ALBERT

Notes: The results of extrapolation procedures for the determination of structural stresses are often questionable due to the fact that the stresses at extrapolation points obtained with finite element analyses can be strongly dependent on the mesh size of finite element model and loading mode. Also, existing design S–N curves are derived mostly on the basis of fatigue testing of joints loaded axially. In the present paper the influence of the finite element mesh size on the structural stress value determined by a linear extrapolation method is analysed. Also, the paper examines the possibility of using existing design S–N curves for cases of bending induced by a force on the welded stiffener. Fatigue test results from aluminium welded components with longitudinal or round pad stiffeners subjected to bending loads have been assessed using a structural stress range approach, and compared with the structural stress design S–N curve FAT 40 (IIW) and the structural stress design S–N curve FAT 44 (Eurocode 9). It is concluded that the more precise estimation of fatigue life of aluminium components subjected to bending can be achieved with structural stress design S–N curve proposed by Eurocode 9. The conclusions also include recommendations for regarding component finite element modelling for the determination of structural stresses in case of bending.

Notes: Summary 1. The butterfly retina exhibits strong interactions between photoreceptor responses recorded intracellularly. In general, the receptor which is locally giving the largest response suppresses the responses of neighbouring receptors that are less strongly stimulated. The effect enhances the differences between the primary photo-receptors and reduces responses to stimuli that excite all receptors together. 2. The interaction is explained in terms of a high extracellular resistance, so that receptor currents pass through other receptors in the opposite direction to those of their own responses. The result is that receptors are actively turned off by colours away from their own peak wavelength. 3. This effect applies most strongly to colour and to polarization plane when the stimulus is a point source on axis, and is therefore strong between the receptors of the same ommatidium. 4. The result is that spectral sensitivity peaks and angular sensitivity peaks are narrower, and polarisation sensitivity is greater, than expected from single retinula cells in isolation. The sensitivities measured electrophysiologically cannot be easily related to the physical properties of the visual pigments. Polarisation sensitivity (PS) can reach 50. 5. There are four types of primary photoreceptor, with peak near 380 nm, 450 nm, 550 nm and 610 nm. Cell marking usually reveals these as single retinula cells. Near the peak spectral sensitivity the responses are up to 60 mV positive-going, but away from the peak they can be negative-going. 6. Anatomically the retina ofPapilio has four distal, four proximal retinula cells, and a ninth basal cell. Narrow pigment cells and tracheoles squeeze through the substantial basement membrane along with each bundle of nine axons. 7. Two of the distal retinula cells contain red pigment grains near the rhabdom. The distal retinula cells are UV or blue sensitive. Green sensitive cells are proximal and can be coupled in opposite pairs. Red sensitive cells are proximal. 8. The UV sensitive cell with peak near 380 nm is the most sensitive of the cell types when measured by the position of theV/logI curve on the intensity axis at the spectral peak of each type. The red-sensitive cells are also sensitive. By its inhibitory effect, interaction between receptors reduces the sensitivity measurement on this scale. 9. Angular sensitivities measured with positive-going responses near the spectral peaks are narrow (Δρ-2°); when measured with negative-going responses they are wider (Δρ=3° to 5°). 10. One type of unit has only negative-going responses to −60 mV, with Δp=2° to 5°, spectral peak near 550 nm and sometimes also 380 nm or 450 nm. This type has not been marked and is regarded as a restricted channel for return current. ItsV/logI curve extends over an intensity range of 106. 11. The variety of the units suggests that their responses are not due to a simple regular network with all units connected indiscriminately to all others at all times through their terminals. There are selective channels for current flow and some retinula cells appear to be little influenced by others. 12. Theory shows that when there is a direct electrical coupling between a pair of retinula cells (not passing through the extracellular space) it is possible to balance out the negative interactions caused by current flow through their terminals. Far from degenerating the signals, direct electrical coupling can cancel the negative interaction, and this may be its normal function.

Notes: Summary 1. Intracellular recordings from the photoreceptors of the locustsLocusta migratoria andValanga irregularis and the flyCalliphora stygia, were used to examine the differences between theV/logI curve (intensity/response function) determined in the dark adapted and the light adapted state. 2. For response amplitudes less than half the maximum the fully light adaptedV/logI curve follows the simple self-shunting model, where the number of channels activated is proportional to the number of photons absorbed. 3. We fitted ourV/logI curves with the commonly employed hyperbolic function Eq. (1), and found a consistent deviation of the experimental data from the predicted curves when responses exceeded half maximal amplitude. 4. The light adaptedV/logI curves of locusts exhibit a ‘kink’ deformation (Fig. 3) and 4 of the 80 cells recorded from locusts had a small spike superimposed on the rising phase of the photo-response (Fig. 1). No kink or spike was observed in the flyCalliphora. 5. Transmembrane recording with a double electrode eliminated the ERG as a candidate for changing the slope and producing the kink or the spike. Stimulation of single ommatidia also ruled out interommatidial interactions. 6. Change in pulse duration did not affect the slope or the shape of theV/logI curves.