SciDAVis for Physics & Engineering: From Sensor Data to Fitted Models

Physics and engineering run on exported data: oscilloscope traces, DAQ logs, thermocouple readings, photodiode voltages, accelerometer streams. SciDAVis — free, open-source, latest release v2.9.2 (precompiled binaries on SourceForge are v2.8) — takes those exports from raw numbers to a publication-quality plot with a fitted physical model, with no license server and no per-seat cost. This guide walks the typical physics workflow: importing and organizing sensor data, putting honest error bars on measurements, finding hidden frequency content with FFT, and extracting material parameters with nonlinear fitting.

Version note: written for v2.9.x (SourceForge binaries are 2.8). The Analysis menu, Fit Wizard dialogs, and fitting engine are unchanged between the two.

Step 1 — Import and organize your data

Most scopes and DAQs export ASCII or CSV: one column of time, one column per channel, often with a header block of instrument metadata. In SciDAVis, File → Import ASCII handles the common layouts (separator, header lines, decimal character). The tricky cases — transposed files, multi-run logs, comment lines mixed with data — are worked through step by step in how to import data into SciDAVis.

After import, tell SciDAVis which column is which: select the columns, then Table → Set Column As → X/Y/Z assigns the roles (time as X, voltage as Y; plain-text comment columns can keep describing run conditions). A little discipline here — one table per run, units in the column comments — pays off when the project grows to dozens of runs and you still need to know which trace was which.

Step 2 — Error bars that mean something

Error bars separate a student plot from a publishable one. In physics they usually come from two places: repeated measurements (the standard deviation of the repeats) or instrument specifications (a scope’s vertical accuracy, a DAQ’s least significant bit). Plot the data first, then Graph → Add Error Bars… lets you attach X errors, Y errors, or both — either as a constant value or from a column of uncertainties. Use a column whenever the uncertainty varies point to point, for example with Poisson counting statistics.

Read the result honestly: error bars are how your data tells the model “this is how far you may miss me.” A model curve that threads every error bar is suspicious (overfit); one that misses systematically is wrong (underfit). The full mechanics of the dialog — constant versus column-based errors, styling — are covered in the error bars tutorial.

Step 3 — Find hidden frequencies with FFT

Many physics signals hide periodic content: 50/60 Hz mains hum on a sensor line, a mechanical resonance in a vibration measurement, the chopper frequency in an optical experiment. Select the curve in the graph and choose Analysis → FFT. SciDAVis computes the spectrum into a new plot, and the peaks mark the characteristic frequencies of your signal. Read off the frequencies, compare with what you expect (mains, rotation rate, drive frequency), and you have either identified your noise source or confirmed your signal.

Two honest caveats. First, the frequency resolution is set by the record length, not the sampling rate — a one-second trace gives 1 Hz bins, so record longer when you need to separate close lines. Second, SciDAVis 2.9.x offers no windowing functions (no Hann or Hamming apodization), so abrupt edges of the record leak a little power into neighboring bins. For long records this is negligible; if it matters, export the CSV and apply windows in Python/SciPy. The detailed frequency-analysis workflow is in the FFT tutorial.

Step 4 — Extract parameters with nonlinear fitting

Physics rarely gives you straight lines: cooling curves are exponential, damped oscillations decay, magnetization saturates. SciDAVis offers two fitting routes, both in the Analysis menu. Quick Fit covers the standard models in one click; the Fit Wizard (Analysis → Fit Wizard, shortcut Ctrl+Y) gives full control — custom formulas, initial guesses, and fit-range limits.

A typical physics workflow: fit the model to the data, then interrogate the residuals. Zoom the graph: residuals should scatter randomly around zero. A wave, a trend, or a funnel in the residuals is the strongest signal to change the model — a high R² with a systematic residual pattern still means the wrong model. The wizard lets you restrict the x-range to the region where the model actually applies; outliers and unrelated shoulders poison the result, so fit only the physics you trust.

A concrete example: a damped oscillation from a torsion pendulum. A custom formula combining an exponential envelope with a sinusoid lets you read the damping coefficient and the natural frequency from the same fit — both parameters of the model, both with standard errors from the fit log. When Quick Fit is not enough, the custom-formula route in the wizard is where real physics fits happen. The full method, including how to read the results log and judge a fit, is in the curve fitting tutorial.

Honest limits for physics users

Two things SciDAVis will not do. It has no automatic peak finder anywhere — in multi-peak workflows you always mark the peaks yourself. And 2.9.x has no hypothesis tests (no t-test, ANOVA, or normality test); for those, export the table as CSV and use Python/SciPy or R/JASP. What it does give you — scriptable, repeatable fitting through a documented Python API — is covered in the curve fitting tutorial, and it is enough to automate a batch of runs without leaving the project.