, simulation and suppression devices. Tests using a line impedance stabilization network (LISN) measure conducted emission from 150 kHz to 30 MHz; antenna measurements cover the radiated emission range. In an electric car, the switching noise from one module can be coupled through common-mode current into another module’s harness. This common-mode current is the reason why a simple differential-mode filter cannot always solve the problem. A complete filter topology should include both common-mode chokes and capacitors. In order to reduce the risk of overdesign, the filter must be chosen according to the actual interference spectrum and the target impedance. For my investigation, I adopted a target-impedance design approach and implemented it using simulation software to obtain quantitative guidance before laboratory validation.
2. EMI generation mechanisms in PTC heaters
2.1 Power-plane noise coupling
Modern integrated circuits in PTC control boards contain millions of transistors, but only a small number of pins are assigned to power supply. All internal logic gates share the same power bus, so noise on that bus couples into every branch of the circuit. The origin of power-plane noise can be divided into several contributors. The voltage regulator itself has a residual ripple and a limited bandwidth; the regulator cannot respond instantaneously to fast load-current changes; and the impedance of the power path creates voltage drops whenever a transient current flows. A simplified expression of the voltage fluctuation ΔV that appears between the power and ground terminals is written as:
$$ \Delta V = Z\,\Delta I $$
where ΔI is the dynamic current drawn by the active devices and Z is the equivalent impedance of the supply network seen from the load. To keep the core voltage of a microprocessor inside its specified tolerance, the supply impedance must be smaller than the so-called target impedance. In an electric car, the temperature-control processor inside the PTC heater might be disturbed by the switching of the power stage that uses the same printed-circuit-board ground plane. This interference can disturb the clock, cause erroneous logic states and even trigger a false protection action.
The noise margin is an important measure of robustness. It is the difference between the minimum acceptable high-level voltage and the maximum actual noise voltage, or the difference between the minimum actual noise voltage and the maximum acceptable low-level voltage. In my calculations, I used the noise-margin concept to determine whether an EMI peak is a real risk. If the measured noise floor is far below the threshold, the circuit has a large margin; if the margin is small, even a modest increase of interference can cause malfunction. This concept also helped me compare the effect of different countermeasures.
2.2 Mechanism of EMI generation
The fundamental source inside a PTC heater is the abrupt commutation of current and voltage. The IGBT in the high-voltage heater bridge is turned on and off according to a fixed switching frequency. During switching transitions, the voltage across the parasitic inductance Lσ and the charging current into the parasitic capacitance Cp follow the laws:
$$ \Delta V = L_{\sigma}\,\frac{\mathrm{d}i}{\mathrm{d}t} $$
$$ \Delta I = C_{p}\,\frac{\mathrm{d}v}{\mathrm{d}t} $$
In these equations, a large di/dt produces a voltage spike if a parasitic inductance exists in the commutation loop, and a large dv/dt produces a current spike if parasitic capacitance couples to the ground. The BUCK-derived topology inside the PTC heater contains two discontinuous current loops. During the on-time of the switch, current flows from the input capacitor through the switch, the inductor and the load before returning to the source. During the off-time, the inductor current freewheels through the diode. The transition between these two states occurs within tens of nanoseconds, so the spectral content of the generated signal extends up to tens of megahertz.
To quantify this mechanism, I selected a representative automotive IGBT module that matches the voltage and current ranges used in PTC heaters. The main parameters that influence switching behaviour are given in table 1. The internal stray inductances LS and LsCE, the input capacitance Cies, the output capacitance Coes and the reverse transfer capacitance Cres determine the switching speed and the ringing frequency.
