Variance Fluctuations of 1/F Noise High-Radiance Small-Area Gallium-Indium-Arsenide 1 06 /im Light - Emitting Diodes
Variance Fluctuations of 1/F Noise High-Radiance Small-Area Gallium-Indium-Arsenide 1 06 /im Light - Emitting Diodes
Variance Fluctuations of 1/F Noise High-Radiance Small-Area Gallium-Indium-Arsenide 1 06 /im Light - Emitting Diodes
Indexing terms: Random noise, Thermal noise The difference in behaviour between the variances of 1 / / ncise and of thermal noise is related to the differences in the ratios of fourth moment to the second moment of the distributions. This ratio is found experimentally to be significantly different in the two cases.
Several authors have shown that the variance of 1//noise is not a constant quantity, but rather is a stochastic quantity subject to fluctuations,12 but, at the same time, it is commonly assumed that the amplitude distribution of 1//noise is Gaussian. The authors are indebted to Prof. T. Lewis for pointing out that the variance of the variance must be a function of the fourth moment of the distribution, and, if a quantity x has a Gaussian distribution, <x4) = 3 x 2 2 .
\ noise source
<
An experiment was therefore set up on the lines shown in Fig. 1. Noise from a chosen source is amplified (with a passband of 1 to 10 kHz) and fed to both inputs of a multiplier to produce x2, and this is similarly applied to another multiplier to produce x 4 . The average values of these two quantities (integrator time constant RC was 20 s, approximately) are sampled simultaneously at intervals of 826 ms and fed alternately to a digital voltmeter; <x4> was fed directly to the d.v.m. and <*2> via a hold circuit, so that the two samples were simultaneous, although they were processed consecutively. The values are punched onto paper tape and a computer programmed to calculate (a) averages, (b) ratio of <x4> to * 2 2 and (c) the percentage by which (6) differs from 3 0. Sets of between 200 and 300 readings were taken every 15 min, so that the behaviour of each resistor was monitored over a period of rather more than 1 h. Preliminary results show that Johnson noise produces the expected ratio within 2 %, but noise due to the passage of current through a carbon resistor, a sample of p silicon of 25 ficm and a sample of n silicon of 4ficm, showed ratios in excess of 3 by 20 to 100%. This suggests that in 1 //noise there is a small, but significant, number of large spikes in excess of the number predicted by the Gaussian formula; the number is small enough to be unnoticed in simple methods of distribution analysis, but large enough to have a marked effect on the ratio of fourth moment to second moment.
D. A. BELL S. P. B. DISSANAYAKE
At present, there is a considerable amount of effort being devoted to the development of components for fibre-optic communication systems. These systems are normally based on the generation and detection of a modulated optical signal with a wavelength of around 0-9//m. However, there are distinct advantages in increasing the working wavelength to some value greater than .1 //m, because of the improved transmission1'2 and dispersion3 characteristics of the fibre at these longer wavelengths. The exploitation of these improved fibre characteristics offers the possibility of systems using increased repeater spacings and wider bandwidths than in systems based on 0-9 fim radiation. There is also an additional advantage of working at longer wavelengths in that the optical attenuation of the fibres has been shown to be less susceptible4 to radiation damage for working wavelengths greater than ] fim than for wavelengths around 0-9 /im or shorter. It is also becoming apparent that, for moderate bandwidth systems, light-emitting-diode sources are superior to c.w. lasers in device life.5 To couple as much light as possible into the fibre, it is necessary to make highradiance-device structures, such structures having been made in GaAs by Burrus6 and Goodfellow.5 This letter reports the fabrication of high-radiance high-speed gallium-indiumarsenide (GalnAs) devices emitting at a wavelength of 1 06 /um. These devices are shown to have radiances comparable with commercially7 available GaAs high-radiance devices, and, in addition, are capable of modulation at speeds in excess of 250 MHz, significantly faster than attainable so far in zincdoped GaAs-based structures. The GalnAs layers were grown on {100} GaAs substrates by a vapour-phase hydride epitaxy process, as described in Reference.8 The p-n junctions were 'grown in' by switching from /7-type, sulphur, to /?-type, zinc, dopant during layer growth. Typically, the layers used in the fabrication of highradiance structures yielded mesa-structure emitters8 with external 300K quantum efficiencies of between 1 and 2% for \06fitn emission. The high-radiance-device structure is shown in Fig. 1 and is the same as that used by Goodfellow.9 A gold integral heatsink was electroplated onto the p side to ease alloy-bonding problems, and, as the {100} substrate orientation was used, it was possible to separate the device chips by scribing and cleaving. The active area was localised by a small-area p contact made through a 50/im-diameter window etched in the SiO2 layer. A well was etched into the /j-type GaAs substrate to permit a multimode optical fibre to be located close to the emitting area for improved coupling. The chip was alloyed to a copper stud S4 package and wire bonds were made to the upper n contact by a silver-loaded epoxy resin.
Electronic Engineering Department University of Hull Hull HU6 7RX, England References
1 BROPHY, j . j . : / . Appl. Phys., 1970,41, pp. 2913-2916 2 PURCELL, w. E.: ibid., 1972, 43, pp. 2890-2895 SiO2 isolation layer electroplated gold pad AuSnaltoy bond
n-type GaAs substrate GalnAs n-type 15pm graded* 15 pm constant composition '6um p-type GalnAs copper stud
Fig. 1
274
High-radiance-device
ELECTRONICS LETTERS