Stimulators for visual electrophysiology

Many different light sources can be used for stimulating the visual system (with varying degrees of success). Here, we will concentrate on those in common use in visual electrophysiology at the time of writing.

Regardless of the source of stimulus energy, the peak intensity or brightness, contrast, spectral content or color, duration and enveloped shape, repetition rate, field or image size, and element size will all affect the recorded response to varying degrees and must be appropriately calibrated. The ISCEV produces guidelines for stimulus calibration, and these form a good starting point for the electrophysiologist who wishes to characterize a stimulus.²

Stimuli can be subdivided into two broad categories: unstructured or spatially structured. The Ganzfeld stimulator that is routinely used in the majority of laboratories for electroretinography is an example of an unstructured stimulus, to be contrasted with any device that displays a pattern. These two categories can then be further subdivided on the basis of light source or technology employed.

A Ganzfeld stimulator is simply an approximation to an integrating sphere, the object of which is to illuminate the entire retina evenly, a task that cannot be achieved by having the patient view a lamp directly. Other techniques are sometimes employed to provide an approximation to uniform retinal illumination. These include 100-diopter lenses mounted on a corneal contact lens electrode, diffusing screens or contact lenses, and half a table tennis ball held in front of the eye and back-illuminated.⁴

The Ganzfeld stimulators (figure 19.1) generally consist of a bowl of around 500-mm diameter, with ports for the injection of light from whatever source is to be utilized, commonly a xenon flash lamp for the stimulus and an incandescent bulb for background illumination. There is another, larger aperture, the exit port, for the patient's head, which is positioned in such a way that the entire visual field is occupied by the inner surface of the bowl. The bowl surface is coated with a high-reflectance white matte paint. In a true integrating sphere, light enters the bowl through one port and is subject to multiple reflections (>200) before exiting the (small) output port, thus smoothing out any irregularities in the illumination. The clinical Ganzfeld stimulator exit port must accommodate the head of a patient and is therefore as large as 300 by 250mm for a typical adult. This significantly reduces the efficiency of a 500-mm sphere. If the sphere is made much bigger to improve the efficiency, more light is required to illuminate the interior. The dimensions given are a practical trade-off between cost and performance.

Figure 19.1.  

The Ganzfeld stimulator.

Ganzfelds for clinical use are generally equipped with filters for modifying the color and intensity of the stimuli, fixation targets and a chin rest to maintain patient position, and closed-circuit infrared video cameras to monitor the patient during the test.

The majority of Ganzfeld stimulators are powered by xenon flash lamps, such as the ubiquitous Grass PS22, although in both custom-built and commercial systems, other sources are used—mainly light-emitting diodes (LEDs) or incandescent lamps and shutter systems.

Xenon flash tubes are essentially arc lamps. They consist of a glass tube, which may be either straight, U-shaped, or coiled. An anode and a cathode (the larger electrode) are placed through the ends of the tube, and a trigger electrode is mounted on the outside of the tube between the main electrodes. The tube is filled with xenon gas at a pressure of several atmospheres.

In operation, energy is stored on a capacitor at high d.c. voltage (several hundred volts), connected in parallel with the tube. To initiate the flash, a pulse of typically 10–20kV is applied to the trigger electrode from a step-up transformer. This causes the xenon gas to ionize and thus become electrically conductive, and the energy stored in the capacitor is rapidly discharged in the form of an arc through the gas, producing a brief, high-intensity flash of light. The wavelength composition of the light depends on the gas and on its pressure. Typically, the light is of a broad spectrum, containing significant quantities of both ultraviolet (UV) and infrared (IR) radiation in addition to the visible spectrum. (While this UV and IR radiation is beyond the human visible spectrum, it is capable of causing damage to the eye, so the output should be filtered accordingly.) The duration of the flash can vary from microseconds to milliseconds depending on the size of the capacitors and the impedance of the discharge circuit. This impedance consists not only of the tube, but also of the capacitors' internal resistance and that of any connecting cables.

The energy stored in the capacitors is given as

where E is joules, C is capacitance (in farads), and V is voltage.

