comnum 0.2.0__tar.gz → 0.3.1__tar.gz

{comnum-0.2.0 → comnum-0.3.1}/Cargo.lock

@@ -95,7 +95,7 @@ checksum = "1c133bc6a41be0d194c306b5506d15e6feeea7b1d6604bd3f8310dfb2ca96486"
 
 [[package]]
 name = "comnum"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "comnum-channel",
  "comnum-core",
@@ -116,7 +116,7 @@ dependencies = [
 
 [[package]]
 name = "comnum-channel"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "comnum-core",
  "comnum-dsp",
@@ -127,7 +127,7 @@ dependencies = [
 
 [[package]]
 name = "comnum-core"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "ndarray",
  "num-complex",
@@ -136,7 +136,7 @@ dependencies = [
 
 [[package]]
 name = "comnum-dsp"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "comnum-core",
  "comnum-math",
@@ -145,7 +145,7 @@ dependencies = [
 
 [[package]]
 name = "comnum-fec"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "comnum-core",
  "comnum-math",
@@ -153,7 +153,7 @@ dependencies = [
 
 [[package]]
 name = "comnum-hardware"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "comnum-core",
  "comnum-dsp",
@@ -162,7 +162,7 @@ dependencies = [
 
 [[package]]
 name = "comnum-mapping"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "comnum-core",
  "comnum-math",
@@ -171,7 +171,7 @@ dependencies = [
 
 [[package]]
 name = "comnum-math"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "comnum-core",
  "ndarray",
@@ -181,7 +181,7 @@ dependencies = [
 
 [[package]]
 name = "comnum-metrics"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "comnum-core",
  "comnum-dsp",
@@ -191,7 +191,7 @@ dependencies = [
 
 [[package]]
 name = "comnum-mimo"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "comnum-core",
  "comnum-mapping",
@@ -201,7 +201,7 @@ dependencies = [
 
 [[package]]
 name = "comnum-py"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "comnum",
  "ndarray",
@@ -214,7 +214,7 @@ dependencies = [
 
 [[package]]
 name = "comnum-sim"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "comnum-channel",
  "comnum-core",
@@ -226,7 +226,7 @@ dependencies = [
 
 [[package]]
 name = "comnum-spec"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "comnum-channel",
  "comnum-core",
@@ -249,7 +249,7 @@ dependencies = [
 
 [[package]]
 name = "comnum-systems"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "comnum-core",
  "comnum-dsp",
@@ -262,7 +262,7 @@ dependencies = [
 
 [[package]]
 name = "comnum-validation"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "comnum",
  "ndarray",
@@ -271,7 +271,7 @@ dependencies = [
 
 [[package]]
 name = "comnum-waveform"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "comnum-core",
  "comnum-dsp",

{comnum-0.2.0 → comnum-0.3.1}/Cargo.toml

@@ -4,7 +4,7 @@ resolver = "3"
 members = ["crates/comnum-core", "crates/comnum-math", "crates/comnum-dsp", "crates/comnum-fec", "crates/comnum-mapping", "crates/comnum-waveform", "crates/comnum-mimo", "crates/comnum-channel", "crates/comnum-hardware", "crates/comnum-metrics", "crates/comnum-sim", "crates/comnum-spec", "crates/comnum-systems", "crates/comnum", "bindings/comnum-py"]
 
 [workspace.package]
-version = "0.2.0"
+version = "0.3.1"
 edition = "2024"
 rust-version = "1.85"          # MSRV, tested in CI
 license = "MIT OR Apache-2.0"
@@ -17,20 +17,20 @@ categories = ["science", "simulation"]
 
 [workspace.dependencies]
 # the crates of the workspace (one version, released together)
-comnum-core = { path = "crates/comnum-core", version = "0.2.0" }
-comnum-math = { path = "crates/comnum-math", version = "0.2.0" }
-comnum-dsp = { path = "crates/comnum-dsp", version = "0.2.0" }
-comnum-fec = { path = "crates/comnum-fec", version = "0.2.0" }
-comnum-mapping = { path = "crates/comnum-mapping", version = "0.2.0" }
-comnum-waveform = { path = "crates/comnum-waveform", version = "0.2.0" }
-comnum-mimo = { path = "crates/comnum-mimo", version = "0.2.0" }
-comnum-channel = { path = "crates/comnum-channel", version = "0.2.0" }
-comnum-hardware = { path = "crates/comnum-hardware", version = "0.2.0" }
-comnum-metrics = { path = "crates/comnum-metrics", version = "0.2.0" }
-comnum-sim = { path = "crates/comnum-sim", version = "0.2.0" }
-comnum-spec = { path = "crates/comnum-spec", version = "0.2.0" }
-comnum-systems = { path = "crates/comnum-systems", version = "0.2.0" }
-comnum = { path = "crates/comnum", version = "0.2.0", default-features = false }
+comnum-core = { path = "crates/comnum-core", version = "0.3.1" }
+comnum-math = { path = "crates/comnum-math", version = "0.3.1" }
+comnum-dsp = { path = "crates/comnum-dsp", version = "0.3.1" }
+comnum-fec = { path = "crates/comnum-fec", version = "0.3.1" }
+comnum-mapping = { path = "crates/comnum-mapping", version = "0.3.1" }
+comnum-waveform = { path = "crates/comnum-waveform", version = "0.3.1" }
+comnum-mimo = { path = "crates/comnum-mimo", version = "0.3.1" }
+comnum-channel = { path = "crates/comnum-channel", version = "0.3.1" }
+comnum-hardware = { path = "crates/comnum-hardware", version = "0.3.1" }
+comnum-metrics = { path = "crates/comnum-metrics", version = "0.3.1" }
+comnum-sim = { path = "crates/comnum-sim", version = "0.3.1" }
+comnum-spec = { path = "crates/comnum-spec", version = "0.3.1" }
+comnum-systems = { path = "crates/comnum-systems", version = "0.3.1" }
+comnum = { path = "crates/comnum", version = "0.3.1", default-features = false }
 # third-party
 num-complex = "0.4"
 ndarray = "0.16"

