Fuel Pump Relays, SSRs & PDMs: PWM Engineering for Motorsport Fuel Delivery

We use pulse-width modulation (PWM) to control high-output motorsport fuel delivery and fuel-system heat generation. Modulating the duty cycle of suitable in-tank turbine fuel pumps—such as Walbro 460, 525 and 535 models or the Bosch BR540—can reduce current draw, fuel recirculation and low-load heat generation.

Duty cycle and frequency describe different things. Duty cycle is the proportion of each cycle for which the power switch is on; frequency is the number of cycles per second. At 500 Hz, each cycle lasts 2 ms. An 80% power-switch duty cycle gives 1.6 ms on and 0.4 ms off. Motor current can continue through the freewheel path during the off interval. This is a timing example, not a recommended frequency or minimum duty for every pump.

A conventional fuel pump relay provides on/off switching. A suitably specified solid-state relay (SSR) uses semiconductor switching and can support PWM without mechanical contacts. A power distribution module (PDM) generally controls and protects multiple electrical circuits; PWM capability, channel limits and inductive-load suitability depend on the specific PDM. A single AI.SSR25 relay is not a complete multi-channel PDM.

Misunderstanding motor inductance, carrier frequency or semiconductor switching can lead to noise, excess heating, pump stalling or controller failure. We explain the electrical and hydraulic considerations here and introduce our Aftermarket Industries AI.SSR25.

We use application examples to explain calibration, not to set universal limits for every pump or ECU. Before commissioning, confirm the component datasheets, wiring instructions, fuel-pressure requirements and thermal limits for your installation. Our product specifications must be read with their operating and installation conditions; they are not a substitute for application-specific testing.

A brushed DC permanent-magnet pump motor can be represented by winding inductance L, winding resistance R and rotational back-EMF. When a solid-state switch chops its supply, winding current cannot change instantaneously.

The inductive voltage law describes current change. Our simplified ripple expression assumes continuous current, an ideal freewheel path and approximately linear current ramps. It is not reliable when long on/off periods allow current to decay to zero or when resistance and back-EMF dominate the waveform. Increasing carrier frequency often reduces ripple, but we verify the actual pump-current waveform rather than extrapolate this approximation outside its assumptions.

Our worked inductance range of 40–80 µH is an illustrative assumption, not a measured or manufacturer-verified value for the Walbro 460, 525 or 535. We cannot use it to prove a minimum PWM frequency for those pumps. At 200 Hz a cycle lasts 5 ms; at 500 Hz it lasts 2 ms. These longer cycles can allow larger current swings than higher-frequency operation, depending on the actual motor and freewheel circuit.

Why low frequency can increase pump heat: winding copper loss depends on RMS motor current, not just average current. At the same average motor current and winding resistance, additional ripple increases I_RMS² × R. The power-stage switching loss may fall at lower frequency while the motor's winding loss rises. A cool relay does not prove that the pump is running efficiently or safely.

Worked illustration—not a Walbro benchmark: a 10 A average motor current with 10 A peak-to-peak triangular ripple has approximately 10.41 A RMS current. Its copper loss is about 8.3% higher than perfectly smooth 10 A DC through the same resistance. We do not assign this ripple or heat increase to a particular pump without measurements. Pump case temperature alone is not winding temperature.

• Variables: V_BAT is battery voltage in volts; D is duty cycle from 0 to 1; L is armature inductance in henries; f_PWM is carrier frequency in hertz; ΔI_pp is peak-to-peak motor-current ripple. R is winding resistance at the relevant operating temperature.
• RMS resistive heating: average DC current alone does not establish winding heat. Ripple affects RMS current, and copper losses increase with its square.
• Brush and commutator loading: torque pulsation and current transitions can increase mechanical and electrical stress. Do not assume a frequency is harmless solely because the motor turns.
• Audible resonance: the pump body or fuel tank can amplify switching-related vibration.

For Walbro/TI Automotive 460, 525 and 535 pumps, we do not present 200–500 Hz as a preferred default, 1 kHz as a proven minimum, or 10 kHz as a proven ideal. The TI Automotive F90000285 and F90000295 public performance sheets cited below do not specify a minimum or optimum power-PWM frequency. We have no matched, pump-specific thermal benchmark establishing those thresholds. Identify the exact part number rather than relying only on a marketing flow name.