| Parameter | Symbol | Typical value | Unit | Condition |
|---|---|---|---|---|
| Collector-emitter voltage | UCE | 750 | V | TJ = 25 °C |
| Continuous collector current | IC | 450 | A | TJ,max = 175 °C |
| Main stray inductance | LS | 20 | nH | — |
| Stray inductance between collector and emitter | LsCE | 8 | nH | — |
| Internal resistance from collector to emitter | RCC+EE | 0.75 | mΩ | TJ = 25 °C |
| Internal gate resistance | RGint | 0.7 | Ω | TJ = 25 °C |
| Input capacitance | Cies | 80 | nF | UGE=0 V, UCE=50 V, f=1 MHz |
| Output capacitance | Coes | 1 | nF | UGE=0 V, UCE=50 V, f=1 MHz |
| Reverse transfer capacitance | Cres | 0.3 | nF | UGE=0 V, UCE=50 V, f=1 MHz |
| Rise time | tr | 80 | ns | UCE=400 V, IC=450 A, Rg=2.4 Ω |
| Fall time | tf | 50 | ns | UCE=400 V, IC=450 A, Rg=5.1 Ω |
I included the junction capacitances CGE, CGC and CCE in the equivalent circuit, together with the internal gate, emitter and collector stray inductances LG, LE and LC. The resulting model makes it possible to simulate how different operating conditions change the collector-emitter voltage spectrum UCE(f), which is the most relevant antenna source in a PTC heater.
2.3 Simulation of the switching waveform influence
Four switching parameters were studied in my simulation: duty cycle, rise time, switching frequency and the ringing that appears on the rising edge. The results are summarized in table 2. The frequency spectrum of a trapezoidal pulse depends strongly on these parameters, and the influence is not limited to the low-frequency envelope; the high-frequency components above 1 MHz also change considerably.
| Parameter under study | Variation range | Observed effect on UCE(f) |
|---|---|---|
| Duty cycle D | 0.1 to 0.8 | A longer on-time moves the first break-point toward lower frequencies and raises the low-frequency magnitude of UCE. |
| Rise/fall time tr | 10 ns to 400 ns | A shorter rise time broadens the spectral envelope, shifting the second break-point from about 0.8 MHz (400 ns) to about 32 MHz (10 ns), thus substantially increasing high-frequency EMI. |
| Switching frequency fsw | 1 kHz, 10 kHz, 100 kHz | The magnitude of UCE increases with switching frequency; the interference is more severe at 100 kHz than at 1 kHz. |
| Ringing on rising edge | with/without ringing | Ringing creates an extra resonance around 62 MHz and significantly extends the spectrum up to 108 MHz. |
My simulation results are useful for the design of the gate driver. For example, if the heater can tolerate a slightly slower turn-on, increasing the external gate resistance can reduce the high-frequency spectral content and therefore reduce the conducted emission measured at the LISN. In many electric cars, the PWM frequency of the PTC heater is fixed for thermal reasons, so the rise-time and ringing parameters are the most practical degrees of freedom.
2.4 LISN measurement concept
Conducted emission measurements on a PTC heater are normally performed with an LISN inserted between the power source and the equipment under test (EUT). The LISN provides a defined high-frequency impedance of 50 Ω, isolates the EUT from the background noise of the power grid, and couples the disturbance voltage to an EMI receiver. For electric cars, separate LISNs are installed in the high-voltage positive and negative lines and in the 12-V power lines. In my study, the LISN model was also used in the simulation environment so that the simulated conducted voltage had the same meaning as the voltage measured in the laboratory. The LISN parameters included a 1-μF coupling capacitor, a 5-μH inductor and an RC branch with 0.1 μF and 1 kΩ; this network shapes the input impedance and allows the noise to be measured over a frequency range of 150 kHz to 30 MHz.
