Advanced Z-Transform¶

Track a continuous signal between sampling instants with a phase-indexed family of Z-transforms.

Version: v1 · 2026-10-04 · History
Domain-specific #: 13708
Domain group: Applied Sciences & Engineering
Origin domain: Engineering & Design (beyond software)
Subdomain: Sampled Data Control → Engineering & Design (beyond software)
Aliases: Modified Z-transform, Delayed Z-transform

Core Idea¶

The advanced, or modified, Z-transform adds a within-period phase to an ordinary Z-transform. With sampling period T>0, a dimensionless phase 0≤m<1, and a continuous-time signal f(t), one unilateral convention is F_m(z)=Σ_{k=0}^∞ f((k+m)T)z^{-k}, on the appropriate region of convergence. For each fixed m, this is the Z-transform of one sampled sequence. Taken as a family across m, the transforms distinguish observations at the sample clock from those between clock ticks. Karpenko, Halverson and Besser's original spacecraft-control analysis explicitly samples at t=(k+m)T and transforms a two-exponential plant response at that phase in its equations (15)–(16), giving a page-verifiable instance of this convention.^[1]

This is a specialist method for sampled-data analysis, not a claim that the m=0 sequence uniquely determines the whole continuous waveform. A dimensional offset τ∈[0,T) gives the equivalent sample time kT+τ when τ=mT. A causal input–output transfer calculation may additionally include an integer delay or a z prefactor, as the spacecraft paper's equation (14) does. Those factors are not silently part of the simple Z-series displayed here; declare the output and index convention before manipulating either expression.^[1]

Structural Signature¶

Sig role-phrases:

Continuous-time carrier — f(t) supplies values between sample instants, not only at integer ticks.
Base sampling period — T partitions time into repeated sampling intervals.
Within-period phase — Dimensionless m selects the instant (k+m)T in every interval; equivalently, dimensional τ=mT selects kT+τ.^[1]
Phase-specific sequence — Holding m fixed yields f(mT), f((1+m)T), ….
Z-series with convention — Transform that sequence using a declared unilateral or bilateral index and a valid region of convergence.
Phase family — Compare the results as m varies when the question concerns intersample behavior.^[2]

What It Is Not¶

Not merely the ordinary Z-transform of one fixed sequence. Each fixed-phase member is one, but the distinctive identity is the linked family indexed by phase.
Not guaranteed continuous-signal reconstruction from clock samples alone. Distinct continuous signals can agree at every kT while differing between those instants.
Not a single universal fractional-delay transfer function. Exact form depends on the plant, hold, initial conditions, sampling convention and delay model.
Not an integer shift in disguise. mT resolves a within-period position, whereas z^{-r} alone encodes an integer sample shift.
Not convention-free. m measured as a fraction and τ measured in time cannot be interchanged without the factor T; a causal transfer may also include an integer delay not present in the raw sequence transform.^[1]

Scope of Application¶

The phase-indexed transform belongs to sampled-data control and related discrete-continuous systems. A digital controller sees discrete measurements, while its plant evolves continuously between them. Examining only the update instants can hide intersample ripples or other behavior. A spacecraft attitude-control study uses phase-shifted samplers and a modified Z-transform to analyze a downsample-and-hold ripple missed by the on-grid model; Haba and colleagues discuss the related method while using a different analysis in their motor experiment.^[1]^[2]

The construction presumes a well-defined underlying signal and appropriate transform convergence. Whether a particular model yields a rational expression in z or a useful exact delay formula is additional mathematics, not part of the generic definition. The 1959 discrete-continuous control literature already describes modified-transform use; the method is not attributed to any one modern application.^[3]

Clarity¶

m is a phase, not a measured delay in seconds, under the convention used here. If T=2 seconds and m=1/4, the samples occur at 0.5, 2.5, 4.5, … seconds. A dimensional offset τ=0.5 seconds means the same sample times. In Karpenko and colleagues' paper, the Z-transform of the phase-sampled plant response (equation 16) is then inserted into a transfer expression with an added causal delay (equation 14). Applying that transfer prefactor to the raw series—or omitting it from the causal system expression—would change the first output sample.^[1]