Microchip's AN905 gives 4–20 kHz as a general brushed-DC motor rule of thumb and recommends experimenting with frequency. This supports evaluating a higher-frequency candidate such as 10 kHz when the pump, controller and ECU allow it; it does not establish a Walbro-approved setting. We treat 1 kHz as an intermediate test point and 10–20 kHz as a candidate range to validate, not as guaranteed pass/fail limits.

Our practical selection process is to check the component limits, then compare current ripple, peak/RMS current, delivered flow, pressure, system power, temperatures and noise at matched hydraulic operating points. Choose the setting that meets the required performance and safety margins; do not select it from a controller-only watt estimate. If the ECU cannot command a suitable validated frequency, use a compatible output/controller arrangement rather than assume the ECU's available range is harmless.

Do not confuse an ECU command frequency with the motor power-carrier frequency. An intelligent pump module may accept a low-frequency speed command while internally switching or filtering power differently. For direct SSR power switching, verify the waveform at the pump terminals; a module's acceptable command frequency is not automatically approval for direct power PWM at that frequency.

At genuine continuous 100% duty there are no ongoing carrier on/off transitions. Frequency comparisons must therefore include partial-duty operation. Also, 10 kHz is still within the human audible range: it can change the sound, but does not guarantee silence. Noise alone does not establish winding temperature or efficiency.

Our AI.SSR25 has a maximum potential PWM frequency of 100 kHz. That potential maximum is distinct from the 10–20 kHz operating range we discuss here and is not a recommendation to command every load at 100 kHz.

A continuous-current label does not establish an SSR's suitability for high-frequency motor PWM. We consider industrial DC solid-state relays such as the Crydom D06D100 and i-Autoc KSJ30D100-L against their own switching-time, inductive-load and thermal specifications—not an advertised current rating alone.

We calculate MOSFET conduction and switching loss using the simplified model below. It is not a complete measurement of module heat: freewheel-diode, gate-driver and other circuit losses must also be considered.

Switching loss occurs during transitions when voltage across the power switch and current through it overlap. Turn-on and turn-off specifications can include control and propagation delays; they are not automatically the voltage/current rise and fall times required by our loss equation. A long command delay alone does not establish how much switching heat a relay produces.

Slow-transition illustration: at 14.5 V and 21.3 A, assuming a combined rise-and-fall time of 150–180 µs gives approximately 5.79–6.95 W of switching-only loss at 250 Hz. With our 35 ns / 40 ns transition example, the same model gives approximately 0.0029 W at 250 Hz. This demonstrates the effect of transition time under matched assumptions; it is not a measured comparison or a verified transition-time specification for either named industrial relay.

Fast switching reduces, but does not eliminate, switching loss. At 14.5 V and 21.3 A, with a 35 ns rise time and 40 ns fall time, we calculate approximately 0.012 W at 1 kHz and 0.174 W at 15 kHz. These are switching-only estimates, not total module heat.

A 1.18 W total-heat figure at 15 kHz requires defined supporting test conditions, including conduction, freewheel-diode and driver losses, operating temperature and mounting. We do not present that figure as a verified measurement or a fixed value across all loads, duties or frequencies.

For a fair device comparison, we separate published specifications, assumed calculation inputs and measured results. We would need matched load voltage, current, frequency, duty cycle, temperature and mounting, plus relevant switching waveforms, to substantiate a numerical total-heat advantage. We do not extend a relay's loss model beyond its permitted frequency or load conditions.

How we use the selection matrix: first check the required PWM frequency, then calculate each pump's peak current draw and check continuous, transient and thermal limits. YES indicates stated controller capability within the applicable limits—not approval for every pump or ECU, a pump-specific minimum, or an optimum frequency. NO indicates a documented frequency limit that rules out that operating range. NOT VERIFIED means we lack the exact-model evidence and must not assume either support or failure. The controller's switching-only estimates do not establish whether a Walbro pump runs hotter at 200–500 Hz than at 1 kHz or 10 kHz.

For an application needing 1 kHz or higher PWM, the Crydom D06D100 is outside its published 700 Hz maximum. Our AI.SSR25 has stated higher-frequency capability, but requires one controller per pump and an individual peak-current suitability check. The i-Autoc KSJ30D100-L datasheet does not specify a PWM frequency rating. Its maximum turn-on and turn-off times are each 0.3 ms at the stated 25°C conditions; these are not separately specified rise/fall transition times and do not establish an approved maximum carrier frequency. We therefore do not mark 1 kHz or 10–20 kHz support as either YES or NO from this timing alone.

