Analysis
F1 telemetry data explained: what the car actually logs
What an F1 car measures, how fast it samples, what reaches the pit wall live, the rule against sending data back, and how engineers read a lap trace.
By CricketTaken EditorialPublished Analysis18 min read
A driver crosses the timing line, and before the car has reached the pit entry the race engineer already knows which corner cost the lap. Not roughly, and not from watching. A line on a screen turned upwards at a specific distance from the start of the lap, the throttle trace beneath it showed a hesitation of about a tenth of a second, and the brake trace showed a release that came off the pedal earlier than the reference. Nobody had to ask.
Understanding F1 telemetry data means understanding three separate things that get muddled together: what the car measures, how much of that measurement reaches a human being while the car is still moving, and what anyone is legally allowed to do about it. A modern Formula 1 car is comprehensively instrumented and records far more than it can transmit. A small fraction of the recording arrives live at the pit wall. And absolutely none of it may travel in the other direction, which is a regulatory decision that shapes how the whole sport is engineered.
What a Formula 1 car measures
Start with the physical instruments, because everything else is built on them.
The power unit is the most heavily monitored assembly on the car. Pressures and temperatures are taken at every point in the fluid systems: oil, water, fuel, air before and after the turbocharger compressor, exhaust gas at the turbine. Rotational speeds are measured on the crankshaft and on the turbo shaft. The energy recovery system carries its own instrumentation for battery cell voltages, cell temperatures, state of charge, and the current flowing into and out of both electrical machines. The mechanical and electrical complexity described in how a Formula 1 power unit works is the reason so much of the channel budget goes here.
The chassis contributes a second large group. Each corner of the car carries suspension position measurement, which is how the team knows the ride height and platform attitude at every instant. Accelerometers sit on the chassis and often on individual components. Brake temperatures are taken at the discs and in the ducts. Wheel speeds come from each hub. Tyre pressures and temperatures are transmitted from inside the wheel assemblies.
Then there are the driver's own inputs, which are among the most valuable channels on the car and among the simplest to measure: throttle pedal position, brake pedal position and pressure, steering angle, gear selected, clutch paddle position, and every switch and rotary the driver touches on the wheel.
Finally there are the channels that are not really measurements at all. A great deal of what appears on an engineer's screen is calculated in real time by the control electronics from several raw inputs at once. Estimated fuel remaining, estimated tyre load at each corner, energy deployment state, and the delta to a reference lap are all derived quantities. They are as useful as any sensor and rather more useful than most, and they exist because somebody wrote a model that turns raw signals into a number a person can act on.
Sample rates, and why nothing is logged at one speed
A channel is defined by what it measures and by how often it measures it, and the second half is where the engineering judgement lives.
Sampling costs storage, bandwidth and processing, so every channel is assigned the lowest rate that still captures the behaviour of interest. The governing principle is the shape of the signal rather than its importance. A quantity that changes slowly can be sampled slowly without losing anything. A quantity that changes faster than the sample rate does not merely lose detail, it produces misleading values, because a sampled signal that moves faster than the sampling can misrepresent itself entirely.
That produces a rough hierarchy. Structural and vibration measurements sit at the top, because the events of interest last milliseconds. Suspension movement and wheel speed sit high, because a wheel locking or a damper crashing through its travel is a fast event. Driver inputs sit in the middle, fast enough to catch a pedal application precisely but nowhere near the top. Fluid pressures sit below that. Bulk temperatures sit at the bottom, because an oil temperature simply cannot change quickly enough to justify anything else.
The numbers above are invented and the spread between them is the real lesson. There are two or three orders of magnitude between the fastest and slowest channels on a car, which is why total data volume is dominated by a small number of high-rate channels while the number of distinct channels is dominated by slow ones.
Rates are also not fixed for a weekend. Teams log differently in practice than in the race, because practice is where diagnostic detail is worth the storage and the race is where a reliable, compact record matters more. Aerodynamic testing runs, with pressure rakes and additional instrumentation fitted, produce configurations that would be unnecessary and impractical to carry on a Sunday.
Two paths: the live link and the download
There are two entirely separate routes by which data leaves the car, and confusing them is the most common misunderstanding about how a team actually works.
The first is the live telemetry link. The car transmits over a dedicated wireless connection to receiving equipment in the garage whenever it is within range of the circuit's infrastructure. This feed is what appears on the screens on the pit wall in real time, and it is what allows an engineer to see a problem while the car is still on the circuit. It is also heavily constrained. Radio bandwidth is finite and shared, so only a selected subset of channels is transmitted, several of them at reduced rates compared with what is being recorded onboard. The live feed is a summary.
