Fact-checked by Grok 2 weeks ago

Decimation

Decimation was a severe form of in , whereby every tenth soldier in a mutinous, cowardly, or disobedient unit was executed by his comrades to enforce discipline and . The term derives from the Latin decimare, meaning "to take a tenth," reflecting the selection of one in ten for execution, and it was regarded as an "ancestral punishment" dating back to at least the fifth century BCE. The procedure typically began with a isolating a subunit, such as a century of roughly 100 men, from the offending or following serious infractions like , in , or . Lots were then drawn to identify the condemned, who were clubbed or stoned to death by the other nine in each group of ten, underscoring the unit's shared guilt. The survivors faced additional degradation, including rations of rather than and exclusion from the main , to reinforce the lesson without fully destroying the unit's fighting strength. Though rare due to its brutality—Polybius noted it was seldom applied in his era—decimation was revived during crises, such as by in 71 BCE, who decimated a unit of 4,000 troops (executing one in ten) after their failure against in the Third Servile War. threatened it in 49 BCE against the 9th Legion for mutiny, and it persisted into the imperial period, with the Third Legion Augusta decimated in 18 CE and the last known instance under Emperor Diocletian in the late third century CE. The practice eventually declined, likely influenced by the and evolving military norms. In contemporary usage, "decimation" has taken on a technical meaning in , where it describes downsampling a signal by an integer factor—low-pass filtering followed by discarding all but every nth sample—to reduce data volume while preserving essential information. This application, unrelated to the punitive origins, highlights the term's evolution into denoting reduction by a tenth or similar proportion.

Etymology and General Definition

Linguistic Origins

The term "decimation" originates from the Latin decimatio, denoting the removal or destruction of one-tenth, derived from decimus, meaning "tenth." This root reflects the broader practice of apportioning a tenth, as seen in decima, which referred to a or exacted as one-tenth of produce or property. In its literal sense, decimatio involved the selection of every tenth item or individual, initially tied to economic or religious obligations rather than violence. The earliest recorded uses of decimatio appear in Roman literature describing events from the 5th century BCE, framing it as an established disciplinary measure. historian , in (Book 2, Chapter 59), recounts its application in 471 BCE following a defeat against the , where Appius selected every tenth soldier by lot for execution to restore order. Similarly, Greek historian , writing in the 2nd century BCE, details the procedure in The Histories (Book 6, Chapter 38), portraying decimatio as a standard punishment for , involving lots to choose approximately one-tenth of the offending unit for beating to death by comrades. These accounts, though composed centuries after the events, indicate the term's antiquity in documenting punitive practices. Over time, decimatio evolved from its neutral connotation of or proportional selection—rooted in agricultural and fiscal customs—into a specifically punitive execution targeting every tenth member of a group for severe offenses. This shift emphasized in contexts, transforming a of into one of and deterrence.

Contemporary Broad Usage

In , "decimation" has evolved from its historical roots in military to signify or a severe in , often implying near-total rather than a literal tenth. This semantic broadening began in the mid-17th century, as English translations of ancient histories, such as those by , popularized the term in hyperbolic contexts describing mass devastation, gradually extending its application beyond precise proportionality. By the , the figurative sense had become entrenched, reflecting a common linguistic pattern where terms for specific acts expand to convey greater intensity. Contemporary dictionaries reflect this shift, defining "decimation" as the killing or destruction of a large proportion rather than exactly one in ten. For instance, the includes senses such as "the action of destroying or killing a large proportion of" and "severe reduction or depletion," acknowledging usages from the onward. Similarly, lists the primary meaning as "to destroy a large part of," with the Roman-specific sense noted as historical. These entries underscore the term's acceptance in its broadened form across formal and informal contexts. In 20th- and 21st-century , , and public discourse, "decimation" frequently describes profound losses in non-military settings. Environmental reports often invoke it to highlight crises, such as the WWF's Living Planet Report, where analyses describe human activities as leading to the "decimation" of global wildlife populations, with monitored species declining by an average of 73% since 1970. In economic commentary, the term captured the fallout from the , with outlets like reporting how real estate losses "decimated" American banks, contributing to widespread recessionary impacts. Literary works, including Jonathan Franzen's (2001), employ "decimation" metaphorically to depict the erosion of family structures under neoliberal pressures, illustrating its versatility in narrative prose. This evolution has sparked linguistic debates between prescriptivists, who argue for restricting "decimation" to its original tenth-based meaning to preserve etymological precision, and descriptivists, who view the broadening as a natural adaptation reflecting actual usage. Prescriptivists, often citing classical sources, contend that loose applications dilute the term's historical specificity, as seen in style guides like those from , which caution against hyperbolic extensions. Descriptivists counter that inherently changes through communal use, pointing to evidence from corpora like the showing the figurative sense dominating since the 19th century, and advocate acceptance to avoid unnecessary pedantry.