{comnum-0.2.0 → comnum-0.3.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: comnum
-Version: 0.2.0
+Version: 0.3.1
 Classifier: Development Status :: 3 - Alpha
 Classifier: Intended Audience :: Science/Research
 Classifier: Topic :: Scientific/Engineering
@@ -47,8 +47,8 @@ the literature.
 <!-- highlights: generated by tools/highlights.py, do not edit -->
 - **Lightweight**: one run-time dependency in Python (numpy), 9 direct crates in Rust and no C, C++ or Fortran library to build; 27 MB installed with its dependencies (the benchmark's environment: komm 22 MB, OptiCommPy 321 MB, Sionna 1004 MB).
 - **Fast**: compiled Rust, Monte Carlo runs on every core; on one CPU thread, the links of the benchmark take 2.1 to 33 times as long with komm, 2.3 to 4.7 times as long with OptiCommPy and 1.1 to 7.4 times as long with Sionna as with comnum.
-- **Safe**: `unsafe` code forbidden in every crate, invalid inputs are errors (one error type), results independent of the thread count; 717 unit tests, 187 validation tests, 272 Python tests.
-- **Validated**: compared with 130 papers, standards and reference tools (82 fully reproduced, 44 partly), 100 inconsistencies found in them; 93 validation cards rerun a comparison without the paper.
+- **Safe**: `unsafe` code forbidden in every crate, invalid inputs are errors (one error type), results independent of the thread count; 749 unit tests, 198 validation tests, 272 Python tests.
+- **Validated**: compared with 139 papers, standards and reference tools (88 fully reproduced, 47 partly), 120 inconsistencies found in them; 104 validation cards rerun a comparison without the paper.
 
 The numbers are recomputed by `tools/highlights.py` from the recorded benchmark and the validation files; the details are on the page [Why comnum](https://com-rs.github.io/comnum/why.html).
 <!-- highlights:end -->
@@ -175,7 +175,7 @@ gives the tests behind every grade and the regimes not covered).
 
 <!-- generated by tools/maturity_table.py from validation/maturity.yaml: do not edit -->
 
-Summary: 71 modules: 66 at A, 3 at B, 2 at C.
+Summary: 75 modules: 70 at A, 3 at B, 2 at C.
 
 **Foundations**
 
@@ -183,6 +183,7 @@ Summary: 71 modules: 66 at A, 3 at B, 2 at C.
 |---|---|---|
 | Constellations, mapping, demapping and closed-form error rates | A | OptiCommPy outputs; closed forms of Proakis & Salehi |
 | Capacity and cutoff rate of CM and BICM | A | Caire, Taricco & Biglieri 1998 Figs. 4–7 |
+| Adaptive MQAM over fading | A | Goldsmith & Chua 1997 Figs. 3, 5, 10, Table II |
 | Sources (bits, PRBS, uniform and shaped symbols) | A | ITU-T O.150 Sec. 5 and Table 1, exact |
 | Differential PSK | A | Pawula, Rice & Roberts 1982 Sec. IV-A; closed forms of Proakis & Salehi |
 | Error rates with an imperfect channel estimate and over Jakes' fading | A | Chang & Su 2002 eqs. 31–53, Figs. 3–4 |
@@ -195,6 +196,8 @@ Summary: 71 modules: 66 at A, 3 at B, 2 at C.
 | Puncturing | A | Yasuda et al. Table I; patterns of IEEE 802.11 and EN 300 421 |
 | Turbo codes (LTE) | A | AFF3CT, K = 40 and 1024 |
 | LDPC codes | A | AFF3CT, PEG and DVB-S2 short code; Hu et al. 2005 |
+| Density evolution of LDPC ensembles | A | Richardson & Urbanke 2001 Tables I-II |
+| EXIT charts of parallel concatenated codes | A | ten Brink 2001 Tables II and IV |
 | Polar codes (SC, CA-SCL) | A | AFF3CT, SC and CA-SCL |
 | 5G NR channel codes: LDPC and polar with rate matching | A | Sionna vectors bit for bit; TS 38.212 tables |
 | Reed–Solomon codes | A | AFF3CT; Lin & Costello; ETSI EN 300 744, CCSDS 131.0-B |
@@ -214,7 +217,7 @@ Summary: 71 modules: 66 at A, 3 at B, 2 at C.
 | Constant-envelope OFDM | A | Thompson et al. 2008 eq. (15); Guerreiro et al. 2016 Fig. 1; López Morales et al. 2022 Fig. 3; Pacheco & Hatzinakos 2006 Fig. 1 |
 | Chirp spread spectrum (LoRa modulation) | A | Chiani & Elzanaty 2019 Table I; Afisiadis et al. 2020 Figs. 2, 7; Vangelista 2017 Fig. 1 |
 | Continuous-phase modulation | A | Murota & Hirade 1981 Table I and Fig. 5; Sundberg 1986 Fig. 5; MSK of Proakis & Salehi |
-| FBMC/OQAM | A | Bouhadda et al. 2014 Table 1, Figs. 7, 10, 11; Nissel & Rupp 2017 eq. (17) |
+| FBMC/OQAM | A | Bouhadda et al. 2014 Table 1, Figs. 7, 10, 11; Nissel & Rupp 2017 eq. (17); Nissel, Schwarz & Rupp 2017 eqs. (13), (17), Figs. 5-8 |
 | UFMC | A | Vakilian et al. 2013 Fig. 3(b) |
 | OTFS and the delay–Doppler channel | A | Raviteja et al. 2019 Theorem 1 and its example; Hadani et al. 2017 Fig. 5 qualitatively |
 
