FEATURE

Shaking down and lighting up

Stephen Mounsey takes a look at the LED-based lighting industry and discovers that, when it comes to testing these light sources, legacy approaches are often insufficient

According to the organisers of the recent Invest in Photonics event held in Bordeaux, France, solid-state lighting looks likely to be a significant sector of the market for photonic technologies over the next few years. Given the concern of environmentalists around the world over the impact of climate change, and the push towards energy efficiency, efficient and long-lasting lighting based on LEDs is an attractive prospect. That efficiency, however, comes as a result of the LED being very different from conventional light bulbs, so specialist techniques are needed when measuring the quality and the power of light produced by the new generation of efficient, solid-state lighting fixtures and components. Spectrometers are used in the industry to check LEDs and LED-based light fittings at every point along their supply chain. Dave Jenkins is president of Orb Optronix, a US-based supplier of LED characterisation apparatus. ‘If one cares about solid-state lighting products having the same lit appearance between units, then one must understand variability,’ he says. ‘The largest culprit in introducing variability in LED illumination systems is often the LED package itself.’

One perceived advantage of LED lighting is that the components can be inexpensively mass-produced using techniques perfected in the semiconductor industry for microprocessors, but there are drawbacks, as Marco Snikkers, commercial director at spectrometer developer Ocean Optics, explains: ‘When you do semiconductor processing, there is always a variation in the processing. Even if you use very high accuracy etching and lithography, you will always get some kind of a spread in the outcome, and for monochromatic LEDs, you will see this variation in the form of LEDs with slightly different colours.’ For this reason, he says, many manufacturers of LEDs – particularly in the Far East – use testing machines to sort LEDs by colour: ‘In China, the largest market [for Ocean Optics’ spectrometers] is in LED sorting machines. We have quite a large market for OEM customers building LED sorting machines to sort and bin the various LEDs when they are sorted and manufactured. There’s a spectrometer in there somewhere, and we want it to be manufactured by Ocean Optics.’

The machines examine and sort each LED, but what properties do these machines measure? ‘For monochromatic LEDs, the manufacturers are interested in relative power output and also x-y coordinates,’ says Snikkers, referring to the x and y coordinates of the Planckian locus – the way in which physicists define colour. ‘The manufactures have certain bins, each of which has target x and y values with a certain spread in them. The chips are placed in the various bins, and the customers can then order a batch of LEDs from a certain bin.’ According to Snikkers, the final colour produced by an LED is a combination of luck and process control, and the technique of sorting and binning is not unique to the LED industry: ‘The more you get into bulk manufacturing, the more reliable the process gets, but even companies like Intel used to [sort and] bin their processor chips into bins of 100MHz, 200MHz, and 300MHz clock speeds; they were all the same chip, but the speed at which they could be clocked without overheating depended on how accurately they were produced. It really is normal procedure for semiconductors – there’s always a statistical spread around a target value. In LEDs, the target value is colour, and you get all kinds of variations around that.’

While monochromatic LEDs have a great many uses in a great many application areas, it is white LEDs that are of interest for applications in solid-state lighting. ‘If you get into the white LEDs, testing gets even more interesting,’ says Snikkers. ‘White LEDs usually consist of a blue LED with a phosphorescent coating. Initially you have the spread [variation] in the blue, and then you have phosphor put on there as well, which can be in the form of a coating or added as a lens. The total amount of luminescence depends on the thickness and quality of the phosphor, and so you’re adding another level of variability to the variability you already have due to the blue LED. You get a certain white spectrum – a blue with phosphor spectrum – and that will give you a certain colour temperature; whether it’s a cold LED or a warm LED.’

Colour temperature, he explains, is an important concept for LED manufacturers hoping to have their products used for everyday lighting applications. In physics, a ‘black body radiator’ is a theoretically perfect emitter of radiation at all wavelengths on the spectrum; the colour of the radiation emitted depends on the temperature of the black body. Confusingly, high-colour temperatures correspond to so-called cool colours like blues and whites, while low-colour temperatures correspond to warm colours such as oranges and reds. Snikkers explains that LEDs are sorted by, and sold in, colour temperatures.

The eye of the beholder

The choice of detector for LED lighting is important. When measuring the total output of a light source – its luminance, given in units of lumens – testers make use of an integrating sphere, such as those provided by US-based Labsphere. Integrating spheres are large enough to contain the light source to be tested, which may be up to two metres in length for some light fixtures. The inside of the sphere is highly reflective, and the sphere only has one exit, which will lead directly to a detector of some kind. Robert Yeo, owner of UK distributor Pro-Lite Technology, describes the detectors that can be used with an integrating sphere: ‘There are two approaches: you can make a lumen measurement using a simple photometer – a lux meter. This uses a silicon detector with a photopic response,’ he says, explaining that ‘photopic response’ is a term used to describe a detector with the same sensitivity to various wavelengths within a spectrum as the human eye. ‘The human eye perceives green light more strongly than red light or blue light. If we plot the spectral sensitivity of the human eye at all wavelengths, we get the CIE standard photopic observer curve. In instrumentation terms, this photopic response is achieved by using a combination of a silicon detector and a glass filter. It must exhibit the same sensitivity as the eye, and simple photometers [also known as lux meters] are able to do exactly that.’

