deep_time/physics/drift.rs
1//! Quadratic polynomial for relativistic corrections, clock drift, and custom timescale steering.
2//!
3//! Used to model the accumulated difference between Proper time (τ)
4//! and a coordinate time such as TT (or any other `Scale`).
5//!
6//! Information on the underlying physical model (the master Lagrangian, different
7//! regimes of behavior, and its relationship to general relativity) can be found
8//! [here](https://github.com/ragardner/deep-time/blob/main/docs/relativity.md).
9
10use crate::{
11 ATTOS_PER_SEC_I128, C_SQUARED, Dt, PLANCK_LENGTH_4, Real, Scale, Spacetime, Velocity, sqrt,
12};
13
14/// Quadratic polynomial that describes the accumulated difference between an
15/// observer’s proper time (the time measured by a real clock moving through
16/// spacetime) and a chosen coordinate time such as TT, TAI, or any other
17/// `Scale`.
18///
19/// The polynomial follows the classic form
20/// Δt = constant + rate·Δt + accel·(Δt)²
21/// where the three coefficients capture any fixed offset, constant drift, and
22/// quadratic acceleration of the clock. This structure is used throughout
23/// spacecraft navigation, GNSS systems, and relativistic timing pipelines to
24/// steer clocks, predict time offsets, and maintain synchronization over long
25/// durations.
26///
27/// All three coefficients are stored using [`Dt`].
28#[cfg_attr(feature = "serde", derive(serde::Serialize, serde::Deserialize))]
29#[cfg_attr(feature = "tsify", derive(tsify::Tsify))]
30#[derive(Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash)]
31pub struct Drift {
32 /// Constant term a₀ expressed in seconds.
33 /// This represents any fixed time offset between the observer’s proper time
34 /// and the chosen coordinate time.
35 pub constant: Dt,
36
37 /// Linear drift rate a₁ expressed in seconds per second.
38 /// This term captures a steady fractional rate difference (for example, a
39 /// clock that runs consistently fast or slow).
40 pub rate: Dt,
41
42 /// Quadratic acceleration term a₂ expressed in seconds per second squared.
43 /// This term accounts for any changing drift rate, such as the gradual
44 /// acceleration caused by relativistic effects or hardware aging.
45 pub accel: Dt,
46}
47
48impl Drift {
49 /// Creates a new `Drift` polynomial from its three coefficients.
50 #[inline]
51 pub const fn new(constant: Dt, rate: Dt, accel: Dt) -> Drift {
52 Self {
53 constant,
54 rate,
55 accel,
56 }
57 }
58
59 /// The zero polynomial representing no correction at all.
60 ///
61 /// Use this when the observer’s clock is already perfectly synchronized with
62 /// the chosen coordinate time.
63 pub const ZERO: Self = Self::new(Dt::ZERO, Dt::ZERO, Dt::ZERO);
64
65 /// Creates a [`Drift`] consisting of a pure constant offset.
66 ///
67 /// This is the most common constructor when only a fixed time bias is known
68 /// (for example, after a one-time clock synchronization or leap-second
69 /// adjustment).
70 #[inline]
71 pub const fn from_constant(c: Dt) -> Drift {
72 Self::new(c, Dt::ZERO, Dt::ZERO)
73 }
74
75 /// Creates a [`Drift`] consisting of a constant offset together with a
76 /// constant linear drift rate.
77 ///
78 /// This form is very common for GNSS receivers and spacecraft clock steering,
79 /// where a steady fractional frequency offset must be corrected in addition
80 /// to any fixed bias.
81 #[inline]
82 pub const fn from_offset_and_rate(offset: Dt, rate: Dt) -> Drift {
83 Self::new(offset, rate, Dt::ZERO)
84 }
85
86 /// Returns the instantaneous proper-time rate `dτ/dt` (dimensionless).
87 ///
88 /// This value tells you how fast a real physical clock (such as a spacecraft
89 /// onboard clock) is advancing compared to coordinate time. A value of
90 /// `1.0` means the clock runs at the normal rate. Values slightly below `1.0`
91 /// are typical when the clock is moving or sitting in a gravitational well.
92 ///
93 /// The rate includes special-relativistic velocity effects, gravitational
94 /// time dilation, and the library’s built-in Planck-scale saturation term.
95 #[inline]
96 pub const fn proper_time_rate(&self) -> Real {
97 f!(1.0) + self.rate.to_sec_f()
98 }
99
100 /// Evaluates the polynomial at the given elapsed coordinate time span.
101 ///
102 /// Returns the accumulated time difference (in seconds) between proper
103 /// time and coordinate time after the interval span has passed.
