Plot frequency data from a CSV on a log X axis

A frequency response from a swept-sine test, a spectrum saved by an analyser or the measured gain of a filter usually comes as a CSV with a frequency column and a magnitude column. Plotted as it is, the X axis often shows row numbers, or the first decades are squeezed into a sliver at the left. This guide shows how to put the frequency column on a log X axis in your browser, with an example file to try, what happens to zero and negative frequencies, why dB values stay on a linear Y axis, why Excel’s line chart has no log X option, and when a file needs an FFT, a PSD or an FRF first.

What a frequency-response CSV looks like

One row per frequency: a frequency column, usually in Hz, and one or more value columns, such as a magnitude in dB, a gain ratio or a phase in degrees. The frequencies are often spaced evenly on a log scale, the same number of rows in each decade, so the steps grow as the frequency rises. These are the first rows and the last row of our example file:

Frequency (Hz),Magnitude (dB)
1,0.01
1.023,-0.01
1.047,-0.01
1.072,0.00
…
1000,-36.68

The example is a made-up frequency response of an accelerometer on a weather mast at Aberdeen: 301 rows from 1 Hz to 1,000 Hz, 100 a decade, with the magnitude in dB. Download the example file (CSV, 3.44 KB) or plot it with the steps below.

Reading the example

On a linear X axis, the 200 rows below 100 Hz, two thirds of the file, share the left tenth of the chart, and the flat part and the rise to the resonance are squeezed against the axis. On a log X axis, 1 to 10 Hz, 10 to 100 Hz and 100 to 1,000 Hz each get a third of the width, and the shape reads at a glance:

Within 1 dB of 0 dB up to about 40 Hz, the response peaks at 18.39 dB at 120.2 Hz, the resonance of the accelerometer’s mounting, and falls about 40 dB a decade above it. The file was made from a second-order system with a resonance at 120 Hz and 6 % damping, plus a ripple of at most 0.05 dB, and SciPy’s freqs of the same system differs from every row by at most 0.05 dB.

Zero, DC and negative frequencies

A log axis has no place for 0 or for negative numbers: the logarithm of 0 is minus infinity. Many spectra start with a 0 Hz row, the DC value, and a two-sided FFT has negative frequencies too. With “Log scale, X” ticked, the plotter leaves out every row whose X is at or below 0, and a note in the chart’s corner counts them, for example “1 row with X at or below 0 is not shown on a log axis”. The rows are not deleted: they stay in the data table and in any CSV you download, and unticking “Log scale, X” draws every point again.

Don’t replace 0 with a tiny frequency such as 0.001 Hz to keep the row on the chart: the axis would then stretch over three empty decades before 1 Hz.

Magnitude in dB: keep the Y axis linear

A value in dB is already logarithmic: 20 × log10 of an amplitude ratio, or 10 × log10 of a power ratio. Plot it on an ordinary linear Y axis, the default, and a log X axis gives the familiar semi-log frequency plot. A log Y axis on top of dB would take the logarithm twice, and has no place for the negative dB of a response below its reference. If the column is a plain ratio instead, such as 0.5 or 8.3, tick “Log scale, left” in the same Axes tab for a log Y axis.

Why Excel’s line chart has no log X axis

In Excel, a line chart’s horizontal axis is a category axis. Microsoft’s documentation says a line chart distributes category data evenly along the horizontal axis, and when the data gives no categories they are made up as 1, 2, 3 and so on: the row numbers again. An XY (scatter) chart has two value axes instead, and Microsoft lists turning the horizontal axis into a logarithmic scale as a reason to choose one. So in Excel, chart the frequency and magnitude columns as a scatter chart, with lines if you like, and set its horizontal axis to a logarithmic scale. Microsoft’s pages, read on 23 October 2026, are listed under Sources.

A time signal needs an FFT, a PSD or an FRF first

If the CSV holds readings over time, a time column and an acceleration, pressure or voltage, it isn’t frequency data yet, and a time axis can’t be logarithmic: the plotter says so. Turn it into frequencies first:

• FFT from CSV: the amplitude of each frequency in a signal, for tones such as a 50 Hz hum.
• The PSD calculator: the power spectral density of random vibration, in g²/Hz from g, with the Grms.
• FRF and coherence: the response of a system, magnitude and phase, from its input and output signals.

Each result plot has Log X and Log Y buttons above it, and its CSV button saves the result, ready to plot on a log X axis here or anywhere else.