3. Signal integrity and power integrity analysis for PTC heaters
3.1 Transmission-line behaviour
In the control board of the PTC heater, the digital signal lines are usually routed as microstrip or stripline structures. At high speed, the physical interconnects must be treated as transmission lines rather than ideal wires. A transmission line can be modelled by distributed resistance R, inductance L, conductance G and capacitance C per unit length. According to the lossless approximation commonly used in high-frequency digital design, the characteristic impedance Z0 is expressed as:
$$ Z_{0} = \sqrt{\frac{L}{C}} $$
If the characteristic impedance is not matched by the source impedance or the load impedance, part of the signal energy is reflected at the discontinuity. The reflection coefficient at the load is:
$$ \rho_{L} = \frac{Z_{L}-Z_{0}}{Z_{L}+Z_{0}} $$
and the reflection coefficient at the source is:
$$ \rho_{S} = \frac{Z_{S}-Z_{0}}{Z_{S}+Z_{0}} $$
When the impedance is matched, the reflection coefficient becomes zero. If the impedance is not matched, multiple reflections produce ringing, overshoot and undershoot. In a PTC heater, these reflections might be observed on the PWM gate signals and on the LIN communication lines. The signal-transition time and the propagation delay determine whether the reflection problem is visible. When the round-trip delay exceeds about one-sixth of the rise time, the reflected waveform no longer merges into the incident edge and the signal integrity degrades significantly. In my simulation I changed the data rate and the line length and observed that the transmitted pulse developed clear overshoot and long-duration ringing when the line delay became comparable with the rise time.
3.2 Suppression of reflection by termination
Four termination strategies were tested in the simulation:
Series termination. A resistor RS is inserted close to the driver. The driver output impedance R0 plus RS is supposed to equal Z0. I selected the resistance as:
$$ R_{S} = Z_{0} – R_{0} $$
In my test, a 30-Ω series resistor produced the best waveform; a 10-Ω resistor only partly reduced the ringing, whereas a 50-Ω resistor introduced an undershoot because the total source impedance became too high.
Thevenin termination. Two resistors, RTH and RTL, are placed at the far end of the line, pulling the node toward the supply rail and ground respectively. Their parallel combination must equal Z0:
$$ \frac{R_{TH}\,R_{TL}}{R_{TH}+R_{TL}} \approx Z_{0} $$
This scheme supplies extra dc current and can be used when the driver has weak output capability. My simulation showed that it suppresses repeated reflections efficiently, but the static power consumption is higher than in series termination.
RC termination. A series resistor R at the far end is followed by a capacitor C to ground. R is equal to Z0, while C blocks dc power dissipation. The RC time constant should be at least twice the load delay. My simulation compared capacitors of 10 pF, 20 pF, 50 pF and 100 pF. A larger capacitance slows the rising edge of the received signal; an excessively small capacitance causes overshoot. The distance between the RC network and the receiver also matters: a longer delay produces more reflection, so the termination components should be placed close to the receiving pin.
Diode termination. Schottky diodes clamp the received voltage to the supply rail and ground. This method does not require exact impedance matching; it prevents the signal from going too far above the high-level limit or below the ground level. My simulation indicated that the waveform at the receiver is effectively clamped, although the reflected energy is not absorbed and eventually decays through resistive losses and diode conduction.
From these results, I concluded that no single termination method is universally optimal. Series termination is recommended for point-to-point traces of moderate length because it is simple, saves power and requires only one resistor. For bidirectional buses or transmission lines where the load impedance cannot be defined precisely, a diode clamp or RC termination provides additional protection. For the PTC heater PWM gate lines, a series resistor close to the gate driver gave the best trade-off between damping and efficiency.
3.3 Power distribution network and target impedance
The power distribution network (PDN) of a PTC heater consists of the voltage regulator, the bulk capacitors, the decoupling capacitors, the PCB planes and the connecting vias. When the output current changes, the impedance of the PDN produces a voltage drop. If the impedance is considered in the frequency domain, the voltage fluctuation is the product of the current spectrum I(f) and the impedance spectrum Z(f):
$$ V(f) = I(f)\,Z(f) $$
To keep the supply voltage stable, the impedance below the target value must remain within the required frequency range. The target impedance can be approximated by:
$$ Z_{\mathrm{target}} = \frac{V_{\mathrm{DD}}\,\mathrm{ripple}}{I_{\mathrm{transient}}} $$
In my design, the peak current of the tested chip was 88 mA and the maximum allowed ripple was 5 % of the 3.3-V supply, which gave a target impedance of about 50 Ω. The original PDN impedance exceeded 80 Ω at several resonance frequencies, so decoupling capacitors were necessary. The selection of the capacitor value was based on the resonant frequency that needs to be damped and on the equivalent series inductance (ESL) of the capacitor. The self-resonant frequency of a capacitor is:
$$ f_{r} = \frac{1}{2\pi\sqrt{L_{\mathrm{ESL}}C}} $$
Below the self-resonant frequency, the capacitor behaves as a capacitor; above it, the parasitic inductance dominates. In the simulation, I used a 470-μF capacitor whose self-resonance lies near 232 kHz, and a 100-nF capacitor whose self-resonance lies near 26 MHz. An important part of my work was to optimize the value and the location of these decoupling elements so that the PDN impedance curve falls below the target over a wide bandwidth.