Historical Military Context

Roman Decimation Practice

Decimation, or decimatio, was a severe disciplinary measure in the ancient Roman army, applied to units guilty of mutiny, cowardice, or desertion. The procedure entailed dividing the offending soldiers into groups of ten and selecting one from each group by lot to be executed immediately by their comrades, who used clubs (bacula) or stones for the killing. This method ensured collective participation, reinforcing the unit's shared responsibility and deterring future infractions through the horror of peer-enforced punishment. The rationale behind decimation stemmed from its role in swiftly restoring amid crises, serving as a stark deterrent against and while emphasizing the army's principle of . Commanders invoked it to reassert authority over wavering legions, viewing the random selection as a way to purge weakness without targeting individuals, thus preserving overall unit cohesion. Prominent leaders such as employed it during turbulent periods of the late , when traditional hierarchies were strained by civil strife and slave revolts. The practice's extremity underscored its use as a last resort, ordered only when lesser punishments failed to quell disorder. Documented instances highlight its application in pivotal conflicts. During the Third Servile War (73–71 BCE), Crassus ordered the decimation of one cohort of approximately 500 men after they fled from Spartacus's forces in 71 BCE, executing 50 by lot to steel the remaining troops for the final campaign that crushed the rebellion. In the era following the of 107 BCE, which professionalized the legions and heightened expectations of loyalty, decimation reemerged sporadically to enforce discipline amid evolving recruitment practices. Later in the Republic, threatened decimation against his Ninth Legion during a 47 BCE in but ultimately commuted it to executing only the ringleaders, demonstrating a shift toward clemency even in extremis. The practice continued into the imperial , though rarely. In 18 , the Legio III Augusta was decimated for in , and the last known instance occurred under Emperor in the late third century . The psychological and social impacts of decimation were profound, instilling terror to bolster short-term obedience but risking long-term resentment and eroded morale within the punished unit. By forcing survivors to kill their fellows, it deepened bonds of fear-based solidarity, yet its brutality often alienated troops, contributing to its rarity—historians document fewer than ten reliable cases across the , primarily in the late , as legal protections like the leges Porciae (195–149 BCE) increasingly limited commanders' arbitrary powers. This infrequency reflected a broader preference for graduated punishments, reserving decimation for moments when the army's survival demanded unyielding terror.

Applications in Other Eras and Armies

While the practice of decimation as a specific punishment largely faded after the classical , echoes of collective disciplinary measures appeared sporadically in later European armies, often as adaptations rather than direct revivals. During , commander employed decimation against mutinous units on the Italian front, drawing lots to execute soldiers among groups suspected of or ; at least two documented instances occurred between and 1917, with dozens executed each time. Similarly, in the early in the war, reports describe a form of decimation in December 1914 near Vieille-Chapelle, where approximately 500-600 soldiers from the 8th Tunisian Tirailleurs Regiment faced selection by lot following a retreat, though only around 50 were ultimately shot. These cases reflected desperate attempts to restore amid high casualties and low , but they were rare and controversial even then. In the , the Soviet during implemented severe collective penalties for desertion, reminiscent of decimation's group accountability, through Stalin's in July 1942, which established blocking detachments to execute retreating or fleeing soldiers. At Stalingrad, these units enforced the policy rigorously, with several hundred executions across the fronts by October 1942, including in the 64th Army, targeting units en masse to deter panic amid the brutal urban fighting. This approach, while not strictly limited to one-in-ten killings, imposed collective terror on entire formations to prevent breakdowns, contributing to the high Soviet casualty rates but ultimately aiding the defense. Non-Western military traditions featured analogous collective punishments, though not termed decimation. In ancient under the (221–206 BCE), the legalist system of lianzuo (mutual responsibility) extended to the army, where soldiers in a unit could face joint liability—including execution or —for one member's failure, such as or disobedience, to enforce strict unity and deter rebellion. This mirrored decimation's intent to bind troops through shared risk, as seen in the harsh during the unification wars. In feudal , samurai codes emphasized individual honor over group penalties, but clans occasionally imposed collective reprisals on retainers' families for betrayal, such as forced or , to maintain loyalty without formal lot-based execution. The decline of such practices accelerated with the rise of modern military codes and , which viewed collective punishments as inhumane and counterproductive to . By the , professional armies like the British emphasized individual courts-martial over group executions, rendering decimation obsolete amid Enlightenment-influenced reforms. The 1949 explicitly prohibited collective penalties in Article 33 of the Fourth Convention, banning punishments for offenses not personally committed to protect prisoners and civilians, a principle extended to armed forces through . This shift prioritized and psychological discipline, eliminating overt revivals. Despite this, the concept endures in military rhetoric, where "decimating the enemy" metaphorically signifies inflicting severe, disproportionate losses, as in descriptions of devastating battles or campaigns.

