`torchsky.log_lambda`¶

Stability: core. Names on this page follow the deprecation policy in contributing.

Unit-integral Gaussian convolution on a uniformly spaced ln λ grid.

This reference is generated from the package source. Names are grouped by the module that defines them; import every name from the public namespace torchsky.log_lambda unless the group text says otherwise.

Top-level namespace¶

Unit-integral Gaussian convolution on a uniformly spaced ln λ grid.

Used for LOSVD broadening and published resolving-power LSFs. The kernel is normalized to sum to one. sigma → 0 is the identity on the sampled grid (an honest sub-pixel statement, not a continuous Dirac identity).

C_KM_S `module-attribute` ¶

C_KM_S = 299792.458

convolve_log_lambda ¶

convolve_log_lambda(
    flux: Tensor,
    dln_lambda: float | Tensor | None = None,
    sigma_kms: Tensor | float | None = None,
    *,
    sigma_px: Tensor | float | None = None,
    kernel_radius: int = 8,
    mean_kms: Tensor | float = 0.0,
    zero_mean: bool = False,
) -> Tensor

Convolve flux with a unit-sum Gaussian in ln λ.

Parameters:

Name	Type	Description	Default
`flux`	`Tensor`	`(..., n)` samples on a uniform `ln λ` grid.	required
`dln_lambda`	`float \| Tensor \| None`	constant grid spacing `Δ ln λ` (required when `sigma_kms` or nonzero `mean_kms` is used).	`None`
`sigma_kms`	`Tensor \| float \| None`	Gaussian width in km/s, broadcastable to `flux[..., 1]` or per-pixel `flux[..., n]`. Mutually exclusive with `sigma_px`.	`None`
`sigma_px`	`Tensor \| float \| None`	Gaussian width in grid pixels, broadcastable to `flux[..., 1]` or per-pixel `flux[..., n]`. Mutually exclusive with `sigma_kms`.	`None`
`kernel_radius`	`int`	half-width of the discrete kernel in pixels.	`8`
`mean_kms`	`Tensor \| float`	optional kernel mean in km/s (LOSVD streaming), broadcastable to `flux[..., 1]` or `flux[..., n]`. The adjoint of a mean-`μ` kernel is the mean-`-μ` kernel.	`0.0`
`zero_mean`	`bool`	if True, rebuild the kernel so the discrete first moment is zero (truncated-Gaussian centering). Incompatible with a nonzero `mean_kms`.	`False`

sigma_R_kms ¶

sigma_R_kms(resolving_power: Tensor | float) -> Tensor

Velocity width of a Gaussian LSF for published resolving power R.

σ = c / (2.355 R) so the FWHM is c / R.