The efficiency of xenon discharge systems varies considerably, but a typical conversion rate (electrical energy to visible light) of 10–20% can be expected. The error sources associated with xenon tubes are numerous and complex. In the short term, the path of the arc through the bore of the tube may vary, resulting in a variation in the light output between flashes of apparent equal input energy. As the tube ages, the arc will erode metal from the electrodes, which will be deposited on the walls of the tube, where it will act as an optical filter. The seal around the electrode wires is subject to considerable thermal stress, which may result in the loss of gas pressure. Capacitors will age, the value of the capacitance and the internal resistance changing with time. All of these factors will affect light output to varying degrees, and several strategies have been developed to minimize these effects. The most effective methodology is to measure the light output of the tube and use this value to switch the current flowing through the tube off when the appropriate amount of light has been emitted. This may be done either by switching the power from the flash tube through a very low-impendence quench tube, thus removing power from the light source, or by briefly switching off the power to the tube for enough time to allow the arc to collapse, thus limiting the duration of the flash and therefore its power content. Both these techniques require power in excess of that nominally required by the flash to be available.

The energy output of the flash can be controlled by three means. The method of pulse width modulation may be used, as in the Metz Mecablitz^TM range of photographic flash lamps, where longer-duration (<20–1000µs) flashes are used. The older PS22 range of Grass photic stimulators used switched capacitors to generate a range of output intensities (1.2log units), while the more recent PS33 models use a variable voltage across fixed value capacitors to the same effect (1.0log units).

Given periodic calibration, the author's experience is that the feedback control of flash lamps is largely unnecessary for routine ERG use. It is also the case that better alternative sources are available (see below). Assuming a stable light output, there are a number of potential problems with the use of xenon flash systems for visual electrophysiology. The first consideration must be safety; the voltage across the tube is high, typically 300–800V, and this in itself leads to a potentially serious safety risk, as the tube can be very close to the patient. The trigger voltage is typically 10,000–20,000V, while the peak current flow through the flash tube can be in excess of 100amps, so all cables and connectors must be insulated and screened to an appropriate standard. Even with a well-constructed system, the potential for the generation of electrical artifacts is significant. The high-frequency and high-energy pulses in the xenon flash system are capable of radiating large amounts of electromagnetic energy both from the tube itself and transmission cables. The very fast rise and fall times of the current pulse make it difficult to filter out, a problem that is compounded by the fact that the artifact is naturally phase-locked to the trigger signal for the averager.

Because of the short duration of the flash produced by xenon flash lamps, calibration requires a photometer capable of tracking the very fast rise and fall times of the light pulse and producing an output proportional to the total emission of the source. Although a slow device that integrated light output with time could theoretically be used, the characteristics of many devices are not suitable for measurement of short, intense flashes. Lists of recommended devices are published on the ISCEV Web site.²

The spectral output of a typical xenon discharge system is illustrated in figure 19.2.

Figure 19.2.  

The spectral distribution of the light emitted by a typical xenon flash tube. The insert shows its physical appearance.

Incandescent lamps and shutter systems have, as was mentioned previously, been used by a number of authors¹² to illuminate Ganzfeld stimulators. Incandescent bulbs, while not having the very high output of xenon flash tubes or arc lamps, can provide a relatively high-energy output from low-voltage power supplies. Unlike the xenon flash systems, an electro-optical feedback system to stabilize the light output is easily implemented, and the power supplies themselves are readily available commercially.

The drawback for routine clinical use is, however, the shutter mechanism. While large-area ferroelectric liquid crystal shutters are available, they present a number of difficulties for use in switching large amounts of light necessary for Ganzfeld illumination, such as transmission efficiency and poor contrast ratios. It should be noted that LCD switching systems are used to great effect in projection systems and may be available in suitable form in the near future. Currently, however, an electromechanical shutter is generally used to gate the light path. The disadvantages that affect electromechanical shutters are simple. Because the shutter blade has mass, it requires energy to accelerate and decelerate. With a good drive amplifier, the acceleration can be fast, but a linear shutter will still give a definite rise and fall time to the light pulse. Very fast mechanisms are available that will, with suitable maintenance, provide an effectively rectangular pulse of light from a continuous source. One major advantage of such systems is the ability to prolong the light pulse, the benefits of which are mentioned elsewhere in this book. While lamp and shutter devices have been used to good effect in several ERG research applications, they have limited application in the clinical laboratory.

Light-emiting diodes (LEDs) are also used to illuminate Ganzfeld and other stimulators. These devices are described in detail elsewhere in this chapter.

