A tour of everything EnginPlot does. Click a number on the screenshot to jump to that feature, or use the chapter list to browse.
RPC3, PLT/SimCreator, fstrm/Moog Stream, CSV and space-separated
text. Channel names and units are detected automatically, including
units written inline as Name [unit] or
Name (unit).
How: File → Open (which also lists recent files), or pass one or more paths on the command line.
The reader works out its own delimiter and decimal separator, so a
European export like 1,5;2,5 reads correctly. It
handles quoted fields, skips instrument metadata above the header,
and reads UTF-8, UTF-16 and BOM-prefixed files.
A cell it can’t read (N/A, a dash, a blank)
becomes a gap in that channel instead of failing the whole file.
A time column written as 2026-08-30T14:03:07.250,
08/30/2026 2:03:07 PM, 30.08.2026 14:03:07,
a date and time in two columns, or a bare time of day that wraps
past midnight all work. Day-first versus month-first is decided from
the whole column, not guessed row by row.
A progress bar with the file name and a Cancel button appears in the plot area while a file loads in the background. The rest of the window stays usable, so you can keep working with files that are already open.
How: press Esc or click Cancel to abandon a load.
Every file shows its channel count, samples, duration, time step, sample rate, FFT settings and recorded date. A Time row says what the time column did, and a Warnings row lists anything the reader had to work around, so a file that loads leniently never loads silently.
Each open file’s row in the tree has a close button. Closing removes its channels from the plot and leaves everything else selected.
Overlay every channel on one plot, stack them one per row, or group rows by engineering unit so each unit gets its own scale.
How: Layout → Overlaid, Stacked, or Stacked by Unit.
When channels span very different ranges, the large ones move to a second, right-hand axis so the small ones stay readable. There is nothing to set up.
Files that aren’t sampled on a clock (CAN bus traffic, network messages, variable-step solvers) plot each sample at its own timestamp. Bursts and idle gaps show up where they really are and for as long as they really lasted.
View time histories, power spectral density, or both together, with or without a histogram panel alongside.
Each segment’s mean is removed before the transform, so a channel riding an offset shows how it varies rather than a spike at 0 Hz taller than anything it actually does.
How: Plots → Time, PSD, Time / PSD, or Time / PSD / Histogram.
The same three channels as above, now in dB. One switch moves every spectrum between engineering units and decibels: PSD plots, the PSD pane of the combined views, CSD overlaid and the magnitude rows of the CSD matrix. Axis labels, the legend and the cursor readout all follow it.
Decibels are 10·log10 of the spectrum, not 20, because a PSD is already magnitude-squared. Empty bins stay at −∞ and land on the plot’s −Inf rail rather than being floored to an invented number.
How: Plots → Spectra in Decibels, or the dB button on the toolbar. The choice is remembered between runs.
Show cross-spectral density as a matrix of every channel pair, or overlaid on one plot. Phase can be unwrapped or wrapped to ±180°, 0 to 360°, ±360°, 0 to 720° and more.
How: Plots → CSD Matrix or CSD Overlaid; Wrap menu for the phase range.
On a frequency plot the preview shows the channel’s spectrum instead of its time trace, on the same engineering-unit or decibel axis as the main plot, and reads it for you in three columns.
Shape gives the peak’s frequency and level, in engineering units and dB, and the roll-off above it in dB per decade. Quality gives the noise floor, the signal bandwidth — the highest frequency still 10 dB clear of that floor — and the dynamic range between them. Calculation gives Nfft, the number of averages, the noise bandwidth, and the random error the averaging leaves on the result.
A channel too slow for the bin spacing to resolve — a vehicle speed, a steering angle — is flagged Not a vibration channel, with the share of its variation sitting below that resolution, rather than being handed a peak that means nothing.
In the CSD plots the preview pairs the channel with every plotted channel from its own file: its own spectrum first, then its cross spectrum with each of them, following the main plot’s decibel and phase-wrap settings. A table underneath gives, for each pair, where the cross spectrum peaks, its level and phase there, the coherence at that peak, and a band coherence weighted so the frequencies where the two channels share power are the ones that decide it.