3.4 Common-mode and differential-mode filter design
The conducted noise from a PTC heater can be separated into differential-mode current that flows in the same direction along the forward and return conductors and common-mode current that flows in the same direction on both conductors and returns through the parasitic ground path. For differential-mode noise, the interference is generated by the discontinuous current loops; for common-mode noise, the high dv/dt nodes inject displacement currents into ground through parasitic capacitances. I first identified the differential-mode and common-mode equivalent circuits of the PTC heater together with the LISN impedance. Then I designed two families of filter topologies: CL, LC, CLC and LCL. The performance criteria were the insertion loss in the LISN voltage and the volume occupied by the magnetic components.
| Filter topology | Differential-mode result near 1 MHz | Common-mode result over 150 kHz-108 MHz |
|---|---|---|
| CL | 46 dB reduction at 0.9 MHz; resonance shifted to 0.96 MHz | Spurious peak at 440 kHz with amplitude 75 dBμV |
| LC | 38 dB reduction at 0.9 MHz; resonances shifted to 0.56 MHz | Good high-frequency attenuation; 62 dB reduction at 30 MHz |
| CLC | 43 dB reduction at 0.9 MHz; resonance shifted to 1 MHz | Highest low-frequency attenuation below 500 kHz; resonance peak at 440 kHz |
| LCL | 36 dB reduction at 0.9 MHz; additional harmonic peaks at multiples of 1 MHz | Most consistent suppression over the full band; largest volume |
From the simulation results it was clear that the LCL filter offers the best differential-mode insertion loss in the frequency range below 2 MHz, but it requires two inductors, which increases the weight and volume of the PTC heater. For the particular layout of an electric-car cabin heater, space is constrained. Therefore I selected the CL topology for the differential path and the LC topology for the common-mode path. The capacitive branches were realized with X and Y safety capacitors that satisfy the leakage-current limits imposed by vehicle standards. After adding the filter, the simulated LISN voltage in the band from 150 kHz to 2 MHz became lower than the limit, and the remaining high-frequency noise could be handled by a well-placed decoupling capacitor.
4. Near-field coupling modelling and suppression
4.1 Near-field and far-field concepts
When the distance r between the radiating source and the observation point is smaller than λ/(2π), the observation point lies in the near-field region. Here the field properties strongly depend on the kind of source. A high-voltage node can be modelled as an electric dipole of length l. If the dipole carries current I, the electric and magnetic field components in the near zone are:
$$ E_{\theta} = -\frac{I l \sin\theta}{4\pi\omega\varepsilon_{0} r^{3}} $$
$$ H_{\varphi} = \frac{I l \sin\theta}{4\pi r^{2}} $$
and the wave impedance is capacitive:
$$ Z_{w} = \frac{E_{\theta}}{H_{\varphi}} = -j\,\frac{1}{\omega\varepsilon_{0} r} $$
A high-current loop of area S behaves as a magnetic dipole. In its near field, the wave impedance is inductive:
$$ Z_{w} = \frac{E_{\varphi}}{H_{\theta}} = j\,\omega\mu_{0} r $$
Thus, for a PTC heater, the IGBT collector node and the switching node of the BUCK converter generate electric-field coupling through the dv/dt, while the discontinuous loop currents generate magnetic-field coupling through the di/dt. The distinction is important because a metal shield that is effective for the electric field may be insufficient for the magnetic field unless the shield provides a low-reluctance path for the magnetic flux. In my study, both coupling types were observed on the power plane, and my countermeasures were chosen accordingly.