The second is the onboard log. Everything, at full rate, is written to storage on the car itself, and it stays there until the car stops. When it returns to the garage the full recording is pulled off, either through a physical connection or over a short-range high-bandwidth link, and only then does the team have the complete picture of what happened. That download is why the garage goes quiet and busy at the same time when a car comes in, and why the genuinely detailed analysis of a practice run begins after the run rather than during it.
The practical consequence shapes a race weekend. Decisions made while the car is out are made on partial information. Decisions made between runs are made on complete information. A good race engineer is fluent in both modes and knows which one they are currently operating in, and a great deal of what looks like caution on the radio is somebody declining to act on a subset until they have seen the whole.
- The sensor produces a signalA physical quantity becomes a voltage, a frequency or a digital message on a bus. Calibration data converts that raw output into engineering units, and a wrong calibration produces a plausible and completely false number.
- The control electronics sample itThe mandated control unit reads the channel at its assigned rate, timestamps it against a common clock, and makes it available both to onboard control functions and to the logging system.
- It is written to the onboard logEvery channel at full rate is stored on the car. This recording is complete and nobody sees any of it until the car has stopped moving.
- A subset goes to the radio linkA configured selection of channels, some at reduced rate, is packaged and broadcast to the garage. The choice of what to include was made before the session and cannot be changed mid-run.
- Garage systems receive and decode itReceivers rebuild the stream, apply calibrations, and compute derived channels the car did not send. The result populates live displays within a fraction of a second of the event happening.
- Engineers and models read it in parallelHumans watch a handful of screens while automated monitors watch every channel at once against expected behaviour, because no person can supervise several hundred signals in real time.
- Something becomes a messageAlmost nothing that arrives is acted upon. What survives the filtering becomes a single sentence on the radio to the driver, because the driver's attention is the scarcest resource in the entire system.
- The full log arrives afterwardsWhen the car returns, the complete recording is downloaded and the reconstruction of what actually happened begins, with detail the live feed never carried.
The path a single channel takes through a Formula 1 team's data system during a session. Each stage adds latency or removes detail, which is why the same channel means different things at different points in the chain.
The last step is the one worth dwelling on. An enormous data operation exists in order to occasionally produce one short instruction to one person. Every stage in between is compression, and the skill being exercised is discarding almost everything.
The standard control unit, and who is allowed to see what
Since the late two thousands, every car has run a single mandated electronic control unit rather than a team-designed one, supplied under an FIA contract by one approved supplier. That decision was taken partly to control cost and partly to control what the electronics were permitted to do, because the previous arrangement made policing driver aids close to impossible.
The standard unit does not make every team's software identical. It provides a common hardware platform, a common set of interfaces, a common logging framework and a common timebase, and teams write their own application layer within it. Two cars on the same grid can behave very differently while running the same box, which is the intended outcome: standardise the thing that needed policing and leave the rest competitive.
The unit is also the reason the governing body has access to what it needs. Certain data are available to the FIA for compliance purposes, which is how technical regulations about engine modes, energy deployment limits and driver aids are enforced at all. A rule that cannot be inspected is not a rule, and control electronics are what makes inspection possible without dismantling a car.
Team access itself is not unlimited either. The Sporting Regulations cap the number of operational personnel a team may have at the circuit, which pushed a large part of every team's analysis capability into remote operations rooms at the factory. Live telemetry arrives at both places at once. Engineers thousands of miles from the circuit watch the same channels as the pit wall, run comparisons the trackside team has no time for, and feed conclusions back. The organisational shape this produces is common to serious teams across sport, and it is described more generally in the piece on how a modern analytics department is structured.
The rule that shapes everything: nothing goes back to the car
Data leaves the car. Nothing of consequence comes the other way, and this is a deliberate regulatory choice rather than a technical limitation.
The sport experimented with two-way telemetry in the early two thousands, allowing the garage to send settings and adjustments to a running car, and then prohibited it. The reasoning was straightforward. A car whose behaviour can be reconfigured remotely by a room full of engineers is a car being driven by more than one person, and the sport decided it wanted a competition between drivers operating machines rather than a competition between control rooms operating machines with drivers aboard.
The consequence is visible in every broadcast and most people miss what they are looking at. When a pit wall wants a change made, someone has to say it out loud, on a radio channel that is recorded, frequently broadcast, and available to rival teams. The driver then has to find the correct control on the steering wheel and make the change themselves, often at speed and in the middle of a battle. Every setting a team might want to alter has to be reachable by a driver's thumb, which is why the wheels carry so many rotaries and buttons and why drivers spend a large part of their preparation memorising positions rather than settings.