How: point at a channel row on any PSD or CSD plot.
Per-channel histograms show how much of the record each channel spent in each band. For event data, each sample counts for the time it covered rather than counting once, so busy bursts don’t drown out long quiet stretches.
Unevenly sampled records are interpolated onto a uniform grid built from their own timestamps before the transform, with idle gaps left empty rather than filled in. The frequency axis reflects the rate the data really has.
Spectra, cross-spectra and histograms are computed on background threads as soon as a file loads, so switching views is usually instant. Cells still being computed say so, and panning, zooming and selecting keep working while they fill in.
Every channel gets a small preview plot in the file tree, so you can see the shape of the data before you select it.
A row of buttons above each file’s channels selects all of them, clears the selection, or picks every channel with a given unit in one click.
Point at a channel in the file tree and it is drawn on its own, in a large framed box over the middle of the plot area, with the main plot dimmed behind it. It follows the pointer down the list, so you can read through twenty channels without plotting any of them. The title bar says what a click will do — Click to add or Click to remove.
The preview always shows the whole record. When the main plot is zoomed in, the zoomed span is shaded on the preview, so it doubles as a map of where you are in the file.
Underneath, a Statistics panel gives the channel’s minimum, maximum and mode; its mean, standard deviation and RMS; and its crest factor, skewness and excess kurtosis. They are taken over the finite samples only, and the heading says so on a channel that has NaNs or infinities. A recording with its own timestamps is weighted by how long each sample was held, so a burst of messages does not outweigh a long steady stretch. A Notices panel beside it lists the channel’s data-quality warnings with their counts.
On a large file still being worked out, the figures show as “…” and fill in on their own — the channel you are pointing at is moved to the front of the queue.
How: point at a channel row. Nothing to turn on.
Values a Y-axis cannot show are not dropped and do not wreck the scale. Each plot reserves a +Inf line under its top edge, a −Inf line along the bottom and a NaN row just above it. Every affected channel marks the rail in its own colour, at the time it happened, so you can see which channel went bad and when while the finite data keeps the full height of the plot.
The rails are per axis, so a dual-axis plot annotates left and right separately, and they only report what is in the window you are looking at.
How: Plots → Show Non-Finite Values toggles it.
Every channel is checked as it loads, and what it fails is drawn as a letter beside its thumbnail: N for NaNs, ∞ for infinities, L for values too large to have come from a real transducer (above 1e12), S for subnormals — numbers so close to zero that they have lost precision — and a flag for a dead channel that never changes value.
Pointing at a channel gives the breakdown, in the preview’s Notices panel: how many samples each check caught and what fraction of the channel that is. The file row totals it up, so a file says “18 red” before you open a single channel.
Hover a series to highlight its trace, left-click to bring it to the front, right-click to remove it from the plot.
How: Plots → Show Legend toggles it.
Drag a rectangle to zoom, middle-drag to pan, or use the scroll wheel. Double-click a point to zoom straight to it: left-click zooms up to that point, right-click zooms in from it.
Move the pointer over a plot and it reads out the value under it on both axes, labelled in engineering units.
On a frequency plot the cursor also marks the harmonics: place it on a peak and a labelled line lands on every integer multiple of that frequency, so you can tell at a glance whether the rest of the spectrum is a harmonic family or something unrelated. Lines that would crowd together are left out rather than drawn as a smear.
Zoom, pan and redraw stay smooth at 150 million plotted points. A 600 MB, ten-channel fstrm recording loads in about half a second. Shown: ten channels of fifteen million samples each.
One self-contained, signed .exe of about 1.3 MB. No installer, no dependencies, no companion DLLs. Runs on Windows 10 and later.
It remembers its window position, size and splitter between runs
in a small settings file under %APPDATA%\EnginPlot.
All screenshots use synthetic data.