Technical Context in Signal Processing

Core Principles of Decimation

In (), decimation refers to the process of downsampling a discrete-time signal by an integer factor M, which involves applying a followed by to retain every Mth sample while discarding the others. This combined operation ensures that the signal's content is appropriately bandlimited before reduction, thereby mitigating . The resulting output signal has a sampling rate reduced to f_s / M, where f_s is the original sampling , allowing for efficient representation of the signal's essential information. The primary purpose of decimation is to lower the rate of a signal, which decreases computational demands, requirements, and in various practical systems. For instance, in , decimation reduces the sampling rate of high-fidelity recordings to suitable levels for or playback without perceptible of . Similarly, in , it optimizes in bandwidth-constrained channels, while in sensor networks, it minimizes volume from continuous devices like accelerometers or . By achieving these efficiencies, decimation enables processing in resource-limited environments, such as devices or systems. A core principle underlying decimation is the Nyquist-Shannon sampling theorem, which states that a continuous-time signal can be accurately reconstructed from its samples if the sampling rate exceeds twice the highest frequency component. In decimation, subsampling without prior filtering risks , where high-frequency components fold into the lower-frequency band of the downsampled signal, causing irreversible distortion. Thus, the must attenuate frequencies above the new of f_s / (2M) to preserve signal integrity, ensuring that the decimated version faithfully represents the original within the reduced . The term "decimation" entered literature in the 1970s, drawing an to the historical of thinning a by selecting every Mth element, though it diverges from the punitive of ancient . This adoption reflected the growing field of multirate , with early formalizations appearing in seminal works on efficient for rate conversion. Unlike pure mathematical reduction, decimation in emphasizes practical measures tailored to real-world signals.

Mathematical Implementation and Filtering

The implementation of decimation in digital signal processing involves two primary steps to ensure the integrity of the signal while reducing its sampling rate by an integer factor M: first, applying a low-pass anti-aliasing filter with a cutoff frequency of \pi/M radians per sample to the input sequence x; second, downsampling by retaining only every Mth sample, effectively discarding M-1 samples out of every M. This process prevents spectral folding, or aliasing, that would otherwise distort the signal's frequency content. Mathematically, the decimated output signal y is defined as y = x[M m], where m is the new time index. The preceding H(e^{j\omega}) must have a that approximates an ideal brick-wall characteristic: |H(e^{j\omega})| \approx 1 for |\omega| < \pi/M (passband) and |H(e^{j\omega})| \approx 0 for \pi/M < |\omega| < \pi (stopband), ensuring that frequencies above the new are suppressed. In practice, the filter's transition band is designed to minimize ripple and attenuation while balancing . For efficient implementation, especially in resource-constrained environments, polyphase filter structures decompose the into M subfilters, each operating at the reduced output rate, significantly lowering the computational load by avoiding unnecessary processing of discarded samples. This approach, introduced in foundational multirate literature, enables decimation with reduced multiplier and adder operations compared to direct filtering. In hardware-oriented applications, such as oversampled analog-to-digital converters, cascaded integrator-comb () filters provide a multiplier-free alternative, consisting of N stages followed by downsampling and N comb stages. The of an Nth-order filter with decimation factor M and differential delay R (typically R=1) is given by H(z) = \left[ \frac{1 - z^{-R M}}{1 - z^{-1}} \right]^N, which yields a sinc-like frequency response suitable for sharp initial decimation stages, though it requires compensation filters for droop in later stages. If decimation proceeds without prior low-pass filtering, aliasing artifacts manifest as overlapping spectral replicas in the baseband, potentially corrupting the desired signal components and degrading overall fidelity. In software tools like MATLAB, the decimate(x, M) function automates this process by applying an 8th-order Chebyshev Type I IIR low-pass filter (with cutoff at $0.8 \pi / M) followed by downsampling, or optionally an FIR filter via the 'fir' flag, providing a practical means to implement decimation while controlling phase distortion through zero-phase filtering. Similar functionality is available in Simulink blocks for system-level simulation.