@@ -228,6 +231,7 @@ Summary: 71 modules: 66 at A, 3 at B, 2 at C.
 | Capacity and precoding | A | Telatar 1999 Theorem 2, eqs. (13), (16)–(18), Figs. 1–4 |
 | MIMO channel estimation | A | Biguesh & Gershman 2006 Figs. 1–4, eqs. (12)–(59) |
 | Multi-user massive MIMO | A | Ngo, Larsson & Marzetta 2013 Props. 1–3, 5–7, Figs. 1 and 4 |
+| Downlink NOMA | A | Ding et al. 2014 Fig. 3 table, Fig. 1(a) |
 
 **Wireless channels**
 
@@ -243,7 +247,7 @@ Summary: 71 modules: 66 at A, 3 at B, 2 at C.
 
 | Module | Level | Strongest evidence |
 |---|---|---|
-| Fibre: linear, split-step Fourier, Manakov, back-propagation | A | OptiCommPy outputs; Häger & Pfister 2021 Fig. 5; Ip & Kahn 2008 Fig. 12 (in part) |
+| Fibre: linear, split-step Fourier, Manakov, back-propagation | A | OptiCommPy outputs; Häger & Pfister 2021 Figs. 5 and 8; Ip & Kahn 2008 Fig. 12 (in part) |
 | GN and EGN models | A | Carena et al. 2014 Fig. 1(a); GNPy outputs; Semrau, Killey & Bayvel 2019 Figs. 4a, 4b, 5 |
 | Raman amplification and ISRS | A | Bromage 2004 eqs. (3)–(5), Figs. 7–8; Hu et al. 2004 eq. (2); GNPy outputs; Semrau & Bayvel 2018 eq. (10) |
 | EDFA and PMD | A | the Maxwellian DGD of Foschini & Poole 1991 Fig. 3; the accumulated ASE |

{comnum-0.2.0/crates/comnum → comnum-0.3.1}/README.md

@@ -16,8 +16,8 @@ the literature.
 <!-- highlights: generated by tools/highlights.py, do not edit -->
 - **Lightweight**: one run-time dependency in Python (numpy), 9 direct crates in Rust and no C, C++ or Fortran library to build; 27 MB installed with its dependencies (the benchmark's environment: komm 22 MB, OptiCommPy 321 MB, Sionna 1004 MB).
 - **Fast**: compiled Rust, Monte Carlo runs on every core; on one CPU thread, the links of the benchmark take 2.1 to 33 times as long with komm, 2.3 to 4.7 times as long with OptiCommPy and 1.1 to 7.4 times as long with Sionna as with comnum.
-- **Safe**: `unsafe` code forbidden in every crate, invalid inputs are errors (one error type), results independent of the thread count; 717 unit tests, 187 validation tests, 272 Python tests.
-- **Validated**: compared with 130 papers, standards and reference tools (82 fully reproduced, 44 partly), 100 inconsistencies found in them; 93 validation cards rerun a comparison without the paper.
+- **Safe**: `unsafe` code forbidden in every crate, invalid inputs are errors (one error type), results independent of the thread count; 749 unit tests, 198 validation tests, 272 Python tests.
+- **Validated**: compared with 139 papers, standards and reference tools (88 fully reproduced, 47 partly), 120 inconsistencies found in them; 104 validation cards rerun a comparison without the paper.
 
 The numbers are recomputed by `tools/highlights.py` from the recorded benchmark and the validation files; the details are on the page [Why comnum](https://com-rs.github.io/comnum/why.html).
 <!-- highlights:end -->
@@ -144,7 +144,7 @@ gives the tests behind every grade and the regimes not covered).
 
 <!-- generated by tools/maturity_table.py from validation/maturity.yaml: do not edit -->
 
-Summary: 71 modules: 66 at A, 3 at B, 2 at C.
+Summary: 75 modules: 70 at A, 3 at B, 2 at C.
 