Timing and heat notes: overall turn-on/off times must not be substituted for rise/fall transition times. The SSR25's 35 ns / 40 ns values are our stated transition example, and its watt figures are calculated switching-only estimates at 14.5 V and 21.3 A. They exclude conduction, diode and driver heat and are not measured competitor benchmarks. NOT SPECIFIED means a feature or operating limit is not stated in the cited manufacturer's datasheet, not proof that every product revision lacks it.

Crydom manufacturer reference: Sensata/Crydom D06D Series, fourth-generation datasheet, input/output specifications and General Note 5. The D06D100's 700 Hz PWM rating applies to resistive loads with at least 8 VDC control voltage; at maximum frequency, output current must be reduced to 75% of its rating. Inductive loads require diode suppression. Datasheet: https://www.mouser.com/datasheet/2/657/d06d-series-dc-panel-mount-1371031.pdf

i-Autoc manufacturer reference: KSJ Series, Rev.4.0 dated 27 July 2022, pages 1–3 and General Notes on page 7. We use the listed KSJ30D100-L variant; confirm the actual relay marking and revision rather than assuming an unsuffixed or customised unit is identical. Its output specifications at 25°C list a 0–24 VDC load range, 100 A maximum load current, 250 A peak surge at 10 ms, and 2.1 mΩ maximum on-state resistance. The control input is 4–32 VDC. The 100 A label requires the thermal derating curves and appropriate heatsink mounting with thermal grease or a pad; it is not a bare-module current guarantee. The 250 A surge rating is a withstand specification, not an active overcurrent-trip threshold or approval for repetitive pump-start pulses.

The i-Autoc datasheet specifies internal overvoltage protection, but still requires an external fast-recovery flywheel diode across an inductive load. Internal TVS protection is not the same as a motor-current recirculation path. The -L variant includes an LED indicator, but the datasheet does not describe it as a latched fault diagnostic or specify an electronic overcurrent-trip response. We retain those distinctions rather than equating an indicator or surge rating with our protection functions.

Inductive flyback: interrupting motor current requires a suitable recirculation or clamping path. We consider possible 80–120 V transients and use an onboard 30 A Schottky freewheel diode in our purpose-built controller. Actual transient voltage and diode loading depend on the circuit. Confirm internal protection or specify an appropriate external suppression network from the controller manufacturer's instructions.

A fuel pump must deliver the required flow against regulator and rail pressure; it cannot safely be calibrated solely like a fan or an LED dimmer. Excessive duty reduction can cause inadequate pressure or flow, with consequences for both the engine and pump cooling.

We use an 11.0 V effective-supply target for our 40–60 psi base-pressure examples. This is an example calibration floor, not a proven universal stall voltage for all listed pumps. PWM behaviour is not identical to simply applying a lower DC voltage; verify pressure, flow and current across the operating envelope.

The illustrative duty calculation gives approximately 79.7% at 13.8 V and 75.9% at 14.5 V. Account for wiring and controller voltage drops and validate the actual system rather than relying on these percentages alone.

• Pre-crank prime: we recommend a 2–3 second, 100% duty example to build rail pressure. Configure prime duration and pressure targets for the actual fuel system and ECU.
• High-side switching: our AI.SSR25 switches the positive supply. Follow our wiring diagram and provide a correctly sized return path; do not treat high-side switching alone as a guarantee against corrosion or every harness fault.
• Logic reference: Pin 4 is GND REF for the ECU ground reference. Verify the exact controller pinout and ECU ground instructions; do not connect motor return current through an ECU sensor-ground circuit.
• Sealed connections: we specify TE Connectivity Deutsch DTP 4-pin connectors with Size 12 contacts. Verify connector/contact ratings, wire size, sealing and mounting for the installation.
• Protection: we specify an autonomous analog eFuse with a response below 4 µs. Electronic protection does not remove the need to follow our upstream fuse and wiring requirements.

Our AI.SSR25 is a purpose-built high-side solid-state controller for suitable motorsport pumps and fans, with 25 A continuous capability and a 33 A latched protection threshold. Apply these specifications within our operating and installation conditions; a protection threshold is not a continuous-current rating.