104 pub const fn time_diff_after(&self, span: &Dt) -> Dt {
105 let dt_attos = span.to_attos();
106 let mut total_attos = self.constant.to_attos();
107
108 if !self.rate.is_zero() || !self.accel.is_zero() {
109 // Linear term: rate * dt
110 let rate_attos: i128 = self.rate.to_attos();
111 let rate_term = rate_attos.wrapping_mul(dt_attos) / ATTOS_PER_SEC_I128;
112 total_attos = total_attos.wrapping_add(rate_term);
113
114 // Quadratic term: accel * dt²
115 let accel_attos: i128 = self.accel.to_attos();
116 let accel_dt = accel_attos.wrapping_mul(dt_attos) / ATTOS_PER_SEC_I128;
117 let accel_term = accel_dt.wrapping_mul(dt_attos) / ATTOS_PER_SEC_I128;
118 total_attos = total_attos.saturating_add(accel_term);
119 }
120
121 crate::dt!(total_attos)
122 }
123
124 /// Evaluates the deterministic relativistic/polynomial correction **and**
125 /// adds a user-supplied stochastic offset (in seconds).
126 ///
127 /// This is the single production method for realistic stochastic clock
128 /// modeling. In real mission pipelines the deterministic part (this
129 /// polynomial) is kept perfectly clean; stochastic noise (white phase noise,
130 /// random-walk frequency noise, Monte-Carlo realizations, Kalman process
131 /// noise, measured clock residuals, etc.) is added at evaluation time.
132 ///
133 /// Pass `0.0` (or simply call the original `time_diff_after`) when you
134 /// want purely deterministic behavior.
135 #[inline]
136 pub fn time_diff_after_with_noise(&self, span: &Dt, stochastic_offset_sec: Real) -> Dt {
137 self.time_diff_after(span).add(Dt::from_sec_f(
138 stochastic_offset_sec,
139 Scale::TAI,
140 Scale::TAI,
141 ))
142 }
143
144 /// Build a linear-rate [`Drift`] from speed (m/s) and SI potential Φ (m²/s²).
145 ///
146 /// Given how fast you move and how deep you sit in gravity, return a
147 /// [`Drift`] whose rate term matches the library’s proper-time model
148 /// (special-relativistic and gravitational effects). Useful when you want
149 /// the rate as a polynomial coefficient rather than integrating a path.
150 ///
151 /// ## `characteristic_length_scale`
152 ///
153 /// Pass **`0.0`** for ordinary weak-field work (Earth orbit, solar system):
154 /// Kretschmann is zero and the rate is the first-order weak-field form.
155 /// Pass a positive length (meters) only if you want the optional curvature
156 /// estimate (see [`Spacetime::kretschmann_from_potential_and_scale`]).
157 pub const fn from_velocity_potential_and_scale(
158 velocity_m_s: Real,
159 grav_potential_m2_s2: Real,
160 characteristic_length_scale: Real,
161 ) -> Drift {
162 let phi = grav_potential_m2_s2 / C_SQUARED;
163 let velocity = Velocity::from_speed(velocity_m_s);
164 let spacetime = Spacetime::from_potential_velocity_and_scale(
165 phi,
166 velocity,
167 characteristic_length_scale,
168 );
169 Self::from_spacetime(&spacetime)
170 }
171
172 /// Canonical low-level constructor that implements the library's general
173 /// relativity formula.
174 ///
175 /// This function is the single source of truth for the proper-time rate
176 /// calculation used throughout the library. Most users will never call it
177 /// directly; the high-level constructors `from_velocity_potential_and_scale`
178 /// and `from_spacetime` are the intended entry points.
179 ///
180 /// The internal expression is
181 /// K_eff = [δ(1 + x) + x(1−δ)²] / (1 + x)
182 /// where δ = α²(1−β²) and x = ℓ_Pl⁴ 𝒦.
183 ///
184 /// The returned rate offset is then applied as a linear term in the `Drift`
185 /// polynomial.
186 pub const fn from_unified_proper_time_rate(u: Real, kretschmann: Real) -> Drift {
187 let delta = u.max(f!(0.0));
188 let x = PLANCK_LENGTH_4 * kretschmann.max(f!(0.0));
189
190 let one_minus_delta = f!(1.0) - delta;
191 let num = delta * (f!(1.0) + x) + x * (one_minus_delta * one_minus_delta);
192 let k_eff = num / (f!(1.0) + x);
193
194 let rate_factor = sqrt(k_eff).max(f!(0.0));
195 let rate_offset = rate_factor - f!(1.0);
196
197 Self::from_offset_and_rate(
198 Dt::ZERO,
199 Dt::from_sec_f(rate_offset, Scale::TAI, Scale::TAI),
200 )
201 }
202
203 /// Creates a `Drift` from a fully resolved `Spacetime` snapshot.