There is exactly one meaningful exception, and it runs on the official system rather than the team's. The marshalling system transmits flag status and safety car information from race control to a display in every cockpit. That is data being sent to a car, and it is permitted because it is a safety function operated by the officials and identical for everyone.
The rule also explains a pattern that puzzles new viewers. A team can see, live, that a component is failing, and can do nothing about it except tell the driver to change a mode, lift and coast, or bring the car in. Knowledge arrives instantly. Intervention has to travel through a human being with a steering wheel.
Where the reference lap comes from
Every delta needs something to be a delta against, and choosing that reference is a decision with consequences.
The obvious candidate is the driver's own best lap of the session so far, which is what the in-car delta display usually shows and what a driver uses to judge whether a push lap is working. It answers a narrow question well: am I ahead of my own best right now.
Engineers frequently want a different reference. Comparing against a team mate's best answers a question about driving. Comparing against a simulated lap generated before the weekend answers a question about whether the car is doing what the model said it would, and the places where the real lap and the simulated one diverge are the places where the team's understanding of its own car is wrong. That is often the most valuable comparison of the weekend, because it points at the simulation rather than at the driver.
Circuit position has to be measured accurately for any of this to work, and it comes from a combination of satellite positioning, wheel speed integrated over time, and correction against fixed timing loops buried in the track surface. Each of the three has a weakness the others cover, which is why teams use all of them rather than trusting the most obvious one.
Reading a lap trace: where the time actually went
Ask an engineer how they find time in a lap and they will describe opening a small number of channels, always the same ones, plotted against distance rather than against time.
Distance is the choice that makes the whole method work. Plot two laps against elapsed time and they drift apart immediately, so the same corner appears at different places on the two traces and comparison becomes impossible. Plot them against distance around the circuit and every corner sits at the same horizontal position on both, no matter how much quicker one lap was. Every overlay in Formula 1 works this way.
The core set is small. Speed shows what the car did. Throttle and brake show what the driver asked for. Steering angle shows how much the car needed. Gear and engine speed show how the power unit was being used. And the delta channel, the running time difference against a reference lap, shows the consequence.
- Delta to reference lap
- hundredths
Invented delta time in hundredths of a second against a reference lap, plotted at successive points around a fictional sequence. Constructed to show how the slope of the line is read, not data from any lap. Positive numbers mean time lost relative to the reference.
Show the numbers
| Item | Delta to reference lap | hundredths |
|---|---|---|
| Braking board, turn 1 | 0 | |
| Turn 1 apex | 4 | |
| Turn 1 exit | 9 | |
| Back straight | 11 | |
| Turn 2 entry | 11 | |
| Turn 2 apex | 6 | |
| Turn 2 exit | -3 | |
| Turn 3 entry | -3 | |
| Turn 3 exit | -8 |
The chart above is invented, and the way to read it is the real technique. The value of the line matters much less than its gradient. Where the line climbs, the lap is losing time at that instant. Where it is flat, the two laps are equivalent regardless of how far apart they already are. Where it falls, this lap is quicker there.
Applied to the figure, the story reads itself. Time is lost from the braking point of the first corner through to its exit, which points at entry rather than at the corner itself. The back straight continues to lose a little, which is the exit deficit still being paid off in speed. Then the second corner is taken better and the third better still, and by the end of the sequence the lap is ahead.
That distinction between where a gap exists and where a gap is created is the single most useful idea in trace reading, and it is where most amateur analysis goes wrong. A driver losing four tenths on the pit straight is almost never losing it on the pit straight.
What each channel tells you about a corner
The traces are read as a set, because each one answers a question the others cannot.
The brake trace shows where the pedal went down, how hard, and how the pressure was released. Engineers look at three separate features: the point of initial application, the peak, and the shape of the release. A driver who releases the brake in one motion is behaving differently from one who trails it into the apex, and the second is usually the faster technique in a car that needs front load on entry.
The throttle trace shows the point of first application and the shape of the pick-up. A tentative, stepped application is a driver managing a rear end they do not trust. A single clean sweep is a driver who knows the car will take it. The moment of first throttle is often the highest-value data point in an entire corner, because a small gain there is multiplied along the whole straight that follows.
The steering trace shows how much lock the corner required and whether the driver had to add any. Extra lock mid-corner is understeer. A rapid correction is the rear stepping out. Comparing the steering traces of two laps often explains a speed difference that the pedal traces alone leave mysterious.