At fixed m, ordinary algebraic properties of the Z-transform apply subject to their usual conditions—convergence, indexing, initial terms and sidedness. Varying m does not license extrapolation of an arbitrary analog waveform from one phase. The phase family represents information provided by the underlying continuous-time model or measurements at those phases.^[1]^[2]

Manages Complexity¶

Sampled-data systems mix two clocks: discrete control updates and continuous plant evolution. The advanced Z-transform packages that mismatch as a family of sequence transforms rather than leaving all intersample behavior outside the discrete analysis. It can reveal how an output differs at, say, the middle of each sample interval from its value at the endpoints.^[2]

The compression requires care. F_0(z) alone is economical but discards phase variation. The full F_m(z) family contains richer information, yet a computation at only finitely many phases still leaves unmeasured times unless the model supplies enough structure. Later intersample modeling research explicitly discusses the limits of finite extra sampling points.^[2]

Abstract Reasoning¶

Let f(t)=e^{-at} for t≥0 and a>0. At phase m, the sequence is e^{-a(k+m)T}=e^{-amT}(e^{-aT})^k. Thus F_m(z)=e^{-amT}/(1-e^{-aT}z^{-1}) for |z|>e^{-aT} under this unilateral convention. Every fixed phase has the same pole but a different amplitude factor. This elementary derivation displays what the phase parameter contributes without claiming every sampled-data transfer function has that form.

The spacecraft paper supplies a more structured case. For its continuous plant response p(t)=ab(e^{-at}-e^{-bt})/(b-a) with distinct plant parameters a,b, phase sampling yields p((k+m)T)=ab[e^{-a(k+m)T}-e^{-b(k+m)T}]/(b-a). Transforming term by term gives P_m(z)=ab/(b-a)[e^{-amT}/(1-e^{-aT}z^{-1})-e^{-bmT}/(1-e^{-bT}z^{-1})] where both series converge. At m=1/2 and the paper's downsampled T=10 s, the two numerator factors become e^{-5a} and e^{-5b}, rather than 1 at m=0. This is the original paper's equations (15)–(16) specialized to half-period phase; its full control-system transfer additionally includes the causal delay and hold factors of equation (14).^[1]

Knowledge Transfer¶

The same operation can be used wherever one continuous signal is examined at a repeated fractional offset from a sampling clock: plant outputs, held-input responses, and delay analysis in sampled-data systems. What transfers is the relationship between the continuous carrier, T, m and the Z-series. One must re-establish convergence, indexing, hold behavior and any causal transfer prefactor for each new setting. A phase-indexed set of numbers not derived from a common signal and clock would miss the distinctive link.^[1]^[2]

Examples¶

Exponential response at an offset¶

For f(t)=e^{-at} and period T, choose any m∈[0,1). The transformed samples yield F_m(z)=e^{-amT}/(1-e^{-aT}z^{-1}) where |z|>e^{-aT}. Setting m=0 recovers on-grid sampling; varying m tracks the exponential's value at a consistent subperiod offset.

Mapped back: Signal → decaying exponential; period → T; phase → m; sequence → e^{-a(k+m)T}; transform → rational expression with phase-dependent numerator and declared convergence region.

Spacecraft attitude-control ripple in an original modified-transform analysis¶

Karpenko, Halverson and Besser model a Lunar Reconnaissance Orbiter attitude-control example with a 5 Hz attitude-control update and a 0.1 Hz downsample-and-hold command interval. Their on-grid discrete model tracks the held response but misses the continuous-time rate ripple between command samples. They introduce phase-shifted output samplers, 0≤m<1, and explicitly transform the plant term p((k+m)T) in equations (15)–(16). For the paper's T=10 s downsample interval, the half-phase expression above supplies two distinct factors e^{-5a} and e^{-5b}; their difference can vary with phase even though the discrete poles remain fixed. The paper's full causal transfer includes an additional delay, and the authors trace the large ripple primarily to a phase-varying gain rather than merely an extra zero.^[1]

Mapped back: Signal → continuous plant/rate response in the attitude loop; period → downsampled command interval T=10 s; phase → output sampler at (k+m)T; sequence → two-exponential plant response evaluated at that phase; transform → original equation (16), inserted into a separately delayed control transfer; boundary → on-grid agreement does not certify intersample slew-rate compliance.

