hilbert — A Hilbert transformer.
asig -- input signal
ar1 -- sine output of asig
ar2 -- cosine output of asig
hilbert is an IIR filter based implementation of a broad-band 90 degree phase difference network. The input to hilbert is an audio signal, with a frequency range from 15 Hz to 15 kHz. The outputs of hilbert have an identical frequency response to the input (i.e. they sound the same), but the two outputs have a constant phase difference of 90 degrees, plus or minus some small amount of error, throughout the entire frequency range. The outputs are in quadrature.
hilbert is useful in the implementation of many digital signal processing techniques that require a signal in phase quadrature. ar1 corresponds to the cosine output of hilbert, while ar2 corresponds to the sine output. The two outputs have a constant phase difference throughout the audio range that corresponds to the phase relationship between cosine and sine waves.
Internally, hilbert is based on two parallel 6th-order allpass filters. Each allpass filter implements a phase lag that increases with frequency; the difference between the phase lags of the parallel allpass filters at any given point is approximately 90 degrees.
Unlike an FIR-based Hilbert transformer, the output of hilbert does not have a linear phase response. However, the IIR structure used in hilbert is far more efficient to compute, and the nonlinear phase response can be used in the creation of interesting audio effects, as in the second example below.
The first example implements frequency shifting, or single sideband amplitude modulation. Frequency shifting is similar to ring modulation, except the upper and lower sidebands are separated into individual outputs. By using only one of the outputs, the input signal can be "detuned," where the harmonic components of the signal are shifted out of harmonic alignment with each other, e.g. a signal with harmonics at 100, 200, 300, 400 and 500 Hz, shifted up by 50 Hz, will have harmonics at 150, 250, 350, 450, and 550 Hz.
Here is the first example of the hilbert opcode. It uses the file hilbert.csd, and drumsMlp.wav.
Example 441. Example of the hilbert opcode implementing frequency shifting.
See the sections Real-time Audio and Command Line Flags for more information on using command line flags.
<CsoundSynthesizer> <CsOptions> ; Select audio/midi flags here according to platform -odac ;;;realtime audio out ;-iadc ;;;uncomment -iadc if realtime audio input is needed too ; For Non-realtime ouput leave only the line below: ; -o hilbert.wav -W ;;; for file output any platform </CsOptions> <CsInstruments> sr = 44100 ksmps = 32 nchnls = 2 0dbfs = 1 instr 1 idur = p3 ; Initial amount of frequency shift. ; It can be positive or negative. ibegshift = p4 ; Final amount of frequency shift. ; It can be positive or negative. iendshift = p5 ; A simple envelope for determining the ; amount of frequency shift. kfreq linseg ibegshift, idur, iendshift ; Use the sound of your choice. ain diskin2 "drumsMlp.wav", 1, 0, 1 ; Phase quadrature output derived from input signal. areal, aimag hilbert ain ; Quadrature oscillator. asin oscili 1, kfreq, 1 acos oscili 1, kfreq, 1, .25 ; Use a trigonometric identity. ; See the references for further details. amod1 = areal * acos amod2 = aimag * asin ; Both sum and difference frequencies can be ; output at once. ; aupshift corresponds to the sum frequencies. aupshift = (amod1 - amod2) * 0.7 ; adownshift corresponds to the difference frequencies. adownshift = (amod1 + amod2) * 0.7 ; Notice that the adding of the two together is ; identical to the output of ring modulation. outs aupshift, aupshift endin </CsInstruments> <CsScore> ; Sine table for quadrature oscillator. f 1 0 16384 10 1 ; Starting with no shift, ending with all ; frequencies shifted up by 2000 Hz. i 1 0 6 0 2000 ; Starting with no shift, ending with all ; frequencies shifted down by 250 Hz. i 1 7 6 0 -250 e </CsScore> </CsoundSynthesizer>
The second example is a variation of the first, but with the output being fed back into the input. With very small shift amounts (i.e. between 0 and +-6 Hz), the result is a sound that has been described as a “barberpole phaser” or “Shepard tone phase shifter.” Several notches appear in the spectrum, and are constantly swept in the direction opposite that of the shift, producing a filtering effect that is reminiscent of Risset's “endless glissando”.
Here is the second example of the hilbert opcode. It uses the file hilbert_barberpole.csd.
Example 442. Example of the hilbert opcode sounding like a “barberpole phaser”.
See the sections Real-time Audio and Command Line Flags for more information on using command line flags.
<CsoundSynthesizer> <CsOptions> ; Select audio/midi flags here according to platform ; Audio out Audio in No messages -odac -iadc -d ;;;RT audio I/O ; For Non-realtime ouput leave only the line below: ; -o hilbert_barberpole.wav -W ;;; for file output any platform </CsOptions> <CsInstruments> ; Initialize the global variables. sr = 44100 ; kr must equal sr for the barberpole effect to work. kr = 44100 ksmps = 1 nchnls = 2 ; Instrument #1 instr 1 idur = p3 ibegshift = p4 iendshift = p5 ; sawtooth wave, not bandlimited asaw phasor 100 ; add offset to center phasor amplitude between -.5 and .5 asaw = asaw - .5 ; sawtooth wave, with amplitude of 10000 ain = asaw * 20000 ; The envelope of the frequency shift. kfreq linseg ibegshift, idur, iendshift ; Phase quadrature output derived from input signal. areal, aimag hilbert ain ; The quadrature oscillator. asin oscili 1, kfreq, 1 acos oscili 1, kfreq, 1, .25 ; Based on trignometric identities. amod1 = areal * acos amod2 = aimag * asin ; Calculate the up-shift and down-shift. aupshift = (amod1 + amod2) * 0.7 adownshift = (amod1 - amod2) * 0.7 ; Mix in the original signal to achieve the barberpole effect. amix1 = aupshift + ain amix2 = aupshift + ain ; Make sure the output doesn't get louder than the original signal. aout1 balance amix1, ain aout2 balance amix2, ain outs aout1, aout2 endin </CsInstruments> <CsScore> ; Table 1: A sine wave for the quadrature oscillator. f 1 0 16384 10 1 ; The score. ; p4 = frequency shifter, starting frequency. ; p5 = frequency shifter, ending frequency. i 1 0 6 -10 10 e </CsScore> </CsoundSynthesizer>
The use of phase-difference networks in frequency shifters was pioneered by Harald Bode1. Bode and Bob Moog provide an excellent description of the implementation and use of a frequency shifter in the analog realm in 2; this would be an excellent first source for those that wish to explore the possibilities of single sideband modulation. Bernie Hutchins provides more applications of the frequency shifter, as well as a detailed technical analysis3. A recent paper by Scott Wardle4 describes a digital implementation of a frequency shifter, as well as some unique applications.
H. Bode, "Solid State Audio Frequency Spectrum Shifter." AES Preprint No. 395 (1965).
H. Bode and R.A. Moog, "A High-Accuracy Frequency Shfiter for Professional Audio Applications." Journal of the Audio Engineering Society, July/August 1972, vol. 20, no. 6, p. 453.
B. Hutchins. Musical Engineer's Handbook (Ithaca, NY: Electronotes, 1975), ch. 6a.
S. Wardle, "A Hilbert-Transformer Frequency Shifter for Audio." Available online at http://www.iua.upf.es/dafx98/papers/.