I Have a Datalog and Don't Understand It: A Symptom-to-Channel Index

If you have a file full of numbers and no idea where to look, the problem is almost never the file. It is that you started from the data instead of the question.

As of September 2026 the fastest method is this: write down in one sentence what the car does wrong and when it does it, then open only the two or three channels that could possibly explain that sentence. Everything else in the log is noise for this particular investigation. A sixty-channel view is unreadable to a professional too — they just filter faster.

This article is that filter, written out. Find your symptom, read the channels it points to, and know what healthy looks like. If you want the mechanics of capturing a good log in the first place, reading your first wide-open-throttle pull covers the recording side, and datalog channels explained covers what each channel is.

The symptom index

Find the row that matches your complaint. Open those channels. Ignore the rest of the file for now.

The car does this	Open these channels first	What you are looking for
Rough or hunting idle	Fuel trims, commanded versus actual idle speed, idle airflow, spark advance	Trims away from zero at idle means fuelling, not idle control; oscillation with flat trims means the control loop
Hesitates or stumbles off idle	Throttle position, manifold pressure, short-term trim, spark	A momentary lean spike or a timing dip exactly at the tip-in point
Feels flat at full throttle	Commanded versus actual air-fuel ratio, spark advance, knock retard, intake air temperature	Retarded timing, a lean or overly rich mixture, or heat-soaked intake air
Pings or rattles under load	Knock retard, spark advance, intake air temperature, load	Retard events repeating at the same load and engine speed rather than scattered
Poor fuel economy	Long-term fuel trims across cruise, coolant temperature, closed-loop status	Persistent positive trims, or a thermostat keeping the engine cold enough to stay in warm-up enrichment
Hard starting when cold	Cranking speed, coolant temperature, injector pulse width during crank, fuel pressure if logged	Whether the calibration is delivering the cranking fuel it should
Hard starting when hot	Fuel pressure if logged, coolant and intake temperature, restart pulse width	Heat-related fuel behaviour rather than a calibration table
Shifts harshly or flares	Transmission line pressure, commanded gear, torque request, engine torque estimate	A torque signal that does not match what the engine is doing
Surges at steady cruise	Short-term trim oscillation, commanded air-fuel ratio, throttle position	A control loop chasing a fuelling error at low load
Check engine light after a tune	Freeze-frame data, the relevant sensor channels for that code, fuel trims	Whether the code is a real fault or a monitor upset by the calibration
Boost falls short of target	Commanded versus actual boost, wastegate duty, intake air temperature	Duty at its limit means a hardware ceiling, not a calibration one
Temperature climbs when towing	Coolant temperature, intake air temperature, spark advance, knock retard, transmission fluid temperature	Whether the calibration is protecting itself by pulling timing

Several of those rows have full articles behind them. Idle complaints are the subject of idle control tuning and, when a camshaft is involved, rough idle after a camshaft swap. Cold and hot start problems are covered in cold start and cranking tuning. Post-tune fault codes are sorted into real and cosmetic in check engine light after a tune.

The ten channels that answer most questions

If you are building a channel list from scratch, start here and add rather than subtract.

Engine speed. The x-axis of almost every diagnostic question. Everything else gets read against it.

Throttle position, or pedal position and throttle position together. On a drive-by-wire car these two differ, and the gap between them is itself information — the computer is deciding how much throttle your request deserves, which is the subject of throttle response tuning.

Manifold pressure or mass airflow, whichever your car uses. This is the air model's input. Which one your car depends on is the single most important structural fact about your calibration, and MAF versus speed density explains the difference.

Short-term and long-term fuel trims. The computer's own report card on its air model. Short-term is the immediate correction; long-term is what it has learned. Healthy is near zero and consistent. Ten percent or more means something is genuinely wrong, and fuel trims explained goes through what each direction means.

Commanded and actual air-fuel ratio. Commanded is what the calibration asked for; actual is what a sensor measured. The gap is the error. At light load the factory sensor is adequate; at full load it is not, which is what the wideband question is about.

Spark advance. How much timing the computer is actually commanding, which is often less than the table says because of corrections.

Knock retard. Timing being removed because the sensors heard something. The most safety-relevant single channel in a performance log.

Coolant temperature and intake air temperature. Both gate large numbers of corrections. A car that never reaches operating temperature behaves differently from one that is too hot, and intake air temperature is the channel that explains why the car felt strong in the morning and flat at noon — see heat soak and density altitude.

That set covers most of what an owner needs. Add transmission channels for shift complaints and boost channels on a forced-induction car, where boost control basics explains what duty and target should look like relative to each other.

How to actually read a trace

Three habits separate people who get answers from people who stare at graphs.

Read the relationship, not the value. A single number in isolation means almost nothing. Knock retard of two degrees is meaningless until you know it happened at full load on a hot day at the top of third gear. Always read a channel against engine speed, load and temperature.