Structural Tensions¶

Compact on-grid algebra versus intersample fidelity. The ordinary on-grid transform is cheaper to compute and may accurately describe the output at update instants, but it omits phase-specific peaks between them. The spacecraft example makes the consequence concrete: its on-grid discrete model missed a downsample-and-hold rate ripple, while the modified-transform analysis exposed phase-varying gain. Evaluating the family requires a continuous plant/hold model and additional phase analysis; sampling finitely many phases alone does not prove a bound at every instant. Diagnostic: Is the required claim only about kT, or about rate and position throughout kT+mT for 0≤m<1?^[1]^[2]

Phase units, first-sample indexing, convergence and any causal z prefactor are correctness conditions, not competing objectives. The original spacecraft paper separates the raw phase-sampled plant transform in equation (16) from the causal transfer expression in equation (14); carrying a factor from one into the other without the declared model would be an error.^[1]

Structural–Framed Character¶

This abstraction is structural within a sampled-data frame: one continuous carrier generates a phase-indexed family of discrete sequences, each mapped by a Z-series. Evaluative weight is low; whether the extra phase information is useful depends on the control question, not the transform's identity. Human modeling practice chooses the sampling clock and phase, while no institution creates the mathematical relation. The vocabulary travels literally among sampled-data analyses preserving the phase and base clock. Importing “advanced Z-transform” to an ordinary Z-transform or to a generic closer inspection is false recognition, not a portable use of the named construction. Its character: an exact mathematical transform whose intersample carrier and timing constraints keep it domain-specific.

Structural Core vs. Domain Accent¶

Skeletal relation. Hold a within-period phase fixed, sample one continuous signal at that offset in every interval, and Z-transform the resulting sequence; then consider the family as phase varies.

Domain-bound condition. Discrete-continuous control systems supply the clock, plant and intersample question. Without a continuous-time carrier or phase-indexed sampling, the entry collapses to the existing ordinary Z-transform.

Prime bar. A generic parameterized transformation might be a future-prime question, not an asserted live parent. The z^{-k} sequence series linked by (k+m)T is a specialist transform construction and does not clear the prime bar by sharing the word “transform.”

This entry presupposes Z-transform.

The Z-transform is a strict prerequisite under composition/presupposes: every fixed-phase member applies its sequence-to-series operation, while ordinary Z-transforms need no common continuous-time carrier or phase family. The family is not a subtype of one transform; its residual identity is the linked variation of within-period phase.

Relationships to Other Abstractions¶

Current abstraction Advanced Z-Transform Domain-specific

Parents (1) — more general patterns this builds on

Advanced Z-Transform presupposes Z-transform Domain-specific

Each member of the advanced phase family applies an ordinary Z-transform to a phase-sampled sequence.

Hierarchy path (1) — routes to 1 parentless root

Advanced Z-Transform → Z-transform → Transformation → Function (Mapping)

Neighborhood in Abstraction Space¶

Advanced Z-Transform sits in a sparse region of the domain-specific corpus (67^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (2551 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With¶

The ordinary Z-transform has one discrete sequence and need not retain a phase family. The discrete-time Fourier transform concerns frequency representation of a sequence under its own existence conditions, not the within-period phase family. A fractional-delay filter may approximate delayed discrete samples; it is a design object rather than this analysis transform. The Zak transform also has time-frequency structure but a different defining construction.^[1]^[2]

References¶

[1] Mark Karpenko, Julie K. Halverson and Rebecca Besser, “Waypoint Following Dynamics of a Quaternion Error Feedback Attitude Control System”, NASA-hosted original author manuscript, pp. 5–6, equations (14)–(18) and Figures 4–6. The manuscript is marked “For Peer Review”; no final-publication pagination is asserted. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m

[2] Haba and colleagues, “Residue Matching: A Method to Determine Intersample Vibrations in Systems With State Feedback”, IET Control Theory & Applications (2025), original research, especially §1.1 on modified/advanced Z-transform and finite-phase sampling limits. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h

[3] “A Contribution to the Statistical Treatment of Discrete-Continuous Sampled Data Control System”, 1959 original research abstract, describing modified Z-transform use; full article not checked. registry ↩