 **Foundations**
 
@@ -152,6 +152,7 @@ Summary: 71 modules: 66 at A, 3 at B, 2 at C.
 |---|---|---|
 | Constellations, mapping, demapping and closed-form error rates | A | OptiCommPy outputs; closed forms of Proakis & Salehi |
 | Capacity and cutoff rate of CM and BICM | A | Caire, Taricco & Biglieri 1998 Figs. 4–7 |
+| Adaptive MQAM over fading | A | Goldsmith & Chua 1997 Figs. 3, 5, 10, Table II |
 | Sources (bits, PRBS, uniform and shaped symbols) | A | ITU-T O.150 Sec. 5 and Table 1, exact |
 | Differential PSK | A | Pawula, Rice & Roberts 1982 Sec. IV-A; closed forms of Proakis & Salehi |
 | Error rates with an imperfect channel estimate and over Jakes' fading | A | Chang & Su 2002 eqs. 31–53, Figs. 3–4 |
@@ -164,6 +165,8 @@ Summary: 71 modules: 66 at A, 3 at B, 2 at C.
 | Puncturing | A | Yasuda et al. Table I; patterns of IEEE 802.11 and EN 300 421 |
 | Turbo codes (LTE) | A | AFF3CT, K = 40 and 1024 |
 | LDPC codes | A | AFF3CT, PEG and DVB-S2 short code; Hu et al. 2005 |
+| Density evolution of LDPC ensembles | A | Richardson & Urbanke 2001 Tables I-II |
+| EXIT charts of parallel concatenated codes | A | ten Brink 2001 Tables II and IV |
 | Polar codes (SC, CA-SCL) | A | AFF3CT, SC and CA-SCL |
 | 5G NR channel codes: LDPC and polar with rate matching | A | Sionna vectors bit for bit; TS 38.212 tables |
 | Reed–Solomon codes | A | AFF3CT; Lin & Costello; ETSI EN 300 744, CCSDS 131.0-B |
@@ -183,7 +186,7 @@ Summary: 71 modules: 66 at A, 3 at B, 2 at C.
 | Constant-envelope OFDM | A | Thompson et al. 2008 eq. (15); Guerreiro et al. 2016 Fig. 1; López Morales et al. 2022 Fig. 3; Pacheco & Hatzinakos 2006 Fig. 1 |
 | Chirp spread spectrum (LoRa modulation) | A | Chiani & Elzanaty 2019 Table I; Afisiadis et al. 2020 Figs. 2, 7; Vangelista 2017 Fig. 1 |
 | Continuous-phase modulation | A | Murota & Hirade 1981 Table I and Fig. 5; Sundberg 1986 Fig. 5; MSK of Proakis & Salehi |
-| FBMC/OQAM | A | Bouhadda et al. 2014 Table 1, Figs. 7, 10, 11; Nissel & Rupp 2017 eq. (17) |
+| FBMC/OQAM | A | Bouhadda et al. 2014 Table 1, Figs. 7, 10, 11; Nissel & Rupp 2017 eq. (17); Nissel, Schwarz & Rupp 2017 eqs. (13), (17), Figs. 5-8 |
 | UFMC | A | Vakilian et al. 2013 Fig. 3(b) |
 | OTFS and the delay–Doppler channel | A | Raviteja et al. 2019 Theorem 1 and its example; Hadani et al. 2017 Fig. 5 qualitatively |
 
@@ -197,6 +200,7 @@ Summary: 71 modules: 66 at A, 3 at B, 2 at C.
 | Capacity and precoding | A | Telatar 1999 Theorem 2, eqs. (13), (16)–(18), Figs. 1–4 |
 | MIMO channel estimation | A | Biguesh & Gershman 2006 Figs. 1–4, eqs. (12)–(59) |
 | Multi-user massive MIMO | A | Ngo, Larsson & Marzetta 2013 Props. 1–3, 5–7, Figs. 1 and 4 |
+| Downlink NOMA | A | Ding et al. 2014 Fig. 3 table, Fig. 1(a) |
 
 **Wireless channels**
 
@@ -212,7 +216,7 @@ Summary: 71 modules: 66 at A, 3 at B, 2 at C.
 
 | Module | Level | Strongest evidence |
 |---|---|---|
-| Fibre: linear, split-step Fourier, Manakov, back-propagation | A | OptiCommPy outputs; Häger & Pfister 2021 Fig. 5; Ip & Kahn 2008 Fig. 12 (in part) |
+| Fibre: linear, split-step Fourier, Manakov, back-propagation | A | OptiCommPy outputs; Häger & Pfister 2021 Figs. 5 and 8; Ip & Kahn 2008 Fig. 12 (in part) |
 | GN and EGN models | A | Carena et al. 2014 Fig. 1(a); GNPy outputs; Semrau, Killey & Bayvel 2019 Figs. 4a, 4b, 5 |
 | Raman amplification and ISRS | A | Bromage 2004 eqs. (3)–(5), Figs. 7–8; Hu et al. 2004 eq. (2); GNPy outputs; Semrau & Bayvel 2018 eq. (10) |
 | EDFA and PMD | A | the Maxwellian DGD of Foschini & Poole 1991 Fig. 3; the accumulated ASE |

{comnum-0.2.0 → comnum-0.3.1}/bindings/comnum-py/src/ais.rs

@@ -134,10 +134,10 @@ fn message_dict<'py>(py: Python<'py>, m: &Message) -> PyResult<Bound<'py, PyDict
 ///
 /// Examples:
 ///     >>> import comnum
-///     >>> bits = comnum.ais_position_report(227000001, lat_deg=48.38, lon_deg=-4.49, sog_knots=8.0)
+///     >>> bits = comnum.ais_position_report(200000001, lat_deg=46.12, lon_deg=-9.88, sog_knots=8.0)
 ///     >>> m = comnum.ais_decode(bits)
 ///     >>> m["msg_type"], m["mmsi"], round(m["position"]["lat_deg"], 2)
-///     (1, 227000001, 48.38)
+///     (1, 200000001, 46.12)
 #[pyfunction]
 fn ais_decode<'py>(py: Python<'py>, bits: &Bound<'py, PyAny>) -> PyResult<Bound<'py, PyDict>> {
     message_dict(py, &message::decode(&bvec(bits)?).py()?)
@@ -266,10 +266,10 @@ fn ais_transmit<'py>(py: Python<'py>, messages: Vec<Bound<'py, PyAny>>, fs: f64)
 /// Examples:
 ///     >>> import comnum
 ///     >>> fs = comnum.ais_front_end_rate()
-///     >>> bits = comnum.ais_position_report(227000001, lat_deg=48.38, lon_deg=-4.49)
+///     >>> bits = comnum.ais_position_report(200000001, lat_deg=46.12, lon_deg=-9.88)
 ///     >>> x = comnum.ais_transmit([bits], fs)
 ///     >>> [p["message"]["mmsi"] for p in comnum.ais_receive(x, fs)]
-///     [227000001]
+///     [200000001]
 #[pyfunction]
 fn ais_receive<'py>(py: Python<'py>, x: &Bound<'py, PyAny>, fs: f64) -> PyResult<Bound<'py, PyList>> {
     let v = cvec(x)?;

