Skip to main content
Track Fast Fourier Transform
12 / 16
Search the archiveLibrary
Library

2002issue C071-4

From the power spectrum to indicator windows

Editorial reading: treat a fast Fourier transform as a construction bench for cycle-aware lookbacks. Detrend, taper, and pad first, then convert the tallest spectral peak into an explicit indicator window.

  • Fourier analysis reconstructs a complex series as sinusoids of amplitude, frequency, and phase, and spectral analysis uses that same construction to identify dominant frequencies in ordered observations.
  • A fast Fourier transform converts time-ordered observations into a power spectrum so a cycle length can be read from frequency and reused as an indicator parameter.
  • Valid construction removes any trend longer than the sample, applies a Hanning window, and pads the array to a power of two such as 64, 128, or 256.
  • Once a cycle length is measured, a moving-average cutoff can be set at half that length so shorter fluctuations are treated as noise.
Entries in this reading3 entries
Fourier analysis14 readingsCurrent trackFast Fourier Transform17 readingsSpectral analysis24 readings

The transform as a construction bench

The editorial reading is that a fast Fourier transform is a construction bench for cycle-aware lookbacks, not a display that already contains a completed signal.

Fourier analysis reconstructs a complex series as a sum of sinusoids, each defined by amplitude, frequency, and phase. Spectral analysis is treated as the same construction as Fourier analysis and is used to identify dominant frequencies or frequency bands in ordered observations.

A fast Fourier transform converts time-ordered observations into a power spectrum whose horizontal axis is frequency and whose vertical axis is cycle power. The power spectrum is a frequency-versus-power plot produced after superimposed cycles have been separated.

Detrend, taper, and pad

Valid construction requires removing any trend longer than the sample, forcing similar start and end values with a Hanning window, and padding the array to a power of two such as 64, 128, or 256.

Detrending removes a component longer than the sample, often by subtracting a linear fit or a moving average, before the transform is run. The Hanning window is a preprocessing taper that forces similar start and end values so endpoint jumps do not distort the transform.

After the Hanning window is applied, peak power remains at the same frequencies even though extra energy appears beside those peaks.

Price series complicate the construction because they trend, need not be periodic, and contain weekend and holiday gaps that can hide parts of a cycle.

Read cycle length from frequency

With a 260-session year, period equals 260 divided by frequency, so a frequency of 20 cycles per year maps to a 13-session period.

In a 64-point summed-wave example, one cycle per window corresponds to a 64-period length and five cycles per window corresponds to a 12.8-period length.

A displayed price series resolved into three ranked cycles of 341, 64, and 102 sessions, with the strongest component shown first.

Convert the peak into a cutoff length

Once a cycle length is read from the transform, a moving-average cutoff can be set at half that length so shorter fluctuations are treated as noise.

The editorial construction reuses that cutoff length as the indicator window, so the measured cycle becomes a lookback instead of remaining only a peak on the power spectrum.

Ford FFT cycles and the half-period indicator windows

On Ford Motor Co., the article names three MetaStock Fourier cycles of 341, 64 and 102 days, strongest at left and weakest at right. The same section turns each peak into a moving-average cutoff by halving the period, so the working lookbacks are 170.5, 32 and 51 days. Cycle lengths come from the prose beside Figure 1; the windows apply that stated cutoff rule.
On Ford Motor Co., the article names three MetaStock Fourier cycles of 341, 64 and 102 days, strongest at left and weakest at right. The same section turns each peak into a moving-average cutoff by halving the period, so the working lookbacks are 170.5, 32 and 51 days. Cycle lengths come from the prose beside Figure 1; the windows apply that stated cutoff rule.F · Daily · 2001-05-01T00:00:00.000Z to 2002-05-31T00:00:00.000Z

Cutoff length equals cycle length divided by two, the article's rule for a noise-rejecting moving average. A 341-day cycle is longer than one year of daily bars; the sidebar warns that a cycle longer than the sample can skew the spectrum unless the series is detrended first.

Educational research material, not investment advice. Historical source context does not establish present-day performance.
12 of 16 in the Fast Fourier Transform track
20031-7 pp.Next on Fast Fourier TransformEndpoint Fast Fourier Transform evaluation with walk-forward mechanical rulesA noise-filtered Fourier curve fitted to a completed price window can appear to lead a major high, while the same transform computed only through that high can point the opposite way.
All readings on this track · 16 readings
  1. 1982Building FFT spectra to size cycle filters
  2. 1988Fourier cycle models break in major swings
  3. 1988Constructing moving average filters from price Fast Fourier Transforms
  4. 1989Staging Fast Fourier construction under memory limits
  5. 1993Constructing forecast inputs with moving averages, Fourier transforms and intermarket spreads
  6. 1994Preprocessing prices so Fourier peaks set moving-average lengths
  7. 1994Constructing a spreadsheet FFT power spectrum from daily prices
  8. 1994Building dominant-cycle spectra with FFT preprocessing
  9. 1994Constructing labeled cycle lengths from FFT spectra
  10. 1999Fast Fourier Transform reconstruction is not a walk-forward decision tool
  11. 1999Walk-forward endpoint Fourier construction as a same-day mechanical procedure
  12. 2002From the power spectrum to indicator windows
  13. 2003Endpoint Fast Fourier Transform evaluation with walk-forward mechanical rules
  14. 2004Constructing signal and noise from market waveforms
  15. 2012A two-stage case study in market cycle analysis
  16. 2015Whitening pink noise to build a near-zero-lag cycle oscillator
By fieldCycles and spectral analysis77 readingsBy year200284 readings
All 17 readings tagged Fast Fourier Transform
Also on Fast Fourier Transform5 readings
1982Building FFT spectra to size cycle filtersFourier analysis1994Constructing a spreadsheet FFT power spectrum from daily pricesFourier analysis1994Constructing labeled cycle lengths from FFT spectraFourier analysis2012A two-stage case study in market cycle analysisFourier analysis1988Constructing moving average filters from price Fast Fourier TransformsFourier analysis