Cathode-Ray Tubes

Cathode-ray tubes (CRTs) consist of an evacuated conical flask shaped glass vessel, the wide front surface of which is coated internally with a phosphor that is excited by an electron beam emitted from a heated cathode in the neck of the flask and accelerated toward an anode by a high voltage (figure 19.3). When the phosphor coating is excited by this stream of electrons, light is emitted.

Figure 19.3.  

Schematic and cathode-ray tube showing the electron beam, lens, and deflector plates.

CRTs are driven in one of three ways. In its simplest mode, an unfocused beam of electrons excites the entire phosphor-coated area of the screen, causing this surface to be illuminated. These glow tubes can be electronically modulated relatively simply, the rise and fall times being governed by a combination of the switching speed of the electron gun and its drive circuitry and the characteristics of the phosphor used.

When used as a vector display, the electron beam is focused to a small spot, and the beam deflected in both the horizontal and vertical axis by an electrical field. In oscilloscopes, the electrodes that focus and deflect the electron beam are internal, and the field is electrostatic. Electrostatic displays are no longer in general use. In the common TV tube, the electrodes are external coils, and a magnetic field is generated by modulating the current flowing in these coils that deflects the beam. By modulating the current in the coils, the beam can be moved about the phosphor surface. All TVs and monitors develop a raster. The horizontal deflection is relatively fast, and repetitive while a much more slowly increasing current in the vertical coils deflects the beam downward, so a series of illuminated lines is drawn on the viewing surface from left to right across the screen. The beam is switched off while the scan returns from right to left. After each line scan the beam moves a step down the screen, progressively writing the entire visible area of the display (figure 19.4). Light emission from any point on the screen is proportional to the beam strength and lasts only for the duration of the excitation (plus the decay time of emission of the phosphor). Various phosphors are used, depending on the characteristics required from the tube. For tubes with slow refresh rates, a phosphor with a long persistence will give a high luminance output, but this would not be appropriate for high raster rate tubes, in which the long decay time would obscure subsequent scans by residual emission during the following frame.

Figure 19.4.  

The raster pattern employed in all TV tubes. The external field coils move the electron beam horizontally from left to right. During the flyback (dotted lines), the electron beam is suppressed and the position deflected downwards. The number of lines in a television raster is greater than indicated. In standard 625-line TV transmission, the frames are interleaved, odd and even lines being transmitted in alternate frames. At the end of the frame, the spot returns to the upper left hand corner.

The phosphor also determines the color of the emitted light. By utilizing three separate electron guns and three separate phosphors (red, green, and blue), a color display can be achieved. The phosphors can be placed on the surface of the tube in the form of either dots or vertical stripes arranged in triads (figure 19.5). In either case, the points on the screen (pixels) are addressed sequentially through a perforated mesh in the case of the dot phosphor or shadow mask tube, or a vertical array of fine wires in the aperture grill tube first popularized by the Trinitron tube from Sony. The shadow mask tube is generally being replaced by the aperture grill system. In the former arrangement, a significant area is inevitably covered by the masking grid arrangement; while this produces more precisely defined pixels, it inevitably reduces the effective emitting area of the display and, with this, the luminance that is available. The striped array in the aperture grill allows for a far greater area of phosphor and thus higher luminance, while advances in control electronics largely mitigate the greater tendency to blur the pixels. A minor problem, in most applications, with the aperture grill tubes is the visibility of the grill support wires running horizontally across the display surface.

Figure 19.5.  

A diagram showing how pixels are produced by the electron beam in color TV. A, The colored phosphor dots on the face of the tube. The red, green, and blue dots are shown as differing shades of gray. Behind them are the holes of the shadow mask. The electron beam passes through the holes and excites the dots to the required extent. The current must be changed rapidly. B, The Trinitron system. The phosphors are laid down in stripes, and the electron beam traverses a support meshwork aperture grill. The horizontal size of the spot is limited to determine the color emitted. With this arrangement a larger proportion of the screen emits light.

The way the display area of the CRT is scanned leads to a number of potential pitfalls in the use of this type of stimulus display. Although none is overly serious, they need to be considered. The time taken for the electron beam to complete a scan of the display area is given as the vertical refresh rate is normally specified in hertz. Older television displays operate at a nominal 50 or 60Hz, depending on the region, and use an alternate-line interlaced scan, whereby the system writes half the lines of the display on alternate vertical scans or fields; thus, in a display made up of a nominal 625 lines, 262 odd-numbered lines are written in the first vertical field, and the next field writes the 263 even-numbered lines (the remaining 100 lines being used for various signaling tasks). Thus, a complete image on a 50-Hz 625-line interlaced system is written at a real rate of 25Hz. Modern stimulus systems utilize computer displays operating in a noninterlaced mode at vertical refresh rates of typically 100Hz, giving an image refresh rate of 10ms.