Ocean’s Snikkers gives an example of the relevance of this photopic response: ‘Lux is a measure of optical energy, but it is calculated with the sensitivity of the eye. It’s therefore a measure of how your eye would perceive and interpret the colour of light. For example, our eyes are very sensitive to green, but not so sensitive to red or blue. This means that if you had a red, blue, and green LED, and if your eye saw them as all being the same, the red and the blue would probably be more powerful in terms of watts.’

When it comes to measuring the luminance of an LED light source, a simple photometer is not suitable: ‘A normal lux meter is a silicon detector with a certain sensitivity,’ says Snikkers, ‘and then an eye-sensitivity curve is added on top of that detector in the form of a filter. This approach works OK with an ideal black body, because it is calibrated with a tungsten lamp, which approximates to a black body. LED light has a lot of jumps in it, because you have a peak at the blue, a phosphor peak as well somewhere in the orange-red area, and so it is not at all like a black body. If you look at this light with a lux meter, you will see the impurities of the filter and detector, and you will get a mismatch against the calibration, which is done with the tungsten bulb. That mismatch can be as high as 10 or 20 per cent.’

Pro-Lite’s Yeo is also familiar with this limitation: ‘As soon as you’re trying to test a light source that has a very peaky spectrum – fluorescent sources, metal halide sources, and in particular LED sources – the shape of the spectrum when measured with a filter-spectrometer can lead to some errors.’ According to Yeo, the American Illumination Engineering Society has gone as far as issuing standards stating that when characterising LED or solid-state lighting, one should use a spectral radiometer. ‘The spectral radiometer uses a diffraction grating to separate the light out into its component wavelengths, which are spread across a detector array, and the device samples what the spectral power is at each wavelength. You measure the spectral distribution and then you apply the photopic response curve – the standard observer function – through software. It’s an inherently more accurate way of doing things,’ he says, adding that the Labsphere integrating spheres that the company distributes are normally supplied with a spectral radiometer. As well as giving an accurate measurement of total luminance, Yeo states that a spectral radiometer can give a reading of colour temperature and colour rendering as well.

Being good with colours

Snikkers describes the meaning of the colour rendering index: ‘The colour rendering index is a measure of how close [the light source] gets to an ideal reflection of a certain standard. You have about 15 different colour rendering index reference tiles, and the indexing is done by experiments with real people. A tile is illuminated by a certain type of light, representing the northern sky sunlight, or equatorial sunlight for example, and then the testers illuminate the same tile with the light source that is being tested – an LED light fitting for example. The tester states whether the tile looks the same as it did under sunlight.’ A subsequent calculation, he explains, leads to a percentage result, with 100 per cent representing an ideal black body, illuminating all colours evenly. ‘Normally, with an index of 80 per cent or better, a normal human being would say that [the colour rendering] is OK,’ he says. In some applications, however, such as graphics studios and clothing shops, the need for perfect rendering is more pronounced. ‘A colour rendering index of 90 or 95 per cent is needed here, and that can be a pretty tough figure to meet, but the latest generation of LEDs are getting there,’ says Snikkers.

Semiconductor manufacturing techniques are inherently variable, and steps must be taken to ensure that the colour and power of LEDs is accurately specified by the manufacturer. At the heart of this LED sorting machine is a spectrometer manufactured by Ocean Optics

Good colour rendering is achieved by mixing LEDs of different wavelengths, or by adding different kinds of phosphors that emit at different wavelengths. ‘To get the best colour temperature or the best colour rendering index, you have to have a very broad wavelength light source, nicely reproducing the black body. Some manufactures add red LEDs to the basic blue-with-phosphor white LED, and some add multiple kinds of phosphors so that they add peaks on top of each other so as to follow the black body curve as closely as possible,’ says Snikkers.

A fitting test

As impressive as high brightness LEDs with perfect colour rendering are, they are unlikely to be sold to consumers as individual units. Instead, lighting manufacturers will integrate them into their own light fitting products. Pro-Lite’s Yeo, describes the importance of specifications to lighting designers: ‘As with all products, you want to know how they’re performing in order to correctly specify them. More specifically with solid-state lighting, the output of the luminaire [light fitting] needs to be expressed in a very specific way.’ In practice, these specifications are given to lighting designers in the form of standard photometric data. ‘A goniophotometer measures the luminance as a function of angle,’ explains Yeo.’ ‘It’s useful to know how much light the light source is emitting, and how much in each direction.’

Producing this data can be expensive for a manufacturer of light fittings: ‘Normally, in order to get accurate data, you have to make measurements a long way from the light source,’ says Yeo, adding that for a 1.5m fixture, such as a fluorescent tube, the measurements would need to be made at a distance of at least 15m. ‘This is the traditional far field approach, and the sort of equipment needed to do that sort of work is expensive – in excess of £100k-£200k – and a huge dark room as well. As a result, most small to medium size lighting companies would rely upon the services of external test companies.’

Pro-Lite offers an alternative in the form of its near field goniophotometer. This uses an imaging photometer to take a picture of the light source from many angles. ‘Using software and a ray-tracing algorithm, we can reconstruct what the equivalent light intensity would be in the photometric far field,’ says Yeo. The customer is able to perform the measurements in a much smaller dark room, with far more affordable equipment. Developments such as this go towards reducing the cost of LED lighting, meaning that the promising energy-efficient technology will be on the shelves of hardware and DIY shops that bit sooner.