Why a driver output is judged only under load: the no-load pass trap
A module can switch a perfect 12 volts with nothing connected and collapse the moment it has to carry current. Learn why the no-load pass is the trap, what a transistor driver actually is, and how an output is proved in the condition it has to work in.
What a module driver output actually is
Inside a control module, the output stage is a transistor - a solid-state switch - that carries power (or ground) to the load: a fuel pump driver feeding the pump, an injector driver feeding the injectors, a relay driver pulling a relay coil. The transistor is rated to carry current, and that is the entire job. The module's logic decides WHEN the switch closes; the silicon decides whether it can carry the load once it does. Those are two different pieces of hardware doing two different jobs, and they fail independently of each other.
Two separate jobs: the command and the delivery
Every module-driven output has a command side and a delivery side, and they must be proved separately:
The command side - the module's logic decides ON or OFF and drives the transistor's control terminal. Scoping the command signal proves the module is asking. It proves nothing about whether the transistor answers.
The delivery side - the transistor itself, the supply that feeds it, the ground return it works against, and the wiring between the output pin and the load. This is the side that carries the amps, and it is the side a no-load test never exercises.
A perfect command signal with a dead delivery side is the classic no-load trap: the module is doing its job and the part doing the real work is not. Blame the module for a delivery failure and you replace the one piece that was working.
Why a no-load pass lies
With no load connected, the output carries essentially nothing - only the microamps the meter itself draws. At that current, almost any transistor can hold voltage, healthy or half-dead. Resistance in the delivery path - a tired transistor junction, a corroded pin, a thin bond wire, a lifting ground - is invisible because voltage lost across resistance is current times resistance (V = IR), and the current is nearly zero. Multiply any resistance by zero amps and the loss is zero volts. The reading looks perfect because the test asked nothing of the part.
The moment the real load connects, current becomes real, and the same resistance drops voltage exactly in proportion to that current. An output that read a clean 12 volts no-load can read 7 volts loaded - not because something changed, but because the test finally asked the question. The no-load reading was never a measurement of the driver. It was a measurement of the meter.
Voltage drop under load: where the proof lives
The number that matters is the output voltage WHILE the output carries its load's current - measured at the output pin, then measured again at the load's terminals. Two comparisons carry the whole judgment:
No-load vs loaded at the pin - this is the trap detector. A healthy driver shows nearly the same voltage loaded as no-load. A multi-volt collapse means the delivery side cannot carry current, whatever the command side is doing.
Loaded at the pin vs loaded at the load terminals - this separates the driver from the harness. If the driver holds at its pin but the load starves at its terminals, the loss is in the wiring between them - the driver was innocent.
Current must be demonstrated, not assumed. A clamp reading during the loaded test proves the load was actually drawing - an output measured "loaded" with a load that never pulled current is just another no-load test wearing a disguise. The measurement discipline behind loaded voltage readings is covered in Voltage drop testing on power feeds; the crank-loading companion is Voltage drop under crank.
Failure modes that only show under load
These are the ways a driver that passes no-load dies when current flows:
Thermal foldback in a weak transistor - the transistor starts the test fine, heats as current flows, and its resistance climbs with temperature. The output sags seconds into the test, not instantly. A one-second reading passes; a ten-second reading fails.
Current-limit protection doing its job - many drivers fold back or shut down when current exceeds a threshold. A load drawing more than the driver can carry - a dragging pump, a partially shorted coil - looks like a dead driver but is really a protection circuit protecting itself. The fix is at the load, not the module.
A high-resistance ground return - on a low-side driver the transistor switches the ground path. Corrosion in the return lets the ground reference lift under load; the output floats up and the load sees less voltage across it. No-load: perfect. Loaded: the ground moves.
Connector pins that make cold and lose under heat - a spread pin touches when cold and backs off as current heats the joint. The output passes the first test and fails the third, and every reading in between is a lie about the connection.
The load changed, not the driver - a pump that drew 8 amps new and draws 16 dying can collapse a driver that was always adequate for the healthy load. Voltage without current is half the story; prove the current against what the load is supposed to draw.
The loaded test: how an output is proved
Proving a driver output follows the order of the circuit, and nothing is skipped:
Confirm the command arrives - before any output is judged, prove the module is actually asking. A driver that was never commanded is not a failed driver.
Prove supply and ground at the module under load - a starving module cannot drive anything. Supply voltage at the module's supply pin and ground rise at its ground pin, both measured while the load draws.
Read the output no-load, then loaded, with current demonstrated - the trap detector. Clamp the current during the loaded reading so the test is known to be real.
Read at the load terminals under load - proves the harness delivered what the driver gave. A driver proved at its pin is not proved at the pump.
Walk what fails, then re-prove the repair loaded - a collapse is walked segment by segment - supply, ground, driver, harness, load - until the lossy piece is found, and the fix is re-proved under load before the case closes.
How SpeedNeeds uses it
A driver output is proved the way it works: commanded, carrying its real current, measured at the load. The comparison that decides is no-load versus loaded with the current shown - a collapse under load is the trap caught, not a surprise. Supply and ground at the module are proved as the driver's support legs before the driver itself is judged, so a starving module is never blamed on the transistor; and the load terminals get their own reading so a harness loss is never blamed on the driver. The measurement discipline behind that comparison is covered in the companion explainers (Voltage drop testing on power feeds and Voltage drop under crank).
Caution and edge cases
Backprobe the connector - never pierce the wire to measure. A pierced output wire corrodes from the inside and becomes the next load-sensitive fault.
Some drivers are current-limited or PWM-controlled by design. A PWM output read with a plain DMM shows an average that means nothing - use a scope. Service information decides what the output is supposed to look like, not habit.
A driver that shuts down on overcurrent is protecting itself. Find why the current is high before condemning the module - the protection is evidence about the load.
Thermal failures need time. A driver that holds for one second and folds at ten is only caught by a test that lasts. Let the loaded reading run long enough to get warm.
An output that holds on the day the fault is hiding proves only that the fault is hiding. Note the conditions - temperature, run time, load state - and measure again when the complaint is active.
Confirm the commanded state before touching anything. On some vehicles the output stays powered after key-off; a surprise live pin is a safety problem, not just a bad reading.
Service information overrides every general rule here - driver type, expected current, command logic, and protection behavior are vehicle-specific.
Closing clarity
The command arriving is not the driver delivering. A module can ask perfectly and still fail to carry the current - and the only reading that catches it is taken with the load connected, the current demonstrated, and the meter on the output. The no-load pass is the trap. The loaded delivery is the proof. Judge the driver in the condition it has to work in, walk what collapses, and re-prove the repair the same way - loaded, not promised.
Stay in the loop
Get new drops in your inbox
New QTS builds, tools, and garage stories — straight to your inbox. No spam. Unsubscribe anytime.