pulp::signal::chardelay::detail Namespace Reference

Typedefs
using	LossSearchVector = std::array<double, kLossSearchDimension>

Functions
template<std::size_t Poles, std::size_t Shelves, typename Target>
LossShape< Poles, Shelves >	fit_loss_shape (Target &&target)
	Generic fit of Poles one-poles and Shelves first-order shelves to a dimensionless target, minimizing the WORST error in dB over log-spaced points.

Typedef Documentation

LossSearchVector

using pulp::signal::chardelay::detail::LossSearchVector = std::array<double, kLossSearchDimension>

Function Documentation

fit_loss_shape()

template<std::size_t Poles, std::size_t Shelves, typename Target>

LossShape< Poles, Shelves > pulp::signal::chardelay::detail::fit_loss_shape ( Target && target )

inline

Generic fit of Poles one-poles and Shelves first-order shelves to a dimensionless target, minimizing the WORST error in dB over log-spaced points.

Minimax rather than least squares because the acceptance criterion is a bound on the worst error at any frequency, and an L2 fit happily trades one large error for many small ones — exactly the wrong bargain. (A tiny L2 term breaks ties between parameter sets sharing a worst point.)