If a stimulus is to be frame-locked, that is, the change in stimulus always occurs at a fixed point, usually the start of the sweep, to give a stable image, then the stimulus rate must be in multiples of the frame period. If the stimulus is not frame-locked, then an irregularity will be seen periodically traversing the screen. The result of using non-frame-locked stimuli affects the responses that are obtained. Assuming that the patient maintains fixation on the center of the screen but the trigger can occur at any point within the frame time, there is a jitter of plus or minus 50% of the frame period introduced. With a 100-Hz vertical refresh rate, a 10-ms period of uncertainty is introduced into the implicit time of the response. Therefore, if responses are averaged, the resultant waveform will be broadened, and the peak amplitude will be reduced. This is in some respects approximately equivalent to using a low-pass filter in the amplifier and removes the high-frequency components of the response.

A number of viewpoints exist as to the optimal trigger location and methodology with CRT displays. The ideal should be that the trigger occurs as the scan crosses the fixation point. In most cases, this will be center of the screen and can easily be achieved by adding a delay of half the frame interval to the frame initialization pulse on the monitor synchronization signal. This is readily achievable in a system in which the operator has adequate direct access to either the hardware or the software but is very difficult to do in many commercial systems. An alternative is to specify the vertical refresh rate of the stimulus display and the point at which the trigger pulse is generated with the laboratory normal values for the test in question.

Considerable effort has been applied to the development of CRT-based monitors for the television and computer displays, such that a wide range of specific characteristics are readily available, with high vertical refresh rates (>200Hz), high luminance output, or precise color control, making these various devices readily adaptable for the display of specialized stimuli.

Liquid Crystal Displays (LCDs)

LCDs are in effect variable transmission optical filters. The principle of operation is that liquid crystals rotate the plane of polarization of light to a degree that depends on the voltage across the liquid crystal layer. The two surfaces of the LCD are coated with a linear polarizer, arranged at 90 degrees to each other (extinction) (figure 19.6). This attenuates the transmission through the device. When a voltage is applied, the plane of polarization within the liquid crystal rotates, and more light is transmitted. A practical LCD device has a matrix of electrodes, in rows and columns, each of which applies charge to a very small area of the filter. Therefore, complex patterns can be displayed by altering the voltage across each pixel. In practice, color filters are incorporated into the display, so each pixel consists of a triad of red-, green-, and blue-transmitting regions. The light transmition can be spatially modulated to display an image. With appropriate masking to prevent light spillage, a colored display is produced. Modern versions of these displays have individual transistors built onto the glass substrate to drive each subpixel.

Figure 19.6.  

Principles of LCD systems. Each cell of an LCD system is coated back and front with a linearly polarizing layer, but the directions of polarization are at 90 degrees. Light transmitted through the back plane is therefore polarized, and if the LCD causes no change in the plane, transmission through the front plane is extinguished (right). However, the LCD can rotate the angle of polarization, so light passes through the front plane (left).

LCD displays of the type commonly used as computer displays and televisions currently have two major drawbacks that limit their application as stimulus displays. The first and less problematic of these concerns color rendition. The LCD panel is normally illuminated with a cold cathode fluorescent lamp. These lamps give a very uneven spectral output, with a number of pronounced peaks. This output is then passed through RGB filters, to provide the colored display. The resultant output, while a reasonable color match, is composed of an irregular makeup of wavelengths.

Figure 19.7 shows the output spectra for a conventional CRT display and an LCD screen, both giving a white output of x = 0.33, y = 0.33 at approximately 60cd.m⁻² or 60cd/m². Although these two outputs are metameric, this means that the white is the same for both devices in humans; this is not necessarily true for other species. Furthermore, the change in the output when the color is varied will be quite different for the two displays. Therefore, the color system must be adjusted so that it is consistent with the type of display used. While this is unlikely to be problematic for black-and-white stimuli, accurate color stimulation could be difficult.

Figure 19.7.  

