Drop into your local DIY store or department store and you will see an alarming array of led lighting products, from halogen replacements to colour mood lighting. LEDs are also being designed into applications ranging from backlighting for lcd tvs to headlight clusters. Even municipal authorities in Europe are using led street lighting to save electricity and CO2 emissions.
The appeal of the technology is obvious when you consider LEDs offer efficiencies up to 100Lumen/W and an average lamp life of up to 100,000hr, while tungsten and halogen lighting offer less than 25Lumen/W and an average lamp life of less than 4000hr.
Two overriding issues should be considered when driving high brightness leds. The first is to ensure that a fairly constant current is available with minimal variation; this is because a fluctuation in current can cause an apparent change in brightness and the LED's nonlinear response to current can exacerbate this. The second issue is to control the LED's temperature. As this affects brightness and longevity, it is important that lighting systems enable heat to be dissipated effectively to control temperature. In addition, it may be necessary to monitor temperature and reduce the current through the LED. This has the effect of lowering its temperature and optimising light output, efficiency and longevity.
There are broadly three kinds of LED control: resistor based; linear; and switching (see figure 1).
LEDs are typically arranged in strings of up to 12. Each LED will have a forward bias voltage (Vf) of typically 3 to 4V, so a string of 12 would require up to 48V. LEDs have ratings starting at about 350mA and can exceed 3A.
Resistor based control is inexpensive and simple to implement, but the downside is that current can vary when Vf changes. A string of up to 12 LEDs could have a significant variation in forward bias voltage, which in turn varies the voltage drop across the resistor. Additionally, placing a resistance in series with the LEDs is inefficient and will generate heat. Linear control is simple to implement but is also relatively less efficient and will still cause heat to be dissipated.
The optimal solution for high brightness LEDs is switching control, which is highly efficient but relatively more expensive and normally requires some form of cpu. An example of this methodology can be found in Cypress' PowerPSoC family of controllers.
With switching control, a constant current regulator controls current through the LED (see figure 2). At time t=0, the FET switch is turned on and current starts to flow through the string. The inductor limits the rate of increase in current. The Rsense resistor is very small and detects the current passing through the string. When the differential voltage drop across Rsense is large enough, the current sense amplifier detects that the threshold is reached and the FET switch is turned off. Energy stored in the inductor allows current to flow through a diode in the return path and current through the string will diminish as energy stored in the inductance is dissipated. When a lower voltage threshold is detected across Rsense, the FET is switched on and current ramps up again. The size of the ripple in the current is determined by the hysteresis in the switching characteristics of the comparator. In PowerPSoC, this feature is programmable and set by d/a converters. Current ripple is generally not detectable by the human eye as long as the ripple is less than 20%.
XHD
LED shortcomings need to be taken into account when applying this methodology. LEDs yield different values of brightness, power and chromaticity in manufacturing and are put into different 'bins'. LEDs with a tighter specification are more expensive; LEDs with a looser spec are cheaper, but control of colour and brightness is more difficult.
This makes it difficult to achieve repeatable colour points. Applying the same current to two sets of red/green/blue LEDs from varying chromaticity bins will produce visibly different hues of the same colour. Similarly, applying the same current to LEDs from two flux bins will produce two levels of brightness. Therefore engineers require flexibility and programmability within their switching regulators to account for these variations.
Additionally, LED output degrades over temperature and different colour LED outputs degrade at different rates. This makes it difficult to maintain the same colour over time, especially when taking into account fixture heating due to LED power dissipation. Thermal feedback capability in a switching regulator ensures that colour variation is kept to a minimum. PowerPSoC implements this feature, allowing engineers to decrease the d/a converter values on the upper and lower Rsense current threshold levels to reduce power dissipation.
Dimming also poses a problem. An LED's colour temperature varies with current and its luminescence is a non linear response to current, making precision analogue dimming difficult. To do this effectively requires knowledge of the LED's transfer function in order to map a linear ramp to the non linear response. For multicoloured LEDs, this is complex.
Most controllers use digital dimming by a PWM to get around this. The PWM mark space ratio and frequency is used to control the brightness of the LED by turning off the FET switch, causing the LED current to go to zero. The LEDs then remain off until the PWM output changes. FET control is then passed back to the comparator function, avoiding any problems associated with non linearities.
The switching frequency must be more than 150Hz to avoid flicker, but if the PWM frequency is too high, it will not allow the current ramp to go to zero and switching losses in the FET might become apparent over time.
One alternative method of digital dimming is Cypress' Precise Illumination Signal Modulation (PrISM), used in the PowerPSoC controller. A normal PWM uses a counter and a programmable combinatorial decode to drive its output. PrISM uses a pseudorandom source based on a shift register and two input EXOR with feedback into the shift register. This enables designers to take the values of the counter over the PWM period 't' and mixes them in an apparently random fashion. The value present on the shift register then feeds into programmable combinatorial logic, which creates a pseudorandom series of pulses. Over the period 't', this modulation guarantees the overall high and low times are maintained.
PrISM is a form of pulse density modulation, which effectively reduces peak emi emissions, especially when switching several strings of LEDs.
LED control is far from straightforward and most engineers will come across some of the issues raised here. PowerPSoC represents an attempt to address designers' issues with a programmable controller that can handle the many variations in LED performance over time, temperature and between bins, eliminating much of the complexity.