Mathematical and Computational Aspects

Sequence Decimation in

In , sequence decimation refers to the operation of extracting a from an infinite (a_n)_{n=0}^\infty by selecting terms whose indices form an with a fixed step size. For a positive M \geq 2, the M-decimated sequence is defined as (b_k)_{k=0}^\infty, where b_k = a_{M k} for each k \geq 0. This process creates a sparser sequence that retains structural properties of the original while facilitating analysis of subsampled behaviors. The operation is distinct from general , as it strictly follows the arithmetic progression starting from index 0, and it applies to sequences over any , such as integers, fields, or rings. Decimation alters key properties of the sequence, particularly regarding periodicity and recurrence relations. For a periodic sequence with period P, the decimated sequence has period P / \gcd(M, P), which can reduce or preserve the original cyclicity depending on the relation between M and P. In number theory, decimation is employed to construct and study arithmetic progressions within sequences, aiding investigations into density and distribution; for instance, applying decimation to sequences generated by primitive polynomials (such as m-sequences) helps analyze the density of elements satisfying specific modular conditions. In combinatorics, decimation serves as a tool for subset selection, where systematically removing terms from a generating sequence produces combinatorial structures like de Bruijn sequences or permutations by deletion, enabling enumeration of distinct subsampled configurations. A representative example is the decimation of the Fibonacci sequence, defined by the recurrence F_0 = 0, F_1 = 1, and F_n = F_{n-1} + F_{n-2} for n \geq 2. The 2-decimated subsequence (F_{2k})_{k=0}^\infty = 0, 1, 3, 8, 21, \dots satisfies its own second-order linear recurrence F_{2k} = 3 F_{2(k-1)} - F_{2(k-2)} for k \geq 2, with initial terms F_0 = 0 and F_2 = 1. This property demonstrates that decimation of linear recurring sequences often yields another linear recurring sequence, preserving the recursive structure while modifying the characteristic equation. More generally, for M-decimation of Fibonacci-like sequences, the resulting subsequence can be expressed using Binet's formula adaptations, highlighting connections to Lucas numbers and golden ratio powers. Theoretically, decimation impacts the generating functions associated with sequences, providing a means to isolate and manipulate arithmetic-progression terms. For the ordinary generating function G(z) = \sum_{n=0}^\infty a_n z^n of the original sequence, the generating function G_M(z) = \sum_{k=0}^\infty b_k z^k of the M-decimated sequence is G_M(z) = \frac{1}{M} \sum_{j=0}^{M-1} G(\omega^j z^{1/M}), where \omega = e^{2\pi i / M} is a primitive M-th root of unity. This expression, derived from the roots of unity filter, enables the extraction of coefficients in residue classes modulo M, with profound implications for solving combinatorial identities and asymptotic analyses of sequence growth. The filter's efficacy stems from the of roots of unity, ensuring precise isolation without overlap from other residue classes.