Comparison of the spectral variation in emission of a TV screen and an LCD screen, both of which appear “white” to the human observer.

The second and more serious problem is that of response time. While high-resolution CRTs can readily sustain refresh rates in excess of 100Hz, the refresh rate of most LCD displays is significantly slower. Most of them currently achieve a quoted rate of 40Hz (25ms), while the faster unit can theoretically achieve a vertical refresh rate. These figures are bound to improve with time (until recently, 34ms was the best achievable). These low rates limit the rate of change of any stimulus. Moreover, there is a marked nonlinearity between the degree of light transmission and the rate of change to another gray level, so pixel switching times are not constant. This can, for example, result in a “fast” luminance transient on the reversal of a checkerboard display. However, an even more serious problem is the efforts of the designers of these displays to provide a smoother motion display. With a CRT display, the signal is passed directly to the screen, albeit through a significant amount of electronics. With the LCD display, the signal is held in memory until it is possible to refresh the screen, in effect a “frame grabber.” The combined result of the low refresh rate and the frame delay mechanism is to introduce an unacceptable degree of uncertainty into when a transition sent from the stimulus generator will actually appear on the screen. As the display technology matures and temporal characteristics improve, such problems will become less severe, but without significant improvements in the operating characteristics, LCD displays are unsuitable for most forms of stimulus display.

Furthermore, although surmountable, problems with LCD panels relate to the comparatively narrow optimum viewing angle, which leads to both contrast and chromatic distortions experienced as the subject moves off the optical axis of the display. Contrast ratios were comparatively low with earlier displays, although this is no longer a major issue.

At the time of press, major technical advances in LCD technology are becoming available, such as 4-ms refresh rates and nonlinear pixel switching, which may make these displays suitable for use as stimulus devices.

LCD shutter devices are also commonly used in data projection systems. The color distortions that are introduced by the use of cold light fluorescent illumination in display monitors are largely overcome by the use of halogen illumination, although the temporal considerations still apply. These data projectors are capable of providing a high-luminance, large-area stimulus display and are therefore of great interest, although care should be taken with respect to the timing considerations. It should also be noted that there are wide variations between different makes and models of projectors, and operating characteristics data are difficult to obtain.

Data projectors are also based around both CRTs and the Texas Instruments digital light projection system (DLP). The CRT-based projectors, while generally large and expensive, approach the conventional CRT display in temporal and chromatic performance, being based on a triplet of very high-output CRTs (one each for red, green, and blue).

Digital light projection projectors are based on a novel chip technology from Texas Instruments. The chip at the heart of these devices consists of a large number of micromirrors, each mounted on an individual memory cell. When the state of the cell changes, the angle of the mirror changes, thus deflecting the illumination beam through the projection lens. Intensity modulation is generated by varying the duty time each pixel is on. Almost all of the current systems based on this technology, with the exception of professional theater systems, use a single chip. The RGB separation is provided by a rapidly rotating filter wheel modulation, while the intensity variation of each color is generated, varying the duty cycle of the individual pixel mirror.

Plasma Displays Panels

Large-area, high-resolution plasma screen video displays are available. The technology that is utilized in these displays is phosphor-based, as in a color CRT device. The display consists of an array of vertical and horizontal electrodes overlaying pixel cells containing xenon or neon gas. When the electrodes that intersect either side of a specific cell are energized, the gas fluoresces (in the same way as in a fluorescent lamp), emitting mostly UV light. This UV radiation is used to excite phosphors to produce the visible light. By arranging RGB phosphor triplets of these cells, a color display similar to that on a shadow mask color CRT is produced.

Input circuitry similar to that utilized in LCD panels is needed to control the electrodes arrays. Although the refresh rate is higher than that in LCD panels, this can lead to temporal problems. While large-area plasma displays offer some advantages for stimulus display, a cautious approach to their application is required.

Absorption and Interference Filters

Filters may be used to modify the characteristics of a stimulus by absorbing or blocking the passage of a portion of the light, thus reducing the energy of all or selective parts of the spectrum, thereby changing the intensity or color of the stimuli. Excluding some of the more specialized filters, such as polarizing types, two types of filters are in common use: absorption and interference filters. Neither of these is ideal. Interference filters provide very narrow pass bands but are sensitive to the direction of the incident light, the transmitted wavelengths varying with the angle of incident of the light. They are also expensive, and the largest size that is commonly available is 50mm in diameter. There are a limited number of colored glasses that can be used as filters, but most absorption filters are made of dyes incorporated into a plastic (or gelatin) base. These absorption filters, such as those from Wratten, are relatively low in cost, are well defined, and are available in a range of sizes. However, they are subject to aging at a rate that depends on the energy they absorb. Many of the filters have quite complex absorption spectra and may pass energy at wavelengths outside the main transmission band.

