pvanal

Phase vocoder analysis utility

The pvanal utility performs phase vocoder analysis on sound files, creating FFT-based analysis files for time-frequency domain processing. This analysis enables time-stretching, pitch-shifting, and spectral transformations.

Syntax

pvanal [options] inputSoundfile outputFFTfile.pvx

Arguments

• inputSoundfile - Input audio file to analyze
• outputFFTfile.pvx - Output phase vocoder analysis file

Options

Basic parameters

Option	Description
-c <channel>	Channel to analyze (default: all channels)
-b <beginTime>	Start time in seconds
-d <duration>	Duration to analyze in seconds
-s <srate>	Override input sample rate

Analysis parameters

Option	Description
-n <frameSize>	FFT frame size in samples
-w <windowOverlap>	Number of overlapping windows (default: 4)
-h <hopSize>	Hop size between frames

Window type

Option	Description
-H	Use Hamming window (default is von Hann)
-K	Use Kaiser window
-B <beta>	Beta parameter for Kaiser window (default: 6.8)

Display options

Option	Description
-g	Enable graphical display
-G <latch>	Graphical display with update rate
-v	Verbose output
-V <txtFile>	Write verbose output to text file

Examples

Basic phase vocoder analysis

pvanal input.wav output.pvx

Analysis with specific FFT size

pvanal -n 2048 input.wav output.pvx

High-quality analysis with more overlap

pvanal -n 4096 -w 8 input.wav output.pvx

Analysis with Kaiser window

pvanal -K -B 8.0 input.wav output.pvx

Analyze specific channel and time range

pvanal -c 1 -b 5.0 -d 10.0 input.wav output.pvx

Stereo analysis with custom hop size

pvanal -h 256 stereo.wav stereo.pvx

Output file format

The .pvx file contains:

• Header: Sample rate, channels, frame size, overlap factor
• Frames: For each analysis frame:
  • Amplitude values for each frequency bin
  • Phase or frequency values for each bin

Frame size considerations

Choosing FFT size

• 512-1024: Good for percussive sounds, fast transients
• 2048: General purpose, balanced time/frequency resolution
• 4096-8192: Better for sustained sounds, pitch analysis
• 16384+: Very fine frequency resolution for long sustained tones

Overlap factor

• 2: Minimal overlap, faster analysis, coarser time resolution
• 4: Standard overlap (default), good balance
• 8: High overlap, smoother time evolution, more frames

Window types

Von Hann (default)

Good general-purpose window with:

• Moderate frequency resolution
• Low sidelobes
• Smooth amplitude envelope

Hamming

Better frequency resolution but:

• Higher sidelobes
• Less smooth amplitude envelope
• Good for tonal analysis

Kaiser

Parametric window with:

• Adjustable sidelobe rejection via beta parameter
• Higher beta = better sidelobe rejection but wider main lobe
• Beta 6.8 (default) approximates Blackman-Harris window

Using PVX files in Csound

Phase vocoder analysis files can be used with Csound opcodes:

• pvsfread - Read phase vocoder analysis file
• pvoc - Classic phase vocoder resynthesis
• vpvoc - Variable-speed phase vocoder
• pvinterp - Interpolate between two PV analysis files
• pvcross - Cross-synthesis between PV files

Technical notes

Memory usage

Memory requirements depend on:

• Duration of analysis
• Frame size
• Overlap factor
• Number of channels

Estimate: duration × samplerate × channels × (1 + overlap) / frameSize × frameSize × 8 bytes

Analysis speed

Larger frame sizes and higher overlap factors increase analysis time but provide better spectral resolution and smoother time evolution.

Related utilities

• atsa - ATS analysis for sinusoidal modeling
• hetro - Heterodyne filter analysis
• lpanal - Linear predictive analysis