Shaping laser-scanner signals by double shadow-pulse differentiation

Abstract

Scanning laser linear-dimension meters are widely used abroad in industry [i]; there have also been Soviet reports on early specimens of such systems [2]. A typical one contains an He-Ne laser, a scanning unit based on a multifaced prism and electric motor, a lens providing parallel displacement of the narrow scanning beam, and a second lens that focusses the beam after transmission through the measurement zone on the semiconductor photocell [3]. The size is determined from the pulse length provided by the shadow on intersection with the component. The shadow length is determined from a set photocurrent threshold, usually 0.5, where there is a direct proportionality between the size and the length. The 0.5 threshold provides maximum accuracy [4], and a linear amplifier has been described [5] to measure the pulse length for a fixed threshold. "However, in that system, there are stringent specifications for the accuracy of the threshold level, and any drift or fluctuation in the radiation power can cause several percent errror in the size measurement. The same effects arise from background illumination, thermal drift in the diode current, and distortion in the pulse crest due, for example, to dirt in the optics. Also, the crest is often distorted by interference fringes arising from reflections in the coherent beam at the component surfaces, which further reduces the accuracy in the threshold. The change in pulse amplitude AU arising from any of these factors results in an error in size measurement of AD = AU/U.d/2, where d is the beam diameter, which is proportional to the edge length in the pulse. For example, with AU/U = 0.i and d = 0.15 mm, the error in measuring the size is ~D = 7.5 #m. Better immunity is provided by a device for recording the edges, which is based on double differentiation of the pulse and isolating the signals for the first derivative dU/dt (Fig. ib) and the second one d2U/dt 2 (Fig. ic) [6]. Here the edge is identified from the point at which the second-derivative signal passes through zero (here the first derivative should exceed the level shown by the dashed line in Fig. Ib). That provides exact coordinate determination no matter what the pulse amplitude and eliminates effects from laser power fluctuations and other forms of interference whose spectra are narrower than the shadow pulse one. Here specifications for the detector components are formulated and tests and an implementation of the double differentiation method are described. From [5] calculations, the edge length in a scanning laser device is about tf = 2 ~sec, and the passband for the photocell system should be fl = i/tf = 0.5 MHz. To provide the maximum signal-noise ratio, the photodiode should be used in a photogalvanic circuit [7], while the amplifier should work in large-signal mode (U ~ 5 V) and provide a rise rate not less than v = 5 V/#sec. Large-signal operation and stage decoupling are provided by operational amplifiers OA. An FDK-142 fast silicon photodiode (~ = 80 nsec) was used, which provided a sensitivity S = i0 mA/im and a low noise level. It in combination with a KR544UD2A operational amplifier connected with negative current feedback provided a noise level in a i MHz passband less by a factor of 1.5-10 by comparison with FD-256 photodiodes (factor of 1.5), FD-7G (factor of 4), FD-I (factor of 8), and FD-24K (factor of 6-10). The diameter of the sensitive area is D = 15 mm, which simplifies the optical design. Tests with fast OA of various types (KR544UD2, KR574UDI, and KR754UD2) showed that in in large-signal mode (5-10 V), the best shapes were obtained for the leading and trailing edges for a pulse corresponding to a Laplace function [4] when the photodiode was used with a KR544UD2. The waveform showed no oscillation due to shock excitation of the OA, and the edges were symmetrical. The maximum passband was attained with weak illumination, flux power not more than 30 #W, in which the output current from the photodiode did not exceed 50 #A.