For simply attenuating light intensity, without altering the wavelength composition, several alternatives to neutral density filters exist. One of the simplest and most stable alternatives is a simple aperture to reduce the area of the source that is able to transmit light into the system, a refinement being a variable aperture similar to that used in cameras. A pair of polarizing filters can be used in which the angle between the planes of polarization increases toward 90 degrees, the amount of light transmitted is reduced. This is the basis on which liquid crystal devices operate, and large-area LCD filters (as opposed to shutter mechanisms) are available as both colored and neutral electronically variable filter mechanisms, although at the time of writing, they are of limited availability, and as with other LCD mechanisms, they are subject to temperature-related variability.

As with the light source itself, calibration of filters is essential.

Light-Emitting Diodes (LEDs)

Light-emitting diodes (LEDs) are nearly ideal light sources for many purposes. They are small, require low voltages and currents to drive them, and can be controlled by simple electronic means to give either continuous light outputs, extremely brief flashes, or complex waveforms (or a mixture of these) over a wide range of intensities. Their light output changes by little in intensity or relative spectral emission over extended use.⁵ These properties not only simplify calibration but also reduce the frequency with which it is needed. Many different spectral outputs are available, ranging from the near ultraviolet through white to infrared in a variety of different packages with differing optical properties. The majority of the available devices are inexpensive. Until recently, the limiting factor to the use of LEDs for visual stimulation has been the relatively low outputs available; however, white LEDs are now being introduced as automobile lamps, in which a cluster of six devices gives an equivalent output of a 20-watt filament lamp. The available output energy is forecast to rise significantly over the next few years, possibly to the point at which these devices may be used as headlights. Nevertheless, even when considerable quantities of light are required (e.g., in electroretinography), LEDs can be used to advantage.

LEDs are members of the family of epitaxial semiconductor junction diodes. A junction is formed by growing a very thin crystal of a semiconductor directly onto another, slightly different, semiconductor surface. The two layers of semiconductor material (frequently gallium aluminum arsenide) contain different impurities (dopants). As a result of these impurities, one layer contains an excess of free electrons, and the other contains an excess of holes (positive charge). The energy that is required to move a charge across the junction against the concentration gradient of free electrons or holes is considerable and larger than in other types of diode (approximately 3:1). This energy band gap must be exceeded if current is to be passed through the junction. When the device is forward biased, electrons move from the negative material to the positive, and a corresponding movement of positive charge or “holes” occurs in the reverse direction. When an electron and hole pair recombine, energy is emitted as a photon. The characteristic wavelength of the photon is determined by the energy band gap. Thus, it is more difficult to produce short-wavelength LEDs, since the higher energy gap must be maintained with the controlled flow of charges. The construction of the epitaxial layer determines the direction of light that is emitted, and the absence of a resonating (reflective) cavity prevents the stimulated emission of radiation (lasing) by the photons, so the light is not coherent and contains a number of differing wavelengths. However, light is emitted over a relatively narrow bandwidth. For a typical red LED (for illustrative purposes, a Stanley HBR5566X), the peak is at 660nm, and the half-power bandwidth is ±30nm. This is a purer red than is readily obtainable with gelatin filters, though expensive, multilayer, complex, band-pass interference filters can produce a better approximation to a monochromatic light.

The construction of a typical LED is shown in figure 19.8. The semiconductor is mounted on a lead frame and encapsulated in a plastic (epoxy) housing with an internal spherical lens. The combination of junction structure, epoxy, and lens type determines the spatial output characteristics of the device, and for many LEDs, including all the “brightest,” the light is concentrated in a cone, which can be represented in a polar diagram (figure 19.9) with a half-power spatial polar distribution of ±7.5 degrees.

Figure 19.8.  

Construction of a typical LED.

Figure 19.9.  

Polar diagram showing the spatial concentration of light from an LED.