Algorithms in Data Processing

In , decimation algorithms facilitate the reduction of a signal's sampling rate by an integer factor M, typically combining low-pass filtering with downsampling to prevent spectral while minimizing computational overhead. These methods are foundational in multirate (), enabling efficient handling of high-rate data in applications such as , image subsampling, and sensor . The core challenge addressed by these algorithms is balancing frequency-domain fidelity with reduced processing complexity, often achieving savings proportional to the decimation factor. The classical decimation algorithm proceeds in two stages: first, a with \pi/M attenuates components that would alias into the upon downsampling; second, the filtered signal is downsampled by retaining every M-th sample, yielding y = x[Mn] where x is the filtered input. This approach, while straightforward, incurs high computational cost since the filter operates at the original high rate, processing M times more samples than necessary for the output. Optimal () filter designs for this process minimize transition-band width or passband under constraints on filter length, as detailed in early work on FIR implementations for decimation. For instance, or equiripple designs ensure sharp with lengths on the order of N \approx 2 f_s / \Delta f, where \Delta f is the transition bandwidth relative to sampling rate f_s, though exact lengths depend on specifications. To enhance efficiency, polyphase decomposition restructures the FIR filter into M parallel subfilters, each operating at the lower output rate. The polyphase filter H(z) = \sum_{k=0}^{M-1} z^{-k} E_k(z^M), where E_k(z) are the polyphase components, allows downsampling to precede most filtering operations, commuting the downsampler through the to eliminate redundant computations on discarded samples. This reduces the multiplier count by approximately a of M, making it ideal for systems; for example, in a decimation-by-4 with a 64-tap , the polyphase form requires only 16 multiplies per output sample instead of 256. Such implementations are widely adopted in software libraries like MATLAB's toolbox and hardware accelerators. A specialized class of decimation algorithms employs cascaded -comb (CIC) filters, which are multiplier-free and suited for high-throughput hardware like FPGAs or . Introduced in , a CIC decimator consists of N stages followed by N comb stages (differencers with delay R M, where R is the differential delay, often 1), separated by the downsampler. The is H(z) = \left[ \frac{1 - z^{-R M}}{1 - z^{-1}} \right]^N, providing a sinc-like with nulls at multiples of the output rate, effectively suppressing without coefficients. CIC filters excel in oversampled systems, such as sigma-delta modulators, where decimation ratios exceed 64; for N=3 and M=8, they achieve over 40 stopband attenuation with zero multiplies, though they require compensation for passband droop via subsequent stages. Their pipelined structure supports clock rates beyond 500 MHz in modern implementations.