The speed trace is the outcome, and it is read for three things: the minimum speed at the apex, the speed at the point of first throttle, and the speed at the end of the following straight. The last of these is the one that matters most, because it captures everything the corner did to the lap.
The ride height and suspension traces show what the platform was doing, which is where trace reading meets aerodynamics. A car bottoming through a compression, or riding higher than the setup intended through a fast corner, is not producing the downforce the driver was promised, and the reasons for that are the subject of how a modern car generates its aerodynamic load.
Driver comparison overlays, and the traps in them
Overlaying two drivers is the most common use of telemetry inside a team and the easiest way to reach a wrong conclusion.
The technique is straightforward. Put both laps on the same distance axis, plot the delta between them, and work backwards from wherever the delta moves to find the channel that explains it. Done well, it produces specific, actionable coaching: brake five metres later here, carry more speed to the apex there, get to full throttle earlier at this exit.
Done carelessly, it produces nonsense, and there are four standard traps.
The first is conditions. Two laps set at different points in a session were driven on a circuit with different grip, different air density and possibly a different wind direction, and none of those appear in the traces as such. A lap set at the end of a long run on a rubbering-in circuit is not comparable to one set an hour earlier.
The second is the car itself. Fuel load changes lap time substantially and changes the aerodynamic platform as well. Tyre age changes grip. Engine and energy modes change straight-line speed. An overlay between two laps run in different states is comparing setups, not drivers.
The third is setup divergence. Team mates rarely run identical cars, and a driver who has asked for a stiffer rear or a different differential setting is driving a different machine. A trace difference may be showing you two drivers adapting correctly to two different cars.
The fourth is the composite lap fallacy. Adding up the best sector of every corner across a session produces a theoretical lap that nobody drove and often nobody could drive, because the fastest approach to one corner sometimes forecloses the fastest approach to the next. Teams compute it anyway, as a boundary rather than a target.
Handled properly, the overlay is still the most powerful tool a driver coach has, because it removes the argument. A driver who believes they braked as late as their team mate can be shown the two brake traces, and the discussion ends.
Reliability monitoring, and predicting a failure before it happens
The most valuable thing telemetry does during a race has nothing to do with lap time.
A failing component almost never fails without warning. It drifts. A pressure that used to recover in a certain number of milliseconds now takes slightly longer. A temperature that used to plateau now keeps climbing at a shallow rate. A vibration signature that used to have its energy at one frequency starts spreading it across others. None of these cross a red line, and any one of them could be noise.
Watching for them is therefore not a matter of alarm thresholds. It is a matter of comparison against expected behaviour. The team knows what each channel does at each point on this circuit, at this fuel load, at this temperature, because it has run the model and has the history. What matters is the residual, the gap between what the channel is doing and what the model says it should be doing, and a residual that grows lap on lap is the signature of something failing regardless of whether the absolute value is anywhere near a limit.
- An automated monitor flags a residualSoftware watching every channel notices one behaving differently from the model's expectation. At this stage nobody has been told anything, because most flags are noise.
- The flag persists across lapsA single anomalous lap is discarded. A trend that repeats and grows over several laps is treated as real, and a systems engineer is alerted to look at it directly.
- A specialist inspects the channel in contextThe engineer responsible for that system checks correlated channels. A rising temperature accompanied by a falling pressure in the same circuit means something different from a rising temperature alone.
- The factory is consultedRemote operations compare the trace against the same component on the same car at previous events, against the other car, and against dyno and rig history for that part.
- A consequence is estimatedThe question becomes what happens if nothing is done, how many laps that leaves, and what changing a setting would buy. A part that will survive to the flag is left alone.
- Options are costed against the raceTurning down a mode, lifting and coasting, or bringing the car in each has a time price. Retirement has the largest price of all, which is why teams tolerate a limping car more often than viewers expect.
- The driver is told one thingThe output is a single instruction, phrased so it can be acted on immediately without explanation. The reasoning stays in the garage.
The escalation path from an automated anomaly detection to an instruction to the driver. Most detections stop at the second or third stage, which is the system working as intended rather than failing.
The economics behind that final judgement are severe, because power unit components are subject to a strict allocation across a season and exceeding it brings grid penalties. Saving a damaged component by retiring a car may be worth more than the points on offer that afternoon, and that calculation lives alongside the tyre and pit stop reasoning described in how a Formula 1 race strategy is built. Reliability monitoring is a strategic function as much as an engineering one.
There is a second, quieter use. The same drift detection running against the full log after a session builds a picture of how components age, which feeds the models that decide how long a part may safely be run at all. A team that understands its own wear curves can run parts closer to their limits than a team that does not, and that is a competitive advantage assembled entirely out of old data.