A typical device requires 30mA at around 2V to produce its maximum output, so it is both easy to control and intrinsically safe to use in a clinical environment. The junction of most modern devices thus exhibits low electrical impedance but is nevertheless fairly robust, being able to tolerate significant overloads for short periods.

An exception and extension to this construction methodology is the white LED. In these devices, a short-wave-length-emitting junction (approximately 470nm) is encapsulated in a phosphor material, similar to that used on the inner surface of black-and-white monochrome CRTs. The diode emits blue light, which excites the phosphor, which in turn emits white light. While this “white” light is a continuous broad-spectrum emission, it normally contains a major peak at the primary wavelength of the diode. The spectrum of a typical device of this type is illustrated in figure 19.10. While this continuous spectral output may offer significant advantages for ERG use when compared to white light simulated by three narrow spectral lines of red, green, and blue light, some care is required, as the phosphor, like that used in CRTs, will have a definite decay time, prolonging the trailing edge of a light pulse.

Figure 19.10.  

The emission spectrum of a typical “white” LED. Note the peak in the blue, which is the primary emission, and the second peak, due to the phosphor emission (see text).

Light-emitting diode technology is advancing at a considerable pace, to the effect that devices that only a few years ago were used as indicator lamps and little else outside the laboratory are expected to be utilized as vehicle head lamps within the next two to three years. Some street lighting systems are already being installed, including an array of fewer than ten devices with suitable heat sinks potentially capable of matching the output of the current generation of xenon lamps.

Recent work suggests, however, that phosphor-enhanced white devices will age rather more rapidly than other LEDs due to a decay of the phosphor with use, similar to that which occurs in CRTs. This decay effect aside, these high output devices have the capability of producing the high luminous flux necessary for electroretinography in standard-sized Ganzfeld stimulators.

Organic LEDs (O-LEDs) are beginning to appear in numerous guises. This family of polymer-based LEDs can be printed to a plastic or glass substrate using inkjet techniques, thereby producing a low cost, high-luminance, high-resolution, RGB color display with none of the geometric limitations of LCD-based devices. It is possible that display devices based on O-LEDs will become the obvious replacement for large format CRT displays, bypassing those problems associated with the current generation of LCD-based monitors.

Various types of LEDs are available. Some are devised for special purposes—for example, alphanumeric display components or indicators. There is a great range of shape, size, and intensity. A number of devices are packaged with additional circuitry that provides a constant current flow or flashes the device on and off. Other devices contain more than one junction and can produce two or even three colors, but these are so specialized that their value for purposes other than those for which they were designed, that is, for instrument displays, is limited.

The colors that are quoted by manufacturers vary from ultraviolet to infrared, with many wavelengths being available across this range. A recent review of devices available showed that steps of 10nm across the visible spectrum were easily achievable, although outputs varied considerably.

Until recently, the longer visible wavelengths (yellow to red) have been the brightest devices available. However, at the behest of the lighting industry, much development has been focused on the shorter (blue) end of the visible spectrum.

Certain problems and their solutions are common across the range of devices available. The relationship between applied current (or voltage) and light output of a typical device is shown in figure 19.11. For a region of about 1.5 decades, it can be seen to be approximately linear. Above or below this region, marked nonlinearities occur. Since in general, users wish to control the intensity over a much larger range, a variety of drive circuits have been devised.

Figure 19.11.  

Relationship between applied voltage (V_f), current, and relative light output (Int) of a typical LED.

For a simple flash stimulator, a voltage drive circuit may be used quite effectively (figure 19.12), and the relationship between light output and applied voltage can be determined by calibration. Better performance can be obtained by replacing the voltage drive with a current source or ideally by placing the LEDs in the feedback loop of a current drive (figure 19.13). This technique, while a significant improvement over the simple circuit in figure 19.17 still limits the range of linearity available to around 2.5 decades. Thus, if a true sinusoidal output is required, the depth of modulation can never be 100%.

Figure 19.12.  

Voltage drive circuit of a simple flash stimulator.

Figure 19.13.  

Feedback loop of a current drive to enhance the performance of an LED.