{comnum-0.2.0 → comnum-0.3.1}/bindings/comnum-py/src/fiber.rs

@@ -10,7 +10,8 @@ use comnum::channel::optical::manakov::ManakovSsfm;
 use comnum::channel::optical::pmd::{ManakovPmd, PmdLink};
 use comnum::channel::optical::raman::Raman;
 use comnum::core::signal::DualPol;
-use comnum::math::nlse::{NonlinearStep, StepRule, StepUpdate};
+use comnum::dsp::compensation::dbp::{Dbp, NonlinearLength};
+use comnum::math::nlse::{NonlinearStep, StepDistribution, StepRule, StepUpdate};
 use comnum::spec::blocks::BuildContext;
 use pyo3::exceptions::PyValueError;
 use pyo3::prelude::*;
@@ -325,6 +326,10 @@ pub(crate) fn raman(py: Python<'_>, v: Option<&Bound<'_, PyAny>>) -> PyResult<Op
 /// Args:
 ///     fiber (FiberParams): the link.
 ///     step_km (float): step size `h` (km).
+///     steps_per_span (int, optional): this number of logarithmic steps per span
+///         instead of `step_km` (Bosco et al., IEEE PTL 12(5), 2000, eq. (6)): every step
+///         carries the same integrated power, short at the start of the span; double
+///         precision only.
 ///     amplification (str): `"edfa"`, `"ideal"` or `"none"`.
 ///     raman (dict, optional): Raman pumps and ISRS in every span, the keys of
 ///         ``fiber.raman`` in a specification (see :func:`raman_profile`); the amplifier
@@ -349,32 +354,39 @@ impl PySsfm {
     fn __repr__(&self) -> String {
         let s = &self.inner;
         format!(
-            "Ssfm(n_spans={}, span_km={}, step_km={}, amplification='{}', raman={})",
+            "Ssfm(n_spans={}, span_km={}, {}, amplification='{}', raman={})",
             s.fiber.n_spans,
             s.fiber.span_km,
-            s.step_km,
+            match s.rule {
+                Some(StepRule::Logarithmic(n)) => format!("steps_per_span={n}"),
+                _ => format!("step_km={}", s.step_km),
+            },
             format!("{:?}", s.amplification).to_lowercase(),
             py_bool(s.raman.is_some())
         )
     }
 
     #[new]
-    #[pyo3(signature = (fiber, *, step_km = 1.0, amplification = "edfa", raman = None, check_sampling = true))]
+    #[pyo3(signature = (fiber, *, step_km = 1.0, steps_per_span = None, amplification = "edfa", raman = None, check_sampling = true))]
     fn new(
         py: Python<'_>,
         fiber: &PyFiber,
         step_km: f64,
+        steps_per_span: Option<usize>,
         amplification: &str,
         raman: Option<&Bound<'_, PyAny>>,
         check_sampling: bool,
     ) -> PyResult<Self> {
+        if steps_per_span == Some(0) {
+            return Err(PyValueError::new_err("steps_per_span must be at least 1"));
+        }
         Ok(PySsfm {
             inner: Ssfm {
                 fiber: fiber.inner.clone(),
                 step_km,
                 amplification: amplification.parse::<Amplification>().py()?,
                 raman: self::raman(py, raman)?,
-                rule: None,
+                rule: steps_per_span.map(StepRule::Logarithmic),
                 check_sampling,
             },
         })
@@ -437,13 +449,23 @@ impl PySsfm {
     ///     nonlinear_scale (float): ξ, multiplies the nonlinear phase (the optimised DBP
     ///         with few steps per span, Ip & Kahn 2008; 0 gives the linear
     ///         back-propagation).
+    ///     step_distribution (str): ``"uniform"`` (equal steps), ``"logarithmic"`` (Bosco
+    ///         et al. 2000: every step carries the same integrated power of the forward
+    ///         span, applied from its end; better with 2 to 8 steps per span, worse with
+    ///         many) or ``"modified_logarithmic"`` (Zhang et al. 2013: the attenuation
+    ///         times ``step_factor``).
+    ///     step_factor (float): the adjusting factor of the modified logarithmic steps.
+    ///     nonlinear_length (str): ``"effective"``, the effective length of a coarse step
+    ///         (the power integrated over it), or ``"step"``, the step itself (the
+    ///         midpoint rule of Häger & Pfister 2021).
     ///
     /// Returns:
     ///     numpy.ndarray: the back-propagated field (complex128).
     ///
     /// Raises:
     ///     ValueError: for an invalid parameter or input (the message names it).
-    #[pyo3(signature = (x, fs, *, steps_per_span = 1, nonlinear_scale = 1.0))]
+    #[pyo3(signature = (x, fs, *, steps_per_span = 1, nonlinear_scale = 1.0, step_distribution = "uniform", step_factor = 0.4, nonlinear_length = "effective"))]
+    #[allow(clippy::too_many_arguments, reason = "the settings of the back-propagation, by keyword")]
     fn backpropagate<'py>(
         &self,
         py: Python<'py>,
@@ -451,10 +473,21 @@ impl PySsfm {
         fs: f64,
         steps_per_span: usize,
         nonlinear_scale: f64,
+        step_distribution: &str,
+        step_factor: f64,
+        nonlinear_length: &str,
     ) -> PyResult<CArr<'py>> {
         let x = cvec(x)?;
-        let s = &self.inner;
-        Ok(carr(py, py.detach(|| s.backpropagate_scaled(&x, fs, steps_per_span, nonlinear_scale)).py()?))
+        let distribution = match step_distribution.parse::<StepDistribution>().py()? {
+            StepDistribution::ModifiedLogarithmic(_) => StepDistribution::ModifiedLogarithmic(step_factor),
+            d => d,
+        };
+        let dbp = Dbp {
+            distribution,
+            nonlinear_length: nonlinear_length.parse::<NonlinearLength>().py()?,
+            ..self.inner.dbp(steps_per_span, nonlinear_scale)
+        };
+        Ok(carr(py, py.detach(|| dbp.run(&x, fs)).py()?))
     }
 }
 