4.2 Near-field interference source identification
For a typical PTC heater, the main near-field sources are the switching node of the DC-DC converter, the gate-drive loop, the freewheeling diode path and the high-voltage busbar that carries the PWM current. The near-field coupling currents flow through a group of parasitic paths between the power module and the digital section of the PCB. Figure 4-6 in the original description of my study showed two alternating conduction states of the input rectifier; during each state, the coupling path was completed through the diode bridge and the parasitic capacitance of the IGBT module. The equivalent circuit of the converter with a boost inductor Lboost and input capacitor Cin was used to deduce the differential-mode voltage gain that determines how efficiently the noise is transmitted to the LISN. In the critical-conduction mode, the influence of diode reverse recovery is smaller; therefore I focused on the magnetic coupling caused by the main current loop and the electric coupling caused by the switching node.
I extracted the relevant signal traces from the PTC controller PCB and imported the layout into the simulation environment. The transmission-line parameters that characterize the traces were obtained from the layer stack, including the characteristic impedance and the propagation delay. I then simulated the signal waveform with and without the near-field coupling contribution. When the coupling was intentionally excluded, the waveform was ideal. When the coupling was activated, the signal exhibited a clear overshoot of about 15% of the supply voltage, a deep undershoot of about 10% below ground, and a ringing oscillation whose amplitude exceeded 20% of the signal swing. The eye diagram at the receiver became much narrower: the eye height decreased by about 40%, and the eye width decreased by about 35%. This large degradation made it clear that the near-field coupling of a PTC element is not a secondary effect. It must be considered during the design phase of any electric-car heater controller that hosts high-voltage power switches and sensitive measurement circuits on the same board.
4.3 Simulation of noise distribution on the PCB
Using Cadence Sigrity, I set up the complete power model of the PTC heater board. For every chip, the current profile was defined; the VRM model was applied to the power input with a very small resistor that emulates the low-impedance supply; voltage observation points were added at critical pins; and the frequency scan was carried out from 0 to 3 GHz. The three-dimensional field result showed that the highest noise area was located around the IC power pins. The impedance between the power and ground planes exhibited two strong resonances at 1.05 GHz and 1.30 GHz. Each resonance exceeded 80 Ω and created a large coupling path for the switching noise.
To suppress the two resonances, I added one decoupling capacitor of 1 nF close to the noise source. The value was chosen from the relation between the resonant frequency and the parasitic inductance of the capacitor. After the capacitor was mounted at the exact noise location, the near-field coupling noise voltage was reduced to about 10% of its original value (an approximate 90% reduction). The self-impedance of the power distribution network dropped below 0.2 Ω over the frequency range of interest, whereas it had been above 80 Ω before the correction. These numbers are summarized in table 4.
| Quantity | Before decoupling | After adding 1 nF decoupling capacitor |
|---|---|---|
| Peak PDN self-impedance | > 80 Ω | < 0.2 Ω |
| Observed noise voltage reduction | Reference level | ~90 % decrease |
| Critical resonances | 1.05 GHz, 1.30 GHz | Resonances removed |
| Near-field coupling effect | Eye height reduction 40% | Signal integrity restored |
This simulation-guided approach is especially valuable for electric cars because a heating unit is frequently installed close to radio antennas and telematics units. The reduction of the PDN impedance at high frequencies reduces the propagation of common-mode surface currents to the harness, which improves the radiated emission behaviour as well as the immunity of the board itself.