Two alternative techniques may be used to obtain linear control over intensity. The first consists of pulse density modulation.⁷ The LEDs are driven by pulses of short duration (100-ns pulses have proved readily achievable in the author's experience), each of fixed power content. Light intensity is altered by changing the repetition rate of the pulses. An upper pulse frequency limit of 5MHz is readily attained. For low intensities, a rate of 50Hz is well above the critical fusion frequency of the human eye; thus, a wide range of intensity of an apparently continuous source can be achieved. If these pulses are derived from a linear voltage-controlled oscillator (VCO), a device in which the frequency of the output is related to the applied voltage, the output pulses are shaped and used to drive the LEDs through a fast switching circuit. Although good VCOs with the required range are difficult to produce, several are available as either integrated or hybrid circuits and can be driven from any waveform source so that very complex temporal changes can easily be produced. The light intensity may also be simply controlled without changing the waveform by passing the output of the VCO through frequency divider circuits. These may be readily produced by using standard logic components. Thus, a visual stimulator with a dynamic range of six orders of magnitude, consistent modulation capability, high stability, and fine control of intensity may be easily produced.

A similar effect can be achieved by using pulse width modulation, as opposed to pulse frequency modulation. This technique has gained favor recently, partially owing to the ease with which control circuits from switch mode power supplies can be utilized to control pulse width, thus eliminating the need for expensive VCOs.

One major drawback to pulse density or pulse width modulation systems derives from the high-speed switching of LEDs, especially if an array dissipating large amounts of energy is required. Each pulse contains frequency components much greater than the pulse repetition rate (5MHz) that are caused by damping inadequacies of the power switching circuits. Thus, the LED array acts as a very high-frequency (VHF) radio transmitter and may radiate tens of watts. (This effect can be more easily reduced in the pulse width modulation systems as the control circuits often include control of the respective rise and fall times of the pulses, thus reducing the tendency to emit the higher harmonic frequencies.)

Because the source is usually very close to both the patient and the preamplifier and because modern clinical amplifiers use high-input impedance field effect transistor input stages, which in practice make very effective FM radio receivers (the common mode rejection ratio is relatively low at very high frequencies), large stimulus artefacts are generated. Therefore, the pulse frequency modulation technique is of use primarily for psychophysical experiments.

An alternative approach is to use a low-frequency current source to drive the LEDs and continuously measure their output with a photo-sensing circuit. The output from the detector circuit is compared with the waveform input signal, and any difference is used to modify the LED drive current and thus the light output. This current modulation approach will operate effectively only over a range of three to four orders of magnitude; to go beyond this would require an unduly complex drive circuitry.

There are two significant defects affecting LEDs that the user should be aware of. The first primarily affects some older short-wavelength devices (GaN blues being a case in point), namely, a variation in emitted wavelength with current and therefore intensity. This change in wavelength can be in the region of 5–10nm over the usable drive current range. It is, however, largely eliminated on modern devices. The shorter-wavelength blue devices also emit some energy in the ultraviolet, which can degrade the epoxy compounds that are used to encapsulate the devices, causing a loss of output through the imposition of a yellow filter. The second and potentially more serious problem affects the longer-wavelength devices. Because of the robust nature of the device junctions, it is always tempting to overdrive them for short periods and thus increase the available light intensity; indeed, most manufactures quote two maximum forward currents; one for continuous use and another significantly higher figure for reduced duty cycle operation. However, even at the maximum continuous forward current, there is a progressive fall in light output for a fixed current due to heating effects within the junction. This effect is very noticeable in the orange and red devices, becoming much less of a problem as the emitted wavelength shortens toward the green portion of the spectra.

Arrays

The low power requirements and general ease of use make LEDs an ideal choice for many types of stimulator. Because of both the size of the radiating area and the relatively low radiant energy of many types of LED, it is generally necessary to use a number of devices to construct an effective stimulus. Here again, the low drive requirements make it a simple matter, provided that some care is taken in the design to interconnect a number of the diodes in an array. By using either the pulse or constant current modulation approach, an array of several hundred LEDs can be assembled to provide the required stimulus. They can be connected in series, parallel, or a combination of both. Thus, if nine devices are required for the stimulus and the power unit has an output of 9V, three LEDs, each with a forward voltage drop of 2V at 50mA, could be connected in series (figure 19.14A), then three identical chains connected in parallel (figure 19.14B) will give a stimulus with maximum light output at three times the maximum continuous forward current of each device. Care should be taken to ensure that the forward voltages of the diodes are closely matched, or, preferably, select a resistor to balance the current through the three chains. This resistor will have the added benefit of protecting the drive circuits should an LED fail and short-circuit.

Figure 19.14.  

Series and parallel connections of LEDs.