@@ -590,8 +623,11 @@ impl PyManakov {
 ///     step_rule (str): ``"fwm"``, the rule of Musetti et al. (2018): the first step of a
 ///         span bounds the four-wave-mixing phase across the signal bandwidth to
 ///         ``fwm_phase``, the next ones grow with the loss (about 0.01 dB of error on the
-///         SNR at 20 rad, whatever the power); or ``"nonlinear_phase"``, OptiCommPy's
-///         ``max_nl_phase`` per step.
+///         SNR at 20 rad, whatever the power); ``"nonlinear_phase"``, OptiCommPy's
+///         ``max_nl_phase`` per step; or ``"logarithmic"``, ``steps_per_span``
+///         logarithmic steps (Bosco et al. 2000).
+///     steps_per_span (int): logarithmic steps per span, with
+///         ``step_rule="logarithmic"``.
 ///     fwm_phase (float): the four-wave-mixing phase of the first step (rad).
 ///     bandwidth_hz (float, optional): the signal bandwidth of the rule ``"fwm"``
 ///         (measured on the launched field if None).
@@ -635,7 +671,7 @@ impl PyManakovSsfm {
     }
 
     #[new]
-    #[pyo3(signature = (fiber, *, step_rule = "fwm", fwm_phase = 20.0, bandwidth_hz = None, max_nl_phase = 2e-2, step_km = None, nonlinear_step = "midpoint", max_iterations = 10, tolerance = 1e-5, amplification = "edfa", raman = None, check_sampling = true))]
+    #[pyo3(signature = (fiber, *, step_rule = "fwm", fwm_phase = 20.0, bandwidth_hz = None, max_nl_phase = 2e-2, steps_per_span = 100, step_km = None, nonlinear_step = "midpoint", max_iterations = 10, tolerance = 1e-5, amplification = "edfa", raman = None, check_sampling = true))]
     #[allow(clippy::too_many_arguments, reason = "the settings of the solver, by keyword")]
     fn new(
         py: Python<'_>,
@@ -644,6 +680,7 @@ impl PyManakovSsfm {
         fwm_phase: f64,
         bandwidth_hz: Option<f64>,
         max_nl_phase: f64,
+        steps_per_span: usize,
         step_km: Option<f64>,
         nonlinear_step: &str,
         max_iterations: usize,
@@ -661,9 +698,10 @@ impl PyManakovSsfm {
                 max_nl_phase_rad: Some(max_nl_phase),
             },
             (None, "nonlinear_phase") => StepRule::NonlinearPhase(max_nl_phase),
+            (None, "logarithmic") => StepRule::Logarithmic(steps_per_span),
             (None, other) => {
                 return Err(PyValueError::new_err(format!(
-                    "step_rule must be 'fwm' or 'nonlinear_phase', got {other:?}"
+                    "step_rule must be 'fwm', 'nonlinear_phase' or 'logarithmic', got {other:?}"
                 )));
             }
         };

{comnum-0.2.0 → comnum-0.3.1/crates/comnum}/README.md

@@ -16,8 +16,8 @@ the literature.
 <!-- highlights: generated by tools/highlights.py, do not edit -->
 - **Lightweight**: one run-time dependency in Python (numpy), 9 direct crates in Rust and no C, C++ or Fortran library to build; 27 MB installed with its dependencies (the benchmark's environment: komm 22 MB, OptiCommPy 321 MB, Sionna 1004 MB).
 - **Fast**: compiled Rust, Monte Carlo runs on every core; on one CPU thread, the links of the benchmark take 2.1 to 33 times as long with komm, 2.3 to 4.7 times as long with OptiCommPy and 1.1 to 7.4 times as long with Sionna as with comnum.
-- **Safe**: `unsafe` code forbidden in every crate, invalid inputs are errors (one error type), results independent of the thread count; 717 unit tests, 187 validation tests, 272 Python tests.
-- **Validated**: compared with 130 papers, standards and reference tools (82 fully reproduced, 44 partly), 100 inconsistencies found in them; 93 validation cards rerun a comparison without the paper.
+- **Safe**: `unsafe` code forbidden in every crate, invalid inputs are errors (one error type), results independent of the thread count; 749 unit tests, 198 validation tests, 272 Python tests.
+- **Validated**: compared with 139 papers, standards and reference tools (88 fully reproduced, 47 partly), 120 inconsistencies found in them; 104 validation cards rerun a comparison without the paper.
 
