Suppressing spectral edge effects in Schroeder’s Harmonic Complex Preprint: Author, JASA-EL
Alessandro Altoè · 2026
Read Suppressing spectral edge effects in Schroeder’s Harmonic Complex Preprint: Author, JASA-EL by Alessandro Altoè (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-23847.txt.
Introduction
Altoè redesigns a classic laboratory sound so that two faint, unwanted steady tones no longer contaminate hearing experiments, in a 2026 paper on psychoacoustics. Schroeder's harmonic complexes are periodic signals that sound like a tone gliding up or down in frequency. Hearing scientists use them to study frequency dispersion in the ear. The catch is that the components at the edges of the signal's band do not share that glide. They sit at fixed frequencies and are easily audible.
Those edge tones can quietly distort behavioral results, since listeners might respond to them rather than to the glide. Altoè presents a variation that largely suppresses the steady components while keeping the time varying frequency character. The signal stays periodic and band limited, so existing experiments can adopt it directly.
Catalog record
| CREATOR | Alessandro Altoè |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 14 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.23847
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Suppressing spectral edge effects in Schroeder’s Harmonic Complex Preprint: Author, JASA-EL Alessandro Altoè Abstract Schroeder’s harmonic complexes are periodic, band-limited signals that are analogous to tones whose frequency increases or decreases over time. As such, they have been widely employed to study phenomena related to frequency-dispersion and frequency-modulation sensitivity in the auditory system. However, Schroeder’s complexes also embed two steady “frequency-fixed” components. Because these components are easily audible, they may complicate interpretation of behavioral experiments. Here I present a variation of Schroeder’s harmonic complex that largely suppresses these undesired steady components. I Introduction In his seminal paper, schroeder1970 Manfred R. Schroeder described how to synthesize periodic, band-limited signals with approximately flat envelope (low crest factor). These signals commonly called Schroeder’s harmonic complexes, can be thought of as tones whose frequency periodically increases (or decreases) from f_{1} to f_{2} over time. As such, they have been broadly employed in hearing research, especially to study the effects of frequency-dispersion (e.g., frequency-dependent delay) in the auditory system. In particular, they have been broadly employed in behavioral studies ( kohlrausch1995 ; carlyon1997 ; summers1998 ; lentz2001 ; oxenhamdau2001 ; henry2023 ) and, to a lesser extent, in physiological studies. summers2003 ; recio2001schroeder ; steenken2022 ; henry2023 A largely overlooked issue with Schroeder’s harmonic complexes, that makes them less than ideal to study frequency-dispersion, is that the signal components at the edge of their frequency band (e.g., at f_{1} and f_{2} ) do not have well-defined group delays and hence they are not well localized in time. As we will see below, Schroeder’s harmonic complexes appear as the superposition of a tone whose frequency changes over time, with two fainter, steady tones. Note1 These steady signal components are audible for certain combinations of stimulus parameters, potentially playing an unrecognized (or undesirable) role for the interpretation of behavioral experiments that make use of these stimuli. Note2 Schroeder’s complexes are closely related to frequency modulated (FM) sweeps, also widely used in auditory research, sinex1981 ; long2008 ; charaziak2026 that are tones whose instantaneous frequency increases (or decreases) from f_{1} to f_{2} —these stimuli are obtained by modulating the frequency of a sinusoidal oscillator with e.g., a linear ramp. Indeed, Schroeder schroeder1970 synthesizes his harmonic complexes essentially by approximating periodic FM sweeps with a sum of sinusoids. While FM sweeps that are made periodic do not manifest the problem noted with the Schroeder’s harmonic complexes, the sudden change in their instantaneous frequency from f_{2} to f_{1} at the beginning of each period, produces a localized broadband response. In laymen’s terms, periodic FM sweeps sound very much like the superposition of a tone whose frequency changes over time, with a click train. Here, I propose a method to design periodic, band-limited stimuli that retain the time-varying frequency characteristic found in Schroeder’s complexes and periodic FM sweeps, while suppressing undesirable steady and click-like signal components found in those stimuli. Note : it is not my intention to review the theory and history of Schroeder’s harmonic complexes—the interested reader is referred to Schroeder’s paper and related literature. schroeder1970 ; blachman1969 ; lentz2001 II Schroeder’s harmonic complexes Schroeder’s