5. Near-field coupling measurement and verification
5.1 Test setup
In order to verify the simulation results, I carried out a near-field coupling measurement on the PTC heater in a semi-anechoic chamber that satisfies the EMC test requirements for automotive components. The ambient temperature was kept between 22 °C and 28 °C, and the relative humidity was maintained between 50 % and 60 %. The heater was powered by a high-voltage DC source at 680 V and by a low-voltage supply at (13.5 ± 0.5) V. All auxiliary instruments, including the EMI receiver, the power supply and the control personal computer, were placed outside the chamber to preserve a low ambient noise floor. The high-voltage lines and the low-voltage lines were laid straight and isolated from the ground plane by insulating supports. A grounded metal plane served as the reference ground. The receiving antenna was located at a distance of 1 m from the heater and at a height of 1 m above the ground plane. Different antennas were used to cover the complete frequency range of interest, as shown in table 5.
| Frequency range | Antenna type | Measured polarization |
|---|---|---|
| 150 kHz – 30 MHz | Rod antenna | Vertical only |
| 30 MHz – 200 MHz | Biconical antenna | Vertical and horizontal |
| 200 MHz – 1 GHz | Log-periodic antenna | Vertical and horizontal |
| 1 GHz – 2.5 GHz | Horn antenna | Vertical and horizontal |
The measurement procedure followed the general principle of the CISPR 25 standard. The EMI receiver recorded both peak and average quasi-peak levels, and the margin was defined as the difference between the measured level and the disturbance limit. During the tests, a CAN communication tool controlled the heater through an optical-isolated interface; a pneumatic blower that was driven by compressed air supplied the airflow to the heater. In this way, the electromagnetic background from blowers and computers was eliminated from the chamber environment.
5.2 Baseline measurements and analysis
The first set of tests was performed without any special countermeasure. The rod-antenna measurement from 150 kHz to 30 MHz revealed several narrowband peaks that exceeded the limit. The major critical frequencies and their margins are shown in table 6. Because the measured peak exceeded the applicable limit by up to 20 dB, the margin was negative and the device was clearly non-compliant in this band.
| Frequency (MHz) | Peak level (dBμV/m) | Limit (dBμV/m) | Margin (dB) |
|---|---|---|---|
| 0.25 | 34.74 | 21.00 | -13.74 |
| 0.99 | 31.19 | 18.00 | -13.19 |
| 1.73 | 46.20 | 38.00 | -8.20 |
| 5.80 | 45.44 | 38.00 | -7.44 |
| 5.80 | 35.36 | 18.00 | -17.36 |
| 5.94 | 41.26 | 38.00 | -3.26 |
| 6.30 | 49.49 | 38.00 | -11.49 |
| 6.32 | 38.61 | 18.00 | -20.61 |
In the 30 MHz to 200 MHz range, the biconical antenna found an obvious parasitic peak at about 173 MHz. The peak value at 173.40 MHz was 23.56 dBμV/m, 7.56 dB above the corresponding limit, and the value at 173.55 MHz was 30.20 dBμV/m, which was 4.20 dB above the limit. Most other frequencies in this range were below the limit by a comfortable margin. In the 200 MHz to 1 GHz range, the log-periodic antenna measurement showed another pair of narrowband peaks around 244 MHz and 245 MHz, with margins of -1.71 dB and -4.79 dB respectively. Above 1 GHz, the horn antenna did not reveal any point that exceeded the limit; all measured amplitudes were more than 5 dB below the threshold. The summarized baseline results are given in table 7.
| Frequency band | Critical frequencies | Maximum excess over limit | Observation |
|---|---|---|---|
| 150 kHz – 30 MHz | 0.25, 0.99, 1.73, 5.80, 5.94, 6.30, 6.32 MHz | up to 20.6 dB | Narrowband switching noise and harmonics |
| 30 MHz – 200 MHz | 173.40, 173.55 MHz | 7.56 dB | Cable resonance near FM band |
| 200 MHz – 1 GHz | 244.75, 244.80 MHz | 4.79 dB | High-order harmonic of the switching frequency |
| 1 GHz – 2.5 GHz | None | Below limit | Compliant |
The patterns in the baseline spectrum confirmed that the most dangerous coupling frequencies are located at the resonance frequencies of the cable harness and the enclosure. The physical explanation is that the PCB and the internal wire harness form a quarter-wave or half-wave structure whose electrical length at those frequencies matches the interference wavelength. The low-frequency peaks below 10 MHz are caused by the conducted switching harmonics that reach the high-voltage bus and are then radiated by the cable. The coupling is therefore a mixture of conducted and radiated mechanisms, which is typical for any power-electronic module in electric cars.