Find the moment, then zoom. Scroll to the few seconds containing the symptom and expand them. Most diagnostic answers live in a two-second window. Looking at a five-minute overview is how people conclude a log is unreadable.

Compare commanded against actual, everywhere you can. Modern calibrations log both sides of many quantities — commanded and actual air-fuel ratio, requested and delivered torque, desired and actual idle speed, target and actual boost. The gap between a pair is a far better diagnostic than either value alone, because it shows you where the calibration's intent and the engine's reality diverge.

Capturing a log that is worth reading

A large number of "I can't tell anything from this log" problems are recording problems.

• Include the symptom. If the complaint is at full throttle, the log needs a real full-throttle pull, taken safely and legally on a road where that is possible, or on a dyno.
• Keep it short. Two to five minutes containing the event beats twenty minutes of commuting.
• Log the right rate. Too few channels at a high rate is more useful than every channel at a crawl, because a fast transient disappears between slow samples.
• Note the conditions. Ambient temperature, fuel in the tank, what gear, whether the engine was fully warm. Write it down at the time; you will not remember.
• Start cold if the complaint is cold. You cannot diagnose a cold-start fault from a log that begins after warm-up.

That last group — the description that travels with the log — is what separates a useful submission from a frustrating one when someone else is reading it. What to send your tuner sets out the full package.

When the log says the problem is not the tune

This is worth stating because it is the most common genuinely useful outcome of reading a log properly.

Fuel trims heavily positive at idle and improving with load is a classic unmetered-air signature — a leak downstream of the airflow sensor, not a calibration error. A misfire counter concentrated on one cylinder is an ignition or mechanical issue, not a table. Fuel pressure collapsing at high injector duty cycle is a supply limit, covered in injector duty cycle and fuel pump headroom. Knock retard that appears only after ten minutes of driving is a heat problem.

None of those are fixed by editing tables, and all of them will be made worse by a calibration that compensates for them. That is precisely why the pre-tune mechanical checklist exists — a tune written on top of a fault becomes a tune that depends on the fault.

Getting a second opinion on a log

Reading your own logs is a skill that builds quickly once you know which channels answer which question, and the index above is most of that knowledge. What takes longer is the pattern recognition to notice the thing you were not looking for.

That is the gap TuneVault fills. It reads your HP Tuners VCM Editor tables, audits them, and verifies them against the log you captured — so instead of "is this number bad", you get an ordered list of what to change and confirmation that the change was safe. A guided tune including a datalog verification pass is $39, or from $19 a month if you are working through a build. The pricing page has the detail, and you can upload a log at the datalog tool.

Either way, the method stays the same: start from the symptom, open three channels, read commanded against actual. A datalog is only overwhelming when you ask it everything at once.

Frequently asked questions

I have a datalog but do not know what I am looking at. Where do I start?

Start from the symptom, not the file. Write down in one sentence what the car does wrong and when. Then open only the two or three channels that relate to that symptom and ignore everything else. A datalog with sixty channels open is unreadable to anyone; the same log filtered to three relevant traces usually answers the question in a couple of minutes.

What are the most important channels to log on any car?

Engine speed, throttle or pedal position, manifold pressure or mass airflow, short and long-term fuel trims, commanded and actual air-fuel ratio, spark advance, knock retard, coolant temperature and intake air temperature. Those ten cover the large majority of diagnostic questions. Add transmission channels if the complaint involves shifting, and boost channels on a forced-induction car.

How long should a datalog be?

Long enough to contain the symptom and short enough to find it. Two to five minutes that includes the exact condition you are investigating beats twenty minutes of general driving. If the complaint happens at wide-open throttle, the log needs a genuine full-throttle pull in it; if it happens at idle when cold, the log needs to start before the engine warms up.

What does a normal long-term fuel trim look like?

Close to zero, typically within a few percent either side, and reasonably consistent across idle, cruise and light load. Trims sitting at ten percent or more mean the computer is working hard to correct an air-model error or a mechanical fault. Positive trims mean it is adding fuel because the mixture read lean; negative means it is removing fuel.

Why does my log show knock retard when the engine sounds fine?

Knock sensors listen for a frequency band, and several things can excite that band that are not detonation — a loose heat shield, an injector tick, a transmission noise. Detonation that shows up repeatably at the same load and engine speed is real. Scattered single-count retard at random points, especially at low load, is usually noise and worth ruling out mechanically before you change any timing.

Do I need a wideband sensor to read a useful datalog?

Not for most part-throttle diagnosis, because fuel trims tell you what you need there. You do need one to see what is genuinely happening at wide-open throttle, because the factory narrowband sensor cannot resolve the rich mixtures used under full load. If the complaint lives at full throttle, a wideband is the difference between a measurement and a guess.

Can someone else analyse my datalog for me?

Yes, and this is a normal part of both professional e-tuning and guided DIY tuning. What makes the difference is the quality of what you send — a log that contains the symptom, taken with the right channels, along with a description of the car and the complaint. A vague log with no context wastes everyone's time regardless of who is reading it.