A techn high-performance servo using the 6-bit channel. The use of Pkwy, Broomfield, CO 80021 que is described which implements burst amplitude demodulation analog-to-digital converter in a digital read digital circuits reduces the size of the channel chip as well al ability compared to the digital area detect standard analog tal area detect performance very close to optimal it includes a servo eneration channels ns in analog cirduplicates functions latest generation of cessing [I]. These annel chips include servo funcIs use a higher percentage of digital on of circuits used for servo and mability in the serv have been used to demodulate g demodulators. Most also use t the demodulator and have an analog circuits to tation.reducing the design time and varianalog circuits. The performance of burst demodulator is compared to techniques. Simulation results show digibetter than peak detect and for a practical burst waveform.the servo burst signal during the readback process. Simulation results are also used to compare the performance of this demodulator to Khat of the optimal amplitude estimator and to that of analog area and peak detectors.The optimal servo burst amplitude demodulator is presented in Section I1 along with its theoretical performance and discussion of practical implementations. In Section I11 a digital burst demodulator is described which gives performance close to the optimal. Simulation results are given in Section IV and conclusions are drawn in Section V,
BURST AMPLITUDE DEMODULATIONFor the typicall servo techniques which use the amplitude of periodic burst signals to measure position error, the servo demodulation process reduces to a problem of amplitude estimation. The optimal estimator of the amplitude of a known signal in white Gaussian noise, r(t)=As(t)+n(t),is well known and is implemented as a correlation scaled appropriately [3]. The variance of' the ML estimator of the amplitude, A, is given bywhere No is the two-sided power spectral density of the additive white Gaussian noise, n(t), and E, is the energy of the underlying burst waveform, s(t).In the present discussion let us consider a burst signal consisting of Lorentzian pulses represented bywhere T is the burst period, p(t) = P (I+(t/W)2)-1 with W equal to half the 50% pulse width, and n(t) is white Gaussian noise with power spectral density NO.For this case the signal energy derived in [4] can be used with (1) to show that the optimal amplitude estimator performance is given by where 7 is the duration of the burst.