Default parameterization using theta = r_s/D_a instead of H_0, and tau instead of z_re and log(A_s) instead of A_s Less general, but should give better performance The well-determined parameter A_s exp(-2tau) should be found by the covariance matrix parameter 3 is 100*theta, parameter 4 is tau, others same as params_H except A->log(A) Theta is much better constrained than H_0 Also a background-only parameterization, e.g. for use with just supernoave etc