Data the car does not send
Not everything a team uses arrives over the telemetry link, and the surrounding sources matter more than they appear.
An accident data recorder is mandated and logs the severe-event channels the medical and safety investigators need, at rates far beyond anything used for performance work. Onboard high-speed cameras record the driver and the cockpit environment. Biometric sensors in the drivers' gloves give the medical team physiological information immediately after an incident, which is designed to inform the response to a crash rather than the running of a race.
Around the car sit weather stations reporting track and air temperature, humidity, pressure and wind, and their readings are as much a part of the data set as anything on the vehicle, because they set the conditions everything else must be interpreted against. Timing and position data come from the official system rather than the team. Tyre data arrives partly from the car and partly from equipment in the garage that measures the tyres before and after they are used.
The output of all of this is not really a set of numbers. It is a model of the car that runs alongside the car, updated continuously, and used to answer questions the sensors cannot answer directly. Nobody measures tyre grip. Nobody measures downforce during a race. Those are estimated by feeding what is measured into a model that has been calibrated against everything the team has ever recorded, and the quality of that model is one of the largest and least visible differences between a good team and a mediocre one.
What to watch for on a broadcast
Telemetry is far more visible on television than most viewers realise, once you know what the signs are.
Listen to how instructions are phrased. A concise, specific radio message with a setting name and a position is a team acting on live data. A vague message asking how the car feels is a team without a channel that answers the question.
Notice when a team asks the driver to change something mid-corner-sequence. That is a decision that had to be made before the next lap, and it means the live feed showed something that would not wait for the download.
Watch for lift and coast. A driver being asked to lift early on straights is almost always managing a temperature or a fuel number that somebody is watching climb in real time.
Watch the graphics comparing two drivers. Broadcast overlays showing throttle and brake side by side are a simplified version of exactly what the engineers are looking at, and the moment where the two traces diverge is where the lap time is being decided.
Treat a retirement as a story that started earlier. A car stopping on lap forty usually appeared on somebody's screen as a residual on lap twenty, and the team spent those twenty laps deciding whether to spend a component or a result.
Watch the garage when a car comes in. The concentrated activity around the laptops in the first minute after a car stops is the full-rate download landing, and it is the point at which the team finally learns what the run actually did.
The habit worth building is to stop thinking of telemetry as surveillance and start thinking of it as compression. Several hundred channels, sampled thousands of times a second, exist so that one person can be told one useful thing at the right moment. More writing on the engineering and rule-making behind the sport is gathered in the motorsport section, and the wider set of explainers across every sport we cover sits on the blog index.
Common questions
What data does an F1 car record during a session?
Several hundred channels covering the power unit, gearbox, hydraulics, brakes, suspension, tyres, aerodynamics and the driver's own inputs, along with position and timing. Some are direct sensor readings such as a pressure or a temperature, and many others are values the control electronics calculate from several sensors at once. The full set is logged onboard at high rate, while a reduced selection is transmitted live.
Do F1 teams get live telemetry during a race?
Yes, in one direction. The car broadcasts a selected set of channels to receivers in the garage over a dedicated wireless link whenever it is within range, and engineers on the pit wall and at the factory watch it in real time. The link is bandwidth limited, so live telemetry is a compressed subset of what the car is actually recording.
Can a team send data or settings to the car?
No. The regulations prohibit telemetry transmitted to the car, so any change to how the car is running has to be made by the driver using the steering wheel controls after being told what to do over the radio. The one permitted transmission is the official marshalling system, which sends flag and safety car information to a display in the cockpit.
What is the delta channel that engineers look at first?
It is the running time difference between the lap being driven and a chosen reference lap, plotted against distance around the circuit rather than against time. Its value at the end of the lap is the lap time gap, and the slope of the line at any point shows exactly where that gap was created or given back.
How do teams compare two drivers' laps?
By overlaying the traces of both laps against distance and reading the delta between them. Speed shows where the difference appeared, throttle and brake traces show what each driver did to cause it, and steering and gear traces show how the car responded. The comparison is only fair if fuel load, tyre age, engine mode and track conditions were similar, which is why raw overlays between sessions are treated with suspicion.
How does telemetry predict a mechanical failure?
By watching for a channel drifting away from the behaviour a model expects rather than by waiting for it to cross a red line. A pressure that recovers more slowly than usual, a temperature climbing at the wrong rate or a vibration signature changing shape are all signs that something has started to fail, and they usually appear laps before the component would actually let go.
Filed under Motorsport·formula 1 · telemetry · data analysis · race engineering · technology