Actual acquisition and FFT settings required

FFT Bin Spacing and Record Resolution

Separate the displayed FFT grid from information supplied by the acquired time record. Zero-padding creates a finer interpolated grid; it does not create additional frequency-resolving information.

Δf_grid=f_s/N_{Rychlá Fourierova transformace (FFT)}Δf_record=f_s/M=1/TNo bearing diagnosis

FFT grid and record-limited quantities

Displayed bin-grid spacing, Δf_grid

Record-limited spacing, Δf_record

DFT record period, T=M/f_s

First-to-last timestamp span

Nyquistova frekvence

One-sided bins (real input)

Highest one-sided bin

Window equivalent noise bandwidth

Approx. null-to-null main-lobe width

Zero-padding ratio

Bin spacing is not the same as resolving two tones

For sample rate f_s, M acquired samples and an FFT of length N_{Rychlá Fourierova transformace (FFT)}, the displayed grid spacing is Δf_grid=f_s/N_{Rychlá Fourierova transformace (FFT)}. The acquired DFT record period is T=M/f_s, so its information spacing is Δf_record=1/T=f_s/M. A real-input one-sided FFT contains floor(N_{Rychlá Fourierova transformace (FFT)}/2)+1 bins including DC; the Nyquist frequency f_s/2 is itself a bin only for even N_{Rychlá Fourierova transformace (FFT)}.

If N_{Rychlá Fourierova transformace (FFT)}>M,zeros add grid samples between the unpadded DFT samples. NIST describes this as frequency-domain interpolation: the padded transform contains the same information as the unpadded transform. Therefore the page reports both spacings and never calls the padded grid an improvement in physical resolution.

Window bandwidth

The rectangular-window equivalent noise bandwidth is one record bin. For a single periodic Hann window this calculator uses ENBW=1.5Δf_record,as tabulated by NIST. Nominal null-to-null main-lobe widths are 2Δf_record for rectangular and 4Δf_record for Hann. These bandwidth scales do not by themselves decide whether two real components are resolvable: amplitude ratio,phase,noise,leakage,averaging and the chosen estimator also matter.

A published NIST measurement example used 801 lines from0 to400Hz with0.5Hz spacing,a2.0s record and a Hann effective bandwidth of0.75Hz. Inputs f_s=800Hz,M=N_{Rychlá Fourierova transformace (FFT)}=1600,Hann reproduce those quantities here.

ISO context

ISO13373-2:2016,edition2 is Published and at stage90.20 (systematic review). Its public scope covers processing,presentation and time/frequency-domain analysis of machinery vibration data. This general calculator is not an implementation or conformity check of the licensed standard,and no clause/table number is asserted.

Not calculated: anti-alias filter passband or stopband,analyzer-specific Fmax/“lines” conventions,overlap,averaging,amplitude/coherent-gain correction,PSD scaling,confidence interval,order tracking,zoom FFT,demodulation,sideband or bearing-fault diagnosis. A universal sampling factor such as2.56 and generic BPFO/BPFI/BSF/FTF multipliers are intentionally not inferred.

Zdroje:NIST Section10—DFT,record spacing,windowing and Hann ENBW;NIST FFT for Experimentalists PartI—zero padding;NIST published0.5Hz/Hann example;NI bin-spacing/record-length overview;ISO catalogue record linked above. Review:12July2026.