5.3 Shielding as a near-field countermeasure
The low-frequency band from 150 kHz to 30 MHz is dominated by magnetic-field coupling, so I used a full-coverage copper shield around the PTC heater. Copper has high electrical conductivity, and eddy currents induced in the shield generate an opposing magnetic flux. The absorption loss increases with frequency, which makes copper suitable for the 150 kHz to 30 MHz region. The PTC heater was wrapped with a grounded copper foil of sufficiently low resistance, and the mechanical assembly ensured continuous electrical contact between the shield and the grounded housing. After the shielding was installed, the rod-antenna measurement was repeated. The sharp peaks that had previously exceeded the limit were all suppressed. The margins in the band were positive; typical values are reported in table 8.
| Frequency (MHz) | Peak level (dBμV/m) | Limit (dBμV/m) | Margin (dB) |
|---|---|---|---|
| 1.2385 | 43.91 | 56.00 | 12.09 |
| 1.2475 | 47.31 | 56.00 | 8.69 |
| 1.2565 | 45.08 | 56.00 | 10.92 |
| 1.2565 | 8.04 | 36.00 | 27.96 |
| 1.4995 | 43.10 | 56.00 | 12.90 |
| 1.7110 | 8.32 | 36.00 | 27.68 |
| 1.7560 | 8.17 | 36.00 | 27.83 |
| 6.1795 | 3.48 | 32.00 | 28.52 |
The shielding solution was effective because it interrupted the magnetic-field path and also reduced the capacitive coupling from the internal switching node to the harness. However, a metal shield is not always the cheapest solution for every band of an electric-car heater. At higher frequencies, the shield’s grounding inductance can excite secondary resonances, so an alternative approach using decoupling capacitors was tested as a second countermeasure.
5.4 Decoupling capacitor network for high-frequency suppression
In the bands above 30 MHz, the problem is associated with high-frequency currents that travel along the power bus and find a return path through parasitic capacitance. To suppress this near-field coupling, I applied the decoupling capacitor network at the power input terminals of the PTC heater. The capacitor network functions as a local energy reservoir and as a low-impedance shunt for the high-frequency disturbance. I selected the capacitor values with the target-impedance method. The target impedance was calculated as:
$$ X_{\mathrm{MAX}} = \frac{V_{\mathrm{DD}}\,\mathrm{Ripple}}{\Delta I_{\mathrm{MAX}}} $$
For frequencies above the bandwidth of the main voltage regulator, the required capacitance is:
$$ C = \frac{1}{2\pi f\,X_{\mathrm{MAX}}} $$
In this case, one large capacitor of 32 μF was combined with 63 small capacitors of 32 nF. The small ceramic capacitors had low equivalent series resistance and low equivalent series inductance. They were placed as close as possible to the fast-switching devices in order to minimize the area of the high-frequency current loop.
After the capacitor network was added, the measurements from 30 MHz to 200 MHz and from 200 MHz to 1 GHz were repeated. The peaks at 173 MHz and 244 MHz were greatly reduced. The margins after modification are reported in table 9 and table 10.