• High-speed gate drive: we use a dedicated 4 A driver, with 35 ns / 40 ns transitions in our switching-loss examples. Maximum potential PWM frequency is 100 kHz; load-specific operating frequency remains a separate decision.
• Thermal performance: fast switching reduces dynamic loss, but total heat depends on current, duty, frequency, temperature, diode losses and cooling. A <1.2 W total-heat figure requires defined supporting test conditions and is not a universal rating.
• Flyback control: we use an onboard 30 A Schottky freewheel path and a 2 oz copper plane. Confirm the application limits and external suppression requirements in our installation documentation.
• Analog eFuse: we specify sub-4 µs protection with a latched red diagnostic LED for the specified overcurrent or fault conditions. Verify reset behaviour and fault coverage before commissioning.
• Static DC and PWM: we use a charge-pump path for continuous 100% on-state operation as well as ECU PWM control. Verify electrical compatibility before replacing a mechanical relay; connector and pinout differences still matter.
• Packaging: we specify a CNC 6061-T6 aluminium base/heatsink, an FKM perimeter gasket and TE Connectivity Deutsch DTP 4-pin interfaces.

A useful benchmark compares the complete pump system, not just relay switching loss. We have not published measured Walbro 460/525/535 frequency-versus-temperature results here. The process below defines the evidence needed before calling a frequency a preferred minimum or ideal. Live-fuel testing belongs in a properly equipped, fuel-safe facility with qualified personnel; do not run a pump dry or bypass its safety interlocks.

Record a continuous-DC reference for context, but do not compare its full-flow temperature directly with a lower-flow PWM point and call that an efficiency result. For each PWM comparison, keep delivered flow and pump pressure rise matched and adjust duty as necessary. Full-flow DC and partial-flow PWM are different hydraulic operating points.

• Identify the exact pump part number and motor architecture. Record fuel composition, supply voltage at the controller, inlet fuel temperature, tank level, plumbing, pressure, flow target, wiring and controller mounting/cooling. Hold these conditions consistent between comparable runs.
• Define permitted test points using the component manufacturers' limits and an approved validation plan. Candidate comparison points include 200/250/500 Hz, 1 kHz, 5 kHz and 10/15/20 kHz, but include only points allowed for the hardware. Never operate a comparison relay above its published frequency or load limits.
• At each frequency, adjust duty to achieve the same target flow and pressure, then repeat at relevant low-load and high-pressure operating points. Equal duty cycle does not necessarily produce equal hydraulic performance at different frequencies.
• Measure motor-current waveforms with suitable bandwidth and record average current, RMS current, peak-to-peak ripple, running peaks and startup/inrush peaks. Battery-side average current is not motor RMS current: the freewheel loop carries current during the off interval. Use appropriate isolated/differential voltage instrumentation and document probe bandwidth and sampling.
• Measure average battery-side input power from voltage and current, plus delivered flow and pump pressure rise. Hydraulic power is flow × pressure rise in consistent SI units. Input power minus hydraulic power represents combined pump/controller losses, not motor winding heat alone. Meter and plumbing losses must be accounted for in the measurement boundary.
• Log controller temperature, pump case temperature and inlet/return fuel temperatures until the defined steady-state condition is met, or report a matched timed interval. Case and fuel temperatures are not direct winding-temperature measurements. Stop before exceeding documented current, pressure or thermal limits, or if pressure/flow becomes unsafe or protection trips.
• Repeat each point at least three times with comparable starting temperatures, varying test order to reveal thermal carry-over. Report repeat variation and the limits of the instrumentation. Do not generalise one sample's result to every pump sold under the same flow name.
• Use one AI.SSR25 per pump. Calculate each pump's peak current draw and verify startup/inrush, continuous/transient and thermal suitability. A 33 A latched protection threshold is not operating-current headroom.
• Confirm the actual power-carrier waveform and ECU output's electrical interface, voltage limits and polarity. Treat 100 kHz as maximum potential capability, not a default setting. Check motor-rated channel suitability, derating and genuine continuous 100% duty.
• Verify inductive recirculation/clamping and diode dissipation from the manufacturer's instructions.
• Validate minimum duty using measured pressure, flow and current—not a universal assumed stall voltage. Configure and verify prime, engine-stall shutdown, ignition-off and command-loss behaviour, and verify pinout, wiring, fusing and connectors before commissioning.
• Publish measured results separately from datasheet ratings and calculated switching-only estimates. Include pump/controller model, test conditions, frequency, duty, flow, pressure, motor RMS/peak current, input power and temperatures. Select a setting only when performance, peak current and thermal margins are acceptable across the intended operating range.