Plugging a scanner into a car takes about forty seconds and produces a list of codes. That list is genuinely useful — it is the car telling you which of its own tests it failed. What it is not, and was never designed to be, is a list of the parts you need to buy.
What is a fault code, exactly?
It is the record of a test the car ran on itself and failed. Which tests a car must run, and what counts as failing them, is not left to the carmaker’s imagination: it is written down. The document that decides it for most of the world’s petrol cars is Title 13 of the California Code of Regulations, section 1968.2. Everything quoted below comes from it, with its own subsection numbers, so that you can go and check.
Start with the section that covers ordinary sensors and actuators, (e)(15.2.1)(A): “The OBD II system shall detect malfunctions of input components caused by a lack of circuit continuity, out of range values, and, where feasible, rationality faults.” Read it slowly. Circuit continuity. Out of range values. Rationality. Not one of those three is a component. They are three ways an electrical signal can be wrong, and every one of them can be produced by the sensor, by the wire, by the connector, by a chafed loom, by a bad earth, or by the control unit at the other end.
For the things the computer commands rather than reads, (e)(15.2.2)(A) is just as revealing: “The OBD II system shall detect a malfunction of an output component/system when proper functional response of the component and system to computer commands does not occur.” The car asked for something and did not get it. That is the whole finding. Whether it did not get it because the part is dead, because the command never arrived, or because something mechanical is holding the part where it is, the regulation does not say — because the car cannot know.
The regulation that invented the fault code was careful never to promise it names a part. Only the people selling the part promise that.
So what does P0420 actually mean?
It means the exhaust is not being cleaned as well as it should be. That is all it means. The requirement, at (e)(1.1), is one line: “The OBD II system shall monitor the catalyst system for proper conversion capability.” The failure threshold, at (e)(1.2.1), is an emissions number, not a component condition — for one class of vehicle the system must flag a malfunction when conversion capability drops far enough that hydrocarbon emissions “exceed 1.75 times the FTP full useful life standards to which the vehicle has been certified”, or when average conversion efficiency “falls below 50” per cent.
Now ask what could put a catalyst below that line. A converter that has genuinely died of age. A converter poisoned by an engine that has been burning oil or coolant for a year. An exhaust leak upstream, feeding the rear sensor air it should not be getting. A lazy rear oxygen sensor reporting a conversion that is actually happening. A misfire dumping raw fuel into a converter that was healthy last month and is not now. Every one of those produces P0420. Only one of them is fixed by a converter, and two of them will destroy the new converter as well.
And a “system too lean” code?
It means the engine ran out of room to correct itself. The fuel system monitor exists, per (e)(6.1.1), to “monitor the fuel delivery system to determine its ability to provide compliance with emission standards.” One of its malfunction criteria, (e)(6.2.2), is exactly this: “if the vehicle is equipped with adaptive feedback control, the OBD II system shall detect a malfunction when the adaptive feedback control has used up all of the adjustment allowed by the manufacturer.”
That is a beautiful piece of drafting, because it describes the finding without a single guess about the cause. The computer has been quietly adding fuel to keep the mixture right, it has added all it is permitted to add, and it is still not right. Unmetered air entering after the mass air flow sensor does that. A weak fuel pump does that. A partially blocked filter does that. A drifting air flow sensor does that — it is not lying about the mixture, it is lying about the air, and the computer believes it. The code cannot distinguish between them, and it does not claim to.
Is that true of every monitor?
It is the pattern throughout. Take the cooling system monitor, which is the one most likely to speak to you in this country. Its requirement, at (e)(10.1.1), is to “monitor the thermostat on vehicles so-equipped for proper operation”, and one of its malfunction criteria, (e)(10.2.1)(A)(ii), is that a fault is detected if “the coolant temperature does not reach a warmed-up temperature within 20 degrees Fahrenheit of the manufacturer’s nominal thermostat regulating temperature.” The car has not found a broken thermostat. It has found an engine that did not get hot enough, fast enough — which a stuck thermostat causes, and so does the wrong thermostat, and so does a coolant temperature sensor reading low, and so does a short school run in February.
