# Analysis software essentials: plots, overlays, XY, histograms, filtering, run notes

> Most analysis is comparison. Know when to plot against time or distance, how to read XY plots and histograms, how filtering changes what you see, and how to keep run notes so data stays tied to a setup.

Area: Data Acquisition & Analysis · Starting level: L1 Foundation · Updated 2026-11-16
Source page: https://race-engineer-academy.vercel.app/lesson/daq-analysis-software-essentials

## L1 · Foundation

**Pick the display for the question.** A logger produces long lists of numbers; the software turns them into pictures.

- **Distance plots** answer *where*: braking point, apex location, where time is lost. Use them for overlays, because a given track location stays at the same distance from lap to lap.
- **Time plots** answer *how long* and *how fast it changes*: shift duration, brake build-up rate, damper response. Overlays against time drift apart as soon as lap times differ.
- **Track maps** show where things happen; colour the map by a channel (gear, braking, damper direction) for quick orientation.
- **XY (scatter) plots** show how two channels relate: lateral vs longitudinal g (the g-g "traction circle"), oil pressure vs lateral g, RPM vs speed (gear chart), front vs rear roll angle.
- **Histograms** show distribution: RPM, throttle position, speed, damper velocity.
- **Run charts** plot one statistic per lap across a run, a day or a season: average understeer angle, fuel per lap, full-throttle %.

**Overlays are the core tool.** Segers puts it bluntly: around 95 % of analysis is comparison, and software that can't overlay laps against distance isn't worth having. Always start with the **speed trace** because every setup or driving change aims to change speed. Then use the time-difference channel to find *where* the gain or loss happens, and only then look at *why*.

**Keep notes with the data.** Every file must link to a setup, fuel load, tyre set, track temperature and driver comment. Put the file name on the setup sheet and use the software's session-constant feature (static ride heights, motion ratios, spring rates, track widths) so math channels update when the setup changes. A file you can't link to a setup is noise.

## L2 · Practitioner

**Templates and workflow.** Build templates before the event, never trackside. Typical pages:

| Page | Channels |
|---|---|
| Vitals | RPM, oil/water temp, oil/fuel pressure, battery voltage, gearbox temp |
| Driver | speed, throttle, brake pressure (front+rear), steering, gear |
| Balance | speed, steering, lat g, understeer angle, yaw rate |
| Braking | speed, long g, brake pressures, brake balance, wheel speeds |
| Platform | speed, damper positions, ride heights, pitch/roll |
| Fuel/engine | fuel used, fuel per lap, lambda, MAP, inlet temp |

Use consistent channel colours (e.g., each wheel always the same colour) and scale axes so the variation you care about fills the graph.

**Filtering.** A moving-average filter replaces each sample by the mean of its neighbours. Its effective time window is

$$t_{\text{filter}} = \frac{n}{f_s}$$

with $n$ the number of samples in the window and $f_s$ the sampling rate. Eight samples at 50 Hz is 0.16 s. That is enough to reveal slow chassis motion in a damper trace but it also removes kerb strikes and shrinks peaks. Rules:

- Filter as little as possible; show the filtered trace on top of the raw one in a lighter colour.
- Use **raw channels inside math channels**, then filter the result if needed. Pre-filtering inputs distorts any non-linear calculation (damper force maps, products, ratios) and misaligns inputs filtered with different windows; for a purely linear channel the order would not matter.
- Every filter delays or smears; never compare event timing across a filtered and an unfiltered channel.

**Getting organised.** Download every run, back it up immediately, name files consistently (event_session_run_car_driver), and export statistics tables to a spreadsheet for run charts. Set the pit-box analysis station where it does not obstruct the mechanics.

## L3 · Advanced

**Choosing the right abstraction.** Expert analysts move between three levels quickly:

1. **Run level (run charts, statistics tables):** is anything trending? Fuel per lap rising, average understeer angle drifting, full-throttle % dropping with tyre age. Run charts expose patterns that single-lap overlays hide – e.g., Segers shows three drivers sharing a car where two trend toward oversteer over a stint and the third toward understeer.
2. **Lap level (overlays, delta time):** where is the difference?
3. **Event level (zoomed time plots, XY of a corner):** why?

**XY plots as diagnostic tools.** A relationship you expect to be fixed becomes a fault detector when it isn't: oil pressure that drops only in right-handers at high lateral g points to scavenging or oil level; RPM vs wheel speed off the gear lines indicates wheelspin or clutch slip; brake pressure vs long g scatter shows tyre/ABS saturation.

**Histogram pitfalls.** Histograms weight by *time* (or samples). A damper histogram from a slow lap has more samples on straights; compare as percentage of time, and only between runs logged at the same rate. For damper velocity use a symmetric bin structure and separate low (< 25–50 mm/s) and high speed zones.

**Filter design beyond moving averages.** A centred moving average has zero phase lag in post-processing but a sinc-shaped frequency response with poor stop-band; a Butterworth applied forward-backward (zero phase) is cleaner for spectral work. Real-time filters in the logger (for dash or control) always add lag – roughly half the window length for a moving average.

**Statistics need context.** A minimum oil pressure of 0 bar in a statistics table may simply be the car stopped in the pit lane. Always cross-check summary numbers with the trace; gate statistics by condition (engine above 3000 rpm, car on a flying lap).

## Key formulas

**Moving-average filter window**

$$t_{\text{filter}} = \frac{n}{f_s}$$

- n: number of samples averaged
- f_s: sampling rate (Hz)
- t_{\text{filter}}: effective filter time (s)

## Trackside takeaways

- Overlay against distance, not time; start with speed, find the loss with delta time, then find the cause.
- Time plots for durations and rates; distance plots for locations.
- Filter as little as possible, show raw underneath, and use raw channels inside math channels.
- Run charts of per-lap statistics reveal trends single-lap overlays hide.
- Every data file must be linked to a setup, tyre set, fuel load and conditions – set this up before the event.

## By car category

### GT3

Pro-Am line-ups make driver-vs-driver overlays a daily job; build a dedicated template with brake pressure, throttle, steer and ABS/TC activity.

### LMP2

Spec Gibson engine and largely common data packages make cross-team benchmarking of overlays possible where data sharing rules allow.

### One-make Cup

Cup series often run identical cars, so overlays with a faster team-mate are the single most powerful tool available.

### Formula 4

With few channels, the speed overlay and delta time do 80 % of the work; drivers should learn to read them themselves.

### Super Formula

Short practice sessions put a premium on pre-built templates and fast run-to-run comparison.

## References

- Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE 2014): Ch. 2 Data Analysis Software Requirements (§2.2–2.8)
- Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE 2014): §2.6 Filtering
- Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE 2014): §3.1 Check the Car’s Vital Signs (oil pressure vs lateral g scatter)