 The numbers are recomputed by `tools/highlights.py` from the recorded benchmark and the validation files; the details are on the page [Why comnum](https://com-rs.github.io/comnum/why.html).
 <!-- highlights:end -->
@@ -144,7 +144,7 @@ gives the tests behind every grade and the regimes not covered).
 
 <!-- generated by tools/maturity_table.py from validation/maturity.yaml: do not edit -->
 
-Summary: 71 modules: 66 at A, 3 at B, 2 at C.
+Summary: 75 modules: 70 at A, 3 at B, 2 at C.
 
 **Foundations**
 
@@ -152,6 +152,7 @@ Summary: 71 modules: 66 at A, 3 at B, 2 at C.
 |---|---|---|
 | Constellations, mapping, demapping and closed-form error rates | A | OptiCommPy outputs; closed forms of Proakis & Salehi |
 | Capacity and cutoff rate of CM and BICM | A | Caire, Taricco & Biglieri 1998 Figs. 4–7 |
+| Adaptive MQAM over fading | A | Goldsmith & Chua 1997 Figs. 3, 5, 10, Table II |
 | Sources (bits, PRBS, uniform and shaped symbols) | A | ITU-T O.150 Sec. 5 and Table 1, exact |
 | Differential PSK | A | Pawula, Rice & Roberts 1982 Sec. IV-A; closed forms of Proakis & Salehi |
 | Error rates with an imperfect channel estimate and over Jakes' fading | A | Chang & Su 2002 eqs. 31–53, Figs. 3–4 |
@@ -164,6 +165,8 @@ Summary: 71 modules: 66 at A, 3 at B, 2 at C.
 | Puncturing | A | Yasuda et al. Table I; patterns of IEEE 802.11 and EN 300 421 |
 | Turbo codes (LTE) | A | AFF3CT, K = 40 and 1024 |
 | LDPC codes | A | AFF3CT, PEG and DVB-S2 short code; Hu et al. 2005 |
+| Density evolution of LDPC ensembles | A | Richardson & Urbanke 2001 Tables I-II |
+| EXIT charts of parallel concatenated codes | A | ten Brink 2001 Tables II and IV |
 | Polar codes (SC, CA-SCL) | A | AFF3CT, SC and CA-SCL |
 | 5G NR channel codes: LDPC and polar with rate matching | A | Sionna vectors bit for bit; TS 38.212 tables |
 | Reed–Solomon codes | A | AFF3CT; Lin & Costello; ETSI EN 300 744, CCSDS 131.0-B |
@@ -183,7 +186,7 @@ Summary: 71 modules: 66 at A, 3 at B, 2 at C.
 | Constant-envelope OFDM | A | Thompson et al. 2008 eq. (15); Guerreiro et al. 2016 Fig. 1; López Morales et al. 2022 Fig. 3; Pacheco & Hatzinakos 2006 Fig. 1 |
 | Chirp spread spectrum (LoRa modulation) | A | Chiani & Elzanaty 2019 Table I; Afisiadis et al. 2020 Figs. 2, 7; Vangelista 2017 Fig. 1 |
 | Continuous-phase modulation | A | Murota & Hirade 1981 Table I and Fig. 5; Sundberg 1986 Fig. 5; MSK of Proakis & Salehi |
-| FBMC/OQAM | A | Bouhadda et al. 2014 Table 1, Figs. 7, 10, 11; Nissel & Rupp 2017 eq. (17) |
+| FBMC/OQAM | A | Bouhadda et al. 2014 Table 1, Figs. 7, 10, 11; Nissel & Rupp 2017 eq. (17); Nissel, Schwarz & Rupp 2017 eqs. (13), (17), Figs. 5-8 |
 | UFMC | A | Vakilian et al. 2013 Fig. 3(b) |
 | OTFS and the delay–Doppler channel | A | Raviteja et al. 2019 Theorem 1 and its example; Hadani et al. 2017 Fig. 5 qualitatively |
 
@@ -197,6 +200,7 @@ Summary: 71 modules: 66 at A, 3 at B, 2 at C.
 | Capacity and precoding | A | Telatar 1999 Theorem 2, eqs. (13), (16)–(18), Figs. 1–4 |
 | MIMO channel estimation | A | Biguesh & Gershman 2006 Figs. 1–4, eqs. (12)–(59) |
 | Multi-user massive MIMO | A | Ngo, Larsson & Marzetta 2013 Props. 1–3, 5–7, Figs. 1 and 4 |
+| Downlink NOMA | A | Ding et al. 2014 Fig. 3 table, Fig. 1(a) |
 
 **Wireless channels**
 
@@ -212,7 +216,7 @@ Summary: 71 modules: 66 at A, 3 at B, 2 at C.
 
 | Module | Level | Strongest evidence |
 |---|---|---|
-| Fibre: linear, split-step Fourier, Manakov, back-propagation | A | OptiCommPy outputs; Häger & Pfister 2021 Fig. 5; Ip & Kahn 2008 Fig. 12 (in part) |
+| Fibre: linear, split-step Fourier, Manakov, back-propagation | A | OptiCommPy outputs; Häger & Pfister 2021 Figs. 5 and 8; Ip & Kahn 2008 Fig. 12 (in part) |
 | GN and EGN models | A | Carena et al. 2014 Fig. 1(a); GNPy outputs; Semrau, Killey & Bayvel 2019 Figs. 4a, 4b, 5 |
 | Raman amplification and ISRS | A | Bromage 2004 eqs. (3)–(5), Figs. 7–8; Hu et al. 2004 eq. (2); GNPy outputs; Semrau & Bayvel 2018 eq. (10) |
 | EDFA and PMD | A | the Maxwellian DGD of Foschini & Poole 1991 Fig. 3; the accumulated ASE |