harmonic complexes are band-limited ( f\in[f_{1},f_{2}] ) periodic signals with repetition rate r and period T=1/r . They are the linear superposition of N harmonics s(t)=\sum_{n=0}^{N}A_{n}\sin(2\pi r(j+n)t+\phi_{n}), (1) with f_{1}/r=j\in\mathbb{N} , f_{2}/r=l\in\mathbb{Z} , and N=l-j+1 . The phase terms \phi_{n} are selected so that each harmonic has a specific group delay; the coefficients A_{n} control the harmonics’ amplitude. Schroeder designed the phase term specifically to minimize the waveform’s crest factor. He achieved that goal essentially by spacing the group delays of the different harmonics across the stimulus period, in a way that depends on the harmonics’ relative power p_{k}=A_{k}^{2}/\sum_{n}A_{n}^{2} . The resulting equation for calculating the phase is \phi_{n}=\phi_{0}-2\pi\sum_{k=0}^{n-1}\sum_{i=0}^{k}p_{i}. (2) Note that in the literature the name “Schroeder complex”, is typically used to describe variations of this stimulus in the special case that all harmonics have the same amplitude (e.g., A(f)=1 with f\in[f_{1},f_{2}] ). In this case, assigning \phi_{0}=0 yields \phi_{n}=-\pi\frac{n(n+1)}{N}, (3) i.e., the stimulus has a quadratic phase curvature and constant amplitude in [f_{1},f_{2}] . This means the resulting signal has approximately flat envelope and instantaneous frequency increasing linearly over time from f_{1} to f_{2} . schroeder1970 Popular variations of these stimuli include additional parameters to controls the signal’s crest factor and phase curvature. lentz2001 The top section of Fig. 1 shows an example Schroeder’s harmonic complex, where f_{1}=1.6 kHz, f_{2}=6.4 kHz and r=50 Hz. Note3 From left-to-right, the figure shows the signal’s waveform, its amplitude spectrum, and its time-frequency representation through a spectrogram. The spectrogram of the Schroeder’s complex is very interesting. It is dominated by diagonal “sawtooth-like” lines, that represent a signal component whose frequency periodically increases linearly over time. However, the spectrogram also shows two not-so-faint horizontal lines centered around f_{1} and f_{2} , meaning that the signal comprises two steady “frequency-fixed” components of frequencies f_{1} and f_{2} . As anticipated in the introduction, when listening to Schroeder’s harmonic complexes, especially at low modulation rates, these two steady components stand out and are audible. Note4 The presence of these frequency-fixed components can be understood by noting that the group delay \tau (within one signal period) can be calculated from the signal’s phase curvature for all harmonics with frequency f\in(f_{1},f_{2}) : \tau_{n}=-\frac{1}{2\pi}\frac{d\phi}{df}=-\frac{1}{2\pi r}\frac{d\phi_{n}}{dn}=T\Big(\frac{n}{N}+\frac{1}{2N}\Big)\qquad\text{for }0<n<N, (4) where the reader must forgive the abuse of notation. Note5 However, because the signal phase is undefined for f<f_{1} and for f>f_{2} , the phase curvature and hence the group delay cannot be calculated at n=0 and n=N . III Periodic FM Sweeps A frequency-modulated (FM) sweep is obtained by modulating the frequency of a sinusoidal oscillator such that its instantaneous frequency “sweeps” from f_{1} to f_{2} over the signal’s period T . The sinusoidal oscillator can be defined as s(t)=\sin(\phi(t)), (5) where \phi(t) indicates the instantaneous phase of the oscillator. The oscillator’s instantaneous frequency is f_{i}=\frac{1}{2\pi}\frac{d\phi}{dt} . In case of a sweep where the instantaneous frequency increases linearly over time from f_{1} to f_{2} over the period T : f_{i}(t)=\frac{1}{2\pi}\frac{d\phi}{dt}=f_{1}+\frac{f_{2}-f_{1}}{T}t\qquad\text{for }0\leq t<T, (6) and the instantaneous phase is \phi(t)=2\pi\int f_{i}dt=2\pi\Big(f_{1}t+\frac{f_{2}-f_{1}}{2T}t^{2}\Big)\qquad 0\leq t\leq T, (7) where, for simplicity, we have assumed that \phi(0)=0 . These sweeps are made periodic simply by “cycling” \phi over one period: \phi(t)=2\pi\Big(f_{1}t+\frac{f_{2}-f_{1}}{2T}(t-T\lfloor\frac{t}{T}\rfloor)t\Big). (8) It is important to appropriately select the frequency limits and period so that \phi(T)=2\pi k with k\in\mathbb{N} , to avoid waveform’s discontinuities at the beginning of each cycle. The middle panel of Fig. 1 shows an example of a linear FM sweep. This signal shares many similarities with the Schroeder’s harmonic complex in the panel above, but also shows some key differences. While the Schroeder’s complex has a perfectly flat frequency spectrum within f_{1} and f_{2} , but presents visible ripple in its envelope, the opposite is true for the FM sweep: its envelope is perfectly flat, but its amplitude spectrum shows visible ripples. More interesting, the horizontal lines visible in the spectrogram of the Schroeder’s complex, are absent from the spectrogram of the FM sweep. However, the latter shows vertical lines at the beginning of each cycle, indicating the presence of a highly localized broad-band signal component. The presence of such component is explained by that, even