A detection apparatus detecting an error component contained in two signals (A, B) approximated by a cosine and sine functions representing an object position, the detection apparatus including an arithmetic portion (3, 4) which reduces an error contained in the signals (A, B) based on an error prediction value to output two error correction signals (A*, B*), a phase arithmetic portion (5) which calculates a phase (θ) based on the error correction signals (A*, B*), an arithmetic storage unit (9, 10) which stores the error correction signals (A*, B*) and a plurality of sampling values of the phase (θ), and a Fourier transform portion (11, 12) which obtains coefficients αk, βk, γk, and δk in the following two expressions: A*=α0+α1 cos θ+β1 sin θ+ . . . +αk cos kθ+βk sin kθ B*=γ0+γ1 cos θ+δ1 sin θ+ . . . +γk cos kθ+δk sin kθ (k≧2) wherein the arithmetic portion (3, 4) updates the error prediction value using the coefficients.