204 ///
205 /// This is the canonical high-level entry point when you already hold a
206 /// `Spacetime` object containing the gravitational lapse factor α, the
207 /// local velocity β, and the Kretschmann scalar. It internally computes the
208 /// unified proper-time rate and packages the result as a `Drift`
209 /// polynomial ready for evaluation at any future time.
210 #[inline]
211 pub const fn from_spacetime(spacetime: &Spacetime) -> Drift {
212 let u = spacetime.alpha * spacetime.alpha * (f!(1.0) - spacetime.beta * spacetime.beta);
213 Self::from_unified_proper_time_rate(u, spacetime.kretschmann)
214 }
215}
216
217impl Dt {
218 /// Builds a clock-drift model in which this [`Dt`] is treated as the
219 /// initial fixed time difference between the observer’s proper time and
220 /// the chosen coordinate time.
221 ///
222 /// In practice you often compute or measure a one-time offset (for example
223 /// after a clock synchronization or a leap-second jump) and then want to
224 /// combine it with a steady rate difference and any quadratic change.
225 /// This method lets you do that directly from a [`Dt`] without having to
226 /// call the more verbose [`Drift::new`].
227 ///
228 /// The other two arguments describe how the difference between the two
229 /// clocks will evolve:
230 /// - `rate` — the constant fractional speed difference (how much faster or
231 /// slower one clock runs compared with the other).
232 /// - `accel` — how quickly that speed difference itself is changing (for
233 /// example because the spacecraft is moving through a varying gravitational
234 /// field).
235 ///
236 /// See [`Drift`] and [`Drift::from_offset_and_rate`] for more background on
237 /// why these three numbers are used to model real clocks.
238 #[inline]
239 pub const fn to_drift_as_constant(self, rate: Dt, accel: Dt) -> Drift {
240 Drift::new(self, rate, accel)
241 }
242
243 /// Builds a clock-drift model in which this [`Dt`] supplies the constant
244 /// fractional rate difference between the observer’s proper time and the
245 /// chosen coordinate time.
246 ///
247 /// If you have already calculated (or measured) a steady rate offset as a
248 /// [`Dt`], you can use this method to attach an initial time offset and a
249 /// quadratic term and obtain a complete [`Drift`] polynomial.
250 ///
251 /// Physically, the rate term captures the fact that two clocks that are
252 /// moving at different velocities or sitting at different gravitational
253 /// potentials will accumulate a steadily growing time difference. The
254 /// other two parameters let you also describe any starting bias and any
255 /// change in that rate over time.
256 ///
257 /// See the documentation on [`Drift`] for the meaning of the three
258 /// coefficients in a relativistic timing context.
259 #[inline]
260 pub const fn to_drift_as_rate(self, constant: Dt, accel: Dt) -> Drift {
261 Drift::new(constant, self, accel)
262 }
263
264 /// Builds a clock-drift model in which this [`Dt`] supplies the quadratic
265 /// term that describes how the rate difference itself is changing.
266 ///
267 /// Some situations (a spacecraft on a highly elliptical orbit, a clock
268 /// whose frequency is aging, or a trajectory that takes it through regions
269 /// of changing gravitational potential) cause the *rate* at which two
270 /// clocks diverge to change over time. If you have computed that changing
271 /// rate as a [`Dt`], this method lets you combine it with an initial offset
272 /// and a base rate to form a full [`Drift`].
273 ///
274 /// The other two arguments are:
275 /// - `constant` — any fixed time bias present at the start.
276 /// - `rate` — the base fractional rate difference that will itself be
277 /// modified by the quadratic term supplied by `self`.
278 ///
279 /// See [`Drift`] for more explanation of why a quadratic model is used for
280 /// relativistic clock predictions.
281 #[inline]
282 pub const fn to_drift_as_accel(self, constant: Dt, rate: Dt) -> Drift {
283 Drift::new(constant, rate, self)
284 }
285
286 /// Advances this `Dt` by the given elapsed duration while applying the relativistic proper-time correction
287 /// derived from the supplied `Spacetime` model.
288 ///
289 /// - This method is intended for simulation of remote clocks (e.g., Earth time as observed from a spacecraft).
290 /// - For a local hardware proper-time clock, use the plain `add` methods instead.