{comnum-0.2.0 → comnum-0.3.1}/crates/comnum-channel/src/multipath/mod.rs

@@ -97,6 +97,30 @@ pub fn max_doppler_hz(speed_mps: f64, fc_hz: f64) -> f64 {
     speed_mps * fc_hz / SPEED_OF_LIGHT
 }
 
+/// The level-crossing rate (crossings per second, in one direction) of a Rayleigh fading
+/// envelope with the Clarke/Jakes spectrum of maximum Doppler shift `doppler_hz`, at the
+/// level $\rho$ = `rho` relative to its RMS value (Zheng & Xiao 2003, eq. (19a)):
+/// $$L = \sqrt{2\pi}\, f_{D}\, \rho\, e^{-\rho^{2}}.$$
+///
+/// # Examples
+/// ```
+/// use comnum_channel::multipath::rayleigh_level_crossing_rate;
+/// // the largest rate, at rho = 1/sqrt(2): sqrt(pi) e^{-1/2} f_D
+/// let l = rayleigh_level_crossing_rate(0.5f64.sqrt(), 1.0);
+/// assert!((l - std::f64::consts::PI.sqrt() * (-0.5f64).exp()).abs() < 1e-15);
+/// ```
+pub fn rayleigh_level_crossing_rate(rho: f64, doppler_hz: f64) -> f64 {
+    (2.0 * std::f64::consts::PI).sqrt() * doppler_hz * rho * (-rho * rho).exp()
+}
+
+/// The average fade duration (s) of a Rayleigh fading envelope below the level $\rho$ =
+/// `rho` relative to its RMS value, maximum Doppler shift `doppler_hz` (Zheng & Xiao 2003,
+/// eq. (19b)): the probability of the fade $1 - e^{-\rho^{2}}$ over the level-crossing rate,
+/// $$T = \frac{e^{\rho^{2}} - 1}{\rho f_{D}\sqrt{2\pi}}.$$
+pub fn rayleigh_average_fade_duration(rho: f64, doppler_hz: f64) -> f64 {
+    ((rho * rho).exp() - 1.0) / (rho * doppler_hz * (2.0 * std::f64::consts::PI).sqrt())
+}
+
 /// The Doppler shift of the specular component of a Ricean tap relative to the maximum
 /// Doppler shift, $f_{S} / f_{D} = 0.7$ (TR 38.901 §7.7.2, for TDL-D and TDL-E).
 pub const LOS_DOPPLER_RATIO: f64 = 0.7;

{comnum-0.2.0 → comnum-0.3.1}/crates/comnum-channel/src/optical/fiber.rs

@@ -375,8 +375,10 @@ pub struct Ssfm {
     pub raman: Option<Raman>,
     /// Another step rule than the fixed steps of `step_km` (OptiCommPy's): the rule of
     /// every propagator of the engine (`docs/adr/0014-one-fibre-propagator.md`), such as
-    /// `StepRule::FWM_DEFAULT`; the back-propagation and the single-precision path keep
-    /// the fixed steps.
+    /// `StepRule::FWM_DEFAULT` or the logarithmic steps of Bosco et al.,
+    /// `StepRule::Logarithmic(n)`; the back-propagation keeps its own steps
+    /// ([`Ssfm::dbp`]), and the single-precision path the fixed steps (it refuses the
+    /// logarithmic ones).
     pub rule: Option<StepRule>,
     /// Refuse a field whose sampling rate or block cannot hold the four-wave mixing and
     /// the walk-off of the link ([`SamplingCheck`]); `false` runs anyway.
@@ -479,7 +481,12 @@ impl Ssfm {
     /// of the lumped amplifiers: [`Ssfm::raman`] is ignored.
     pub fn dbp(&self, steps_per_span: usize, xi: f64) -> Dbp {
         let f = &self.fiber;
-        let forward = self.steps().ok().map(|(n, h)| Steps::new(n, h));
+        // the steps of the forward model, only when they are the fixed steps of step_km
+        // (another rule propagates with other steps: no exact inverse)
+        let forward = match self.rule {
+            None => self.steps().ok().map(|(n, h)| Steps::new(n, h)),
+            Some(_) => None,
+        };
         let mut dbp = Dbp { steps_per_span, xi, forward, ..Dbp::new(f.medium(), f.span_km, f.n_spans, 1.0) };
         dbp.span_gain = self.amplification.field_gain(f, steps_per_span as f64 * dbp.step_km());
         dbp
@@ -515,6 +522,9 @@ impl Ssfm {
     /// precision only).
     pub fn propagate_f32(&self, x: &mut Vec<C64>, fs: f64, rng: Option<&mut Rng>) -> Result<()> {
         ensure(self.raman.is_none(), || "the Raman amplification is simulated in double precision only".into())?;
+        ensure(!matches!(self.rule, Some(StepRule::Logarithmic(_))), || {
+            "the logarithmic steps are simulated in double precision only".into()
+        })?;
         let (n_steps, h) = self.steps()?;
         let n = x.len();
         if n == 0 {