References

  1. [1]
    Decimation - Livius.org
    Jul 10, 2020 · Decimation: punishment in the Roman army. Of every ten soldiers, one was executed. Decimation was never a common punishment: it was too harsh ...
  2. [2]
    Decimation - Etymology, Origin & Meaning
    Decimation originates from mid-15c. Old French and Late Latin, meaning the removal or tax of one-tenth, derived from Latin decimare (to take a tenth).
  3. [3]
    Decimation | UNRV Roman History
    Decimation was a brutal form of military punishment employed in ancient Rome to maintain discipline and order within the ranks of the Roman army.
  4. [4]
    Decimation - dspGuru
    Decimation is the process of reducing the sampling rate. In practice, this usually implies lowpass-filtering a signal, then throwing away some of its samples.
  5. [5]
  6. [6]
  7. [7]
    Decimate - Etymology, Origin & Meaning
    Originating c.1600 from Latin decimatus, meaning "removal of one-tenth," decimate means to select and kill every tenth man, reflecting its numeric root "ten ...
  8. [8]
  9. [9]
  10. [10]
    Decimatio: Myth, Discipline, and Death in the Roman Republic
    Dec 13, 2022 · The punishment was possibly used as an extraordinary measure a couple of times in the fifth and fourth centuries BC. It soon fell into total ...Missing: earliest | Show results with:earliest
  11. [11]
    Regarding the Incorrect Use of 'Decimate' - Merriam-Webster
    Decimate had the specific meaning, in ancient Rome, of killing one of every ten soldiers, as a form of military punishment.
  12. [12]
  13. [13]
  14. [14]
    Humans are decimating wildlife, report warns ahead of U.N. ...
    Oct 26, 2022 · Wildlife populations tracked by scientists shrank by nearly 70%, on average, between 1970 and 2018, a recent assessment has found.
  15. [15]
    A Rate of Zero Percent From the Fed? Some Analysts Say It Could ...
    Oct 29, 2008 · Like Japan's, American banks have become so decimated by losses in real estate that they are either unable or unwilling to resume normal lending ...
  16. [16]
    "Neoliberalism in Contemporary Literature: The Nuclear Family's ...
    I contend that Franzen's novel, despite being a work of fiction, is an accurate portrayal of the nuclear family's decimation during the era of neoliberalism.Missing: modern | Show results with:modern
  17. [17]
    Prescriptivism and Language Change - Arrant Pedantry
    Dec 9, 2016 · Arguing that decimated really should mean “reduced by one-tenth” is based on a desire to hang on to rules that one was taught in one's youth.
  18. [18]
    LOOSE DECIMATION. - languagehat.com
    Mar 25, 2013 · John McIntyre's latest column demolishes a reader who insists that decimate must be limited to reduce by a tenth”: There you have the ...
  19. [19]
    Decimation in the Roman Republic (Classical Journal, Winter 2015)
    Jun 23, 2015 · The research finds that in 49 BCE, Caesar threatened decimatio during a mutiny but later spared his troops, executing only the ringleaders. His ...
  20. [20]
    Discipline and Military Justice (Italy) - 1914-1918 Online
    Oct 8, 2014 · Cadorna also imposed decimation, that is the drawing by lots of the soldiers to be executed by firing squad from among the suspects in the case ...
  21. [21]
    lianzuofa 連坐法, law on collective liability - Chinaknowledge
    May 25, 2016 · An very special kind of collective liability was kin liability, the punishment of three generations (sanzu 三族), meaning the extirpation of the ...Missing: ancient army
  22. [22]
    Decimation Filter - an overview | ScienceDirect Topics
    A decimation filter is defined as a filter that bandlimits an input signal prior to downsampling to prevent aliasing, ensuring that the output signal ...
  23. [23]
    Decimation — decrease sample rate by integer factor - MATLAB
    Decimation reduces the sample rate of a signal by a factor of r, shortening the signal by that factor. It is the opposite of interpolation.Missing: definition | Show results with:definition
  24. [24]
    Interpolation and decimation of digital signals—A tutorial review
    In this paper we present a tutorial overview of multirate digital signal processing as applied to systems for decimation and interpolation.
  25. [25]
    [PDF] Chapter 9 – Multirate Digital Signal Processing
    Thus, the formal definition of M-fold decimation, or down-sampling, is defined by Equation 9.1. In decimation, the sampling rate is reduced from Fs to Fs/M ...
  26. [26]
    Design of Decimators and Interpolators - MATLAB & Simulink
    This example shows how to design filters for decimation and interpolation of discrete sequences.
  27. [27]
    [PDF] On the Distinctness of Decimations of -Sequences
    Jan 18, 2002 · Recall that the d-decimation of the sequence a = a0,a1, ททท is the sequence ad = a0,ad,a2d ททท. On the basis of extensive experimental ...Missing: pure | Show results with:pure
  28. [28]
    [PDF] ON DECIMATION OF LINEAR RECURRING SEQUENCES
    ... decimation of s by d, s^, is defined by s(d)(t) = s(td), t > 0. Analogously, given a set S of sequences over F, the decimation of S by d,. S(d\ is defined by ...Missing: pure | Show results with:pure
  29. [29]
    [2010.15215] Decimation and Interleaving Operations in One-Sided ...
    Oct 28, 2020 · We study a family of decimation operations, which extract subsequences of symbol sequences in infinite arithmetic progressions, and show they ...
  30. [30]
    On Minimal Strings Containing the Elements of S_n by Decimation
    The permutations by decimation problem is thought to be applicable to computer graphics, and raises interesting theoretical questions in combinatory theory.
  31. [31]
    Fibonacci numbers - OeisWiki
    Apr 20, 2011 · Even indexed Fibonacci numbers. The even indexed Fibonacci numbers are given by the recurrence. F 0 = 0 ,: F 2 = 1 ,: F 2 n = 3 ...Recurrence equation · Generating function · Formulae · Fibonacci numbers mod n
  32. [32]
    [PDF] generatingfunctionology - Penn Math
    May 21, 1992 · This book is about generating functions and some of their uses in discrete mathematics. The subject is so vast that I have not attempted to give ...Missing: decimation | Show results with:decimation
  33. [33]
  34. [34]
    Perform polyphase FIR decimation - Simulink - MathWorks
    The FIR Decimation block performs an efficient polyphase decimation using an integer downsampling factor M along the first dimension.
  35. [35]
    A Beginner's Guide To Cascaded Integrator-Comb (CIC) Filters
    Mar 26, 2020 · The passband gain of a 1st-order CIC decimation filter, derived in Reference [1], is equal to the comb filter delay D = NR at zero Hz (DC).