| Frequency (MHz) | Peak level (dBμV/m) | Limit (dBμV/m) | Margin (dB) |
|---|---|---|---|
| 54.35 | 27.71 | 40.00 | 12.29 |
| 54.80 | 26.91 | 40.00 | 13.09 |
| 55.75 | 28.47 | 40.00 | 11.53 |
| 91.05 | 6.02 | 18.00 | 11.98 |
| 94.05 | 5.55 | 18.00 | 12.45 |
| 162.80 | 22.32 | 27.00 | 4.68 |
| 171.00 | 15.56 | 16.00 | 0.44 |
| 171.65 | 22.66 | 26.00 | 3.34 |
| Frequency (MHz) | Peak level (dBμV/m) | Limit (dBμV/m) | Margin (dB) |
|---|---|---|---|
| 200.00 | 13.89 | 16.00 | 2.11 |
| 200.05 | 21.54 | 26.00 | 4.46 |
| 208.00 | 21.20 | 26.00 | 4.80 |
| 208.00 | 12.22 | 16.00 | 3.78 |
| 224.00 | 12.52 | 16.00 | 3.48 |
| 244.25 | 10.54 | 16.00 | 5.46 |
| 245.00 | 17.80 | 26.00 | 8.20 |
| 300.35 | 25.11 | 44.00 | 18.89 |
Both countermeasures therefore achieved their purpose. The shielding approach was more suitable for the low-frequency magnetic field, whereas the decoupling capacitor network was cheaper and easier to apply at higher frequencies. In practice, a PTC heater for electric cars might combine both methods: a compact shield over the power switch area and a well-designed array of ceramic capacitors near the power connector. The final measured spectrum satisfied the standard limits in every evaluated frequency band.
Conclusion
In my study, I investigated the EMI behaviour of PTC heaters used in electric cars and proposed a systematic workflow that moves from noise-source analysis to signal-integrity simulation, power-integrity optimization, near-field-coupling modelling and finally laboratory verification. The principal results can be summarized as follows.
First, the switching operation of the IGBT in a PTC heater generates a disturbance whose spectral content is strongly affected by the duty cycle, rise time, switching frequency and ringing. The simulation model with parasitic capacitances and inductances made it possible to explain why the radiated EMI from the heater is spread over a wide frequency band. A smaller rise time and multiple ringing on the rising edge increase the high-frequency envelope and make compliance more difficult.
Second, the transmission-line reflections inside the controller PCB were analysed using four termination approaches. Series termination was the most efficient for the gate-drive signals, while the diode clamp provided an additional safeguard. The PDN impedance was optimized through a target-impedance method and common-mode and differential-mode filters were designed using CL, LC, CLC and LCL topologies. The CL type was chosen for the differential-mode filter because it offers sufficient suppression with a small volume; the LC type was selected for the common-mode filter for similar practical reasons.
Third, the near-field coupling model showed that the layout of a PTC heater should not be treated as several isolated subcircuits. I observed a reduction of the eye height by about 40% when the coupling was included, and the power-plane self-impedance exceeded 80 Ω at two frequencies above 1 GHz. By inserting a 1-nF decoupling capacitor at the noise source, the near-field coupling noise voltage was reduced by about 90% and the impedance was lowered below 0.2 Ω. This result confirmed that the position of the decoupling capacitor is more important than merely increasing the total capacitance.
Finally, the laboratory test in a semi-anechoic chamber validated the numerical predictions. The PTC heater without countermeasure failed in several frequency bands: multiple peaks from 150 kHz to 30 MHz, a peak near 173 MHz, and a pair of peaks around 244 MHz. A copper shield completely cured the low-frequency band, while the decoupling capacitor network reduced the high-frequency peaks to acceptable levels. After the countermeasures, all measured margins became positive. These findings demonstrate that a combination of source-level filtering, low-impedance power distribution and careful shielding can make a PTC heating unit compliant with automotive EMC limits, thereby helping to protect the safe and reliable operation of electric cars.
There are still topics that deserve further investigation. The interaction between the PTC power stage and the EMI filter is a complex three-dimensional electromagnetic problem that cannot always be described by a simple equivalent circuit. In the future, I plan to extend the near-field-coupling model to include the housing, the wiring harness and the complete vehicle environment. I also want to study the coupling between the main power converter and the control section when the DC-DC stage operates in different modes, because the switching-frequency dithering that is often used in electric cars may create additional sidebands around the expected harmonics. Another promising direction is the use of active gate-drive shaping to reduce the dv/dt and di/dt of the IGBT without sacrificing the switching loss. Combining such a source-level method with the passive filter and shield measures discussed in this thesis could lead to a more compact, cheaper and more reliable EMC solution for PTC heaters and for many other power converters of electric cars.