291 #[inline]
292 pub const fn adjusted_advance(&mut self, elapsed: &Dt, spacetime: &Spacetime) {
293 let dtau = elapsed.add(Drift::from_spacetime(spacetime).time_diff_after(elapsed));
294 *self = self.add(dtau);
295 }
296
297 /// Advances this `Dt` by the given elapsed duration while applying the relativistic proper-time correction
298 /// from a pre-computed `Drift` value.
299 ///
300 /// - This is an optimized variant of [`Dt::adjusted_advance`](../struct.Dt.html#method.adjusted_advance)
301 /// for callers that already hold a [`Drift`] instance.
302 /// - This method is intended for simulation of remote clocks (e.g., Earth time as observed from a spacecraft).
303 /// - For a local hardware proper-time clock, use the plain `add` methods instead.
304 #[inline]
305 pub const fn adjusted_advance_using_drift(&mut self, elapsed: &Dt, drift: &Drift) {
306 let dtau = elapsed.add(drift.time_diff_after(elapsed));
307 *self = self.add(dtau);
308 }
309
310 /// Converts this instant to any other [`Scale`] while applying an exact quadratic relativistic
311 /// or clock-drift correction defined by a [`Drift`] model relative to a reference instant.
312 pub const fn convert_using_drift(self, reference: Dt, drift: &Drift) -> Dt {
313 let span = self.to_diff_raw(reference);
314 let correction = drift.time_diff_after(&span);
315 self.add(correction)
316 }
317
318 /// Performs the inverse conversion of [`Dt::convert_using_drift`], recovering the original proper
319 /// time on the source clock scale.
320 ///
321 /// A fixed-point iteration (at most 16 steps) is used to solve the implicit equation. For the common
322 /// case of a pure constant offset the function returns immediately without iteration.
323 pub const fn convert_back_using_drift(self, reference: Dt, drift: &Drift) -> Dt {
324 if drift.rate.is_zero() && drift.accel.is_zero() {
325 return self.sub(drift.constant);
326 }
327 let mut guess = self;
328 let mut i = 0u32;
329 while i < 16 {
330 let span = guess.to_diff_raw(reference);
331 let correction = drift.time_diff_after(&span);
332 guess = self.sub(correction);
333 i += 1;
334 }
335 guess
336 }
337}
338
339#[cfg(feature = "wire")]
340impl Drift {
341 /// Current wire format version.
342 pub const WIRE_VERSION: u8 = 1;
343
344 /// Size of the canonical wire representation in bytes.
345 pub const WIRE_SIZE: usize = 3 * Dt::WIRE_SIZE;
346
347 /// Serializes this [`Drift`] polynomial into a fixed buffer.
348 ///
349 /// The layout is the concatenation of the three `Dt` fields.
350 pub fn to_wire_bytes(&self) -> [u8; Self::WIRE_SIZE] {
351 let mut buf = [0u8; Self::WIRE_SIZE];
352 let c = self.constant.to_wire_bytes();
353 let r = self.rate.to_wire_bytes();
354 let a = self.accel.to_wire_bytes();
355
356 buf[0..Dt::WIRE_SIZE].copy_from_slice(&c);
357 buf[Dt::WIRE_SIZE..2 * Dt::WIRE_SIZE].copy_from_slice(&r);
358 buf[2 * Dt::WIRE_SIZE..].copy_from_slice(&a);
359 buf
360 }
361
362 /// Deserializes a [`Drift`] from exactly `WIRE_SIZE` bytes of wire data.
363 ///
364 /// Returns `None` if any nested `Dt` fails validation or if the version
365 /// byte is unknown.
366 ///
367 /// ## Security
368 ///
369 /// Composes the safety guarantees of
370 /// [`Dt::from_wire_bytes`](../struct.Dt.html#method.from_wire_bytes).
371 ///
372 /// Fixed size and layered validation make it safe for untrusted input.
373 pub fn from_wire_bytes(bytes: &[u8]) -> Option<Self> {
374 if bytes.len() != Self::WIRE_SIZE {
375 return None;
376 }
377
378 if bytes[0] != Self::WIRE_VERSION {
379 return None;
380 }
381
382 let constant = Dt::from_wire_bytes(&bytes[0..Dt::WIRE_SIZE])?;
383 let rate = Dt::from_wire_bytes(&bytes[Dt::WIRE_SIZE..2 * Dt::WIRE_SIZE])?;
384 let accel = Dt::from_wire_bytes(&bytes[2 * Dt::WIRE_SIZE..])?;
385
386 Some(Self::new(constant, rate, accel))
387 }
388}