| What the car recorded | What the regulation says it means | What can produce it |
|---|---|---|
| A catalyst efficiency code | Conversion capability fell past an emissions threshold. | A dead converter, a poisoned converter, an exhaust leak, a lazy rear sensor, or a misfire. |
| A fuel system code | Adaptive fuel control “has used up all of the adjustment allowed by the manufacturer”. | Unmetered air, a weak pump, a restricted filter, or an air flow sensor reading wrong. |
| A sensor circuit code | “Lack of circuit continuity, out of range values, or rationality faults”. | The sensor, the wiring, the connector, the earth, or the control unit input. |
| A thermostat code | The coolant did not warm up to within 20°F of the nominal regulating temperature. | A stuck thermostat, the wrong thermostat, a low-reading sensor, or a drive too short to warm the engine. |
What else is in the car besides the codes?
A good deal, and this is the part a forty-second scan skips. Three things in particular change a diagnosis.
- Freeze frame
- The regulation requires it at (d)(2.2.7): “A manufacturer shall store and erase freeze frame conditions present at the time a malfunction is detected.” It is the engine’s readings at the instant the fault was recorded — so it can tell you the car was hot, loaded and at speed rather than cold on a driveway, which is often the difference between two candidate causes.
- Pending versus confirmed
- A test can fail once without lighting the lamp. A pending code is a fault the car has noticed and the driver has not, and it is the earliest thing on this list — it is also, on a car you are thinking of buying, the most interesting.
- Readiness monitors
- Each monitor reports whether it has run to completion since memory was last cleared. A car whose monitors are all sitting at “not ready” has had its memory cleared recently — which on a car with no codes is not the same thing as a car with no faults.
Why is clearing a code not a repair?
Because the car decides when a fault is gone, and it takes its time about it. Under (d)(2.3), once the lamp is on it “may be extinguished after three subsequent sequential driving cycles during which the monitoring system responsible for illuminating the MIL functions and the previously detected malfunction is no longer present.” Under (d)(2.4), the stored code itself can only be erased by the car “if the identified malfunction has not been again detected in at least 40 engine warm-up cycles”.
Which means the quiet dashboard after a clear tells you nothing for several days of ordinary driving. It also means the reverse: a lamp that came back is not a second fault, it is the same fault, confirmed by the car a second time. And on cars from roughly 2010 there is a category of code the scan tool cannot touch at all — we set that out in the article on what a pre-purchase inspection catches, because that is where it matters most.
What will a scan never see?
Anything the regulation did not ask the car to watch. The scope, at (e)(15.1.1), is components that “can affect emissions during any reasonable in-use driving condition” or that are “used as part of the diagnostic strategy for any other monitored system or component.” CARB’s own public fact sheet says the same thing in plainer words — the system is there to “monitor virtually every component that can affect emission performance”. That is a wide net over the engine and a very narrow one over everything else.
- A noise. Bearings, bushes, drop links, heat shields and engine mounts announce themselves acoustically and store nothing.
- Play. A worn ball joint, a tired steering rack and a loose wheel bearing are all measured by hand, on a lift.
- A leak. Oil, coolant, power steering and gearbox fluid leave evidence on the floor and under the car, not in memory.
- Wear. Brake pad and disc thickness, tyre condition and belt condition are inspected, not interrogated.
- A fault that has not met its own threshold yet. A converter at 55 per cent efficiency is failing and is not, by the regulation’s definition, a malfunction.
So what does a real diagnosis look like?
- Read everything, not just the code list: stored codes, pending codes, freeze frame, readiness monitors and live data, before anything is cleared.
- Reproduce the conditions in the freeze frame. A fault recorded hot and loaded is not a fault you find at idle on a cold engine.
- Test the circuit the code names, at the connector — voltage, ground and signal — rather than replacing the part hanging off the end of it.
- Separate the symptom from the system. A lean code is a measurement of the fuel system. The cause is usually somewhere else, and it usually has a pressure, a vacuum or a resistance that can be measured.
- Fix the cause, then let the car re-run its own monitors and agree with you. The car’s agreement is the proof, and it takes driving cycles, not a clear.