newsletter link
mondo arc

Driving Responsibly

Issue 48 Apr / May 2009

Geoff Archenhold explains why LED driver reliability will be essential for SSL to succeed.

The editor at mondo*arc has received quite a few requests and comments on manufacturers’ claims regarding the reliability of LED systems and so I have decided to devote the next few SSL articles to LED system reliability and more particularly LED drivers and their reliability.

Before I launch into lots of technical maths and the theory of system reliability I have been asked to reproduce the five simple quality questions which I presented at the recent IALD conference at The ARC Show. The five simple questions allow anyone in the lighting supply chain to quiz their LED suppliers to ensure they end up only purchasing quality products. If your LED supplier cannot provide the information then it is likely their products are purchased from an unreliable source which could mean problems for your projects in the future. The five questions are as follows:

1.    Which LED and Phosphor does your fixture use?

  • Ensures the product has a quality LED to start off with and give the end user a high assurance that the LEDs will have a good lumen maintenance performance.
  • If the supplier cannot answer this question convincingly then be very careful as the LEDs may be low cost and not perform well for long.

2.    Please show me the independent photometric test report?

  • Proves real performance of the LED lighting fixture in operation. Most quality LED fixture suppliers will use reputable test laboratories such as BSI, 42 Partners, PC(UK) or the Lighting Association in the UK.


3.    Can you show me the standard safety test reports for all components?

  • Quickly reduces sub-standard LED products – provides the end users with the assurance that the products, especially LED drivers, are safe!

4.    What lifetime testing information can you provide as assurance?

  • If no test data – do you believe the warranty? Most companies will not have any lifetime test data for 50,000 hours but if they could demonstrate something which is a few thousand hours of lifetime data then this will ensure that they have done some reliability engineering

5.    What IPR or licenses come with your products?

  • LED IP space is complex so check if they have a licence or they have their own IPR, otherwise the products you are selling could be infringing on someone else’s technology which could be very costly in the future!

To start off this month, I thought I would mention what I hope will become a trend from LED manufacturers who publish datasheets. A major breakthrough has been achieved as the LED manufacturer, Bridgelux, has been the first to publish the performance data of their LEDs when operated at what could be considered “Real” temperatures rather than the unrealistically low temperatures usually quoted at 25oC.

Up until now LED datasheets referred only to the lumen characteristics of a single LED die at 25oC in short pulsed mode (so called “Cold Lumens”) so when the LEDs were formed into an array and operated at a nominal current the lumen output of the array was always difficult to compute due to the temperature differences seen in practice from those found in the LED datasheet. This meant that the majority of LED fixture manufacturers would simply compute the lumen output of their LED fixture by calculating the number of LEDs within the fixture and multiplying it by the minimum “cold” lumens figure per LED. Of course this is totally incorrect and has resulted in many lighting designers getting very excited over claims by manufacturers regarding the light output of LED fixtures and then getting very disappointed when they observe most LED products do not measure up to their claimed light output due to the “cold” lumen calculations falling short from real life operating conditions.

To my knowledge the first LED fixture manufacture to provide real lumen performance data was IST Lighting with their DL LED downlighter range which measured the performance of their products at standard operating conditions rather than the “cold” lumen figures. With the advent of LED manufacturers such as Bridgelux leading the way in discussing “hot” lumen performance data then I hope the trend will be for other LED lighting manufacturers to follow suit so that designers and specifiers can believe with certainty the LED datasheets.

Bridgelux has launched a series of high power and high efficiency LED arrays (See figure 1) that break the traditional price/performance ratios of LED systems by placing more than one high power LED die within a package. Although not the first LED manufacturer to do this, others include ENFIS, Edison Opto and indeed OSRAM, the products are specifically tailored for high volume lighting applications and are highly cost effective. The arrays come in two formats - a star or square tile configuration - which have options of Cool, Neutral and Warm White from 400 Lumens up to 2000 Lumens dependant upon system configurations.

Table 1 (see over) highlights the various light output flux characteristics for the LED arrays and importantly there are several columns covering the minimum “cold” lumen values as found in other manufacturers datasheets such as Philips Lumileds and Cree as well as the typical “cold” luminous flux at 25oC and the typical “hot” luminous flux at 60oC measure at the board level.

Incredibly, the “hot” lumen output is only slightly poorer than the “cold” lumen output eg; approximately 10% less. This is unusual as most LED systems suffer ~30% thermal degradation from “cold” lumen data but the Bridgelux team state that they have improved the performance of LED arrays through the introduction of their proprietary Metal Bond Technology helping to reduce the thermal resistance of their LED arrays with figures down to just 0.5°C/W (this compares to ~9°C/W for most LED emitters) for the 2000 lumen arrays.  The second advantage of the arrays is that the system reliability and costs are reduced because the LED arrays can be attached directly to the heatsink and so there is one less thermal interface required compared to the current LED arrays created from more than one LED emitter. The component placement costs and the MCPCB board are also taken away from the Bridgelux system making the products more cost effective.

This technology improvement enables the Bridgelux family to compete against a raft of conventional lamp technologies in terms of raw lumen output but with improved lumens per watt efficiency as shown in table 2.

The launch of new standards such as LM-79-08 entitled “Electrical and Photometric Measurements of Solid-State Lighting Products” enables the lumen depreciation or lumen maintenance of LEDs to be measured providing a measure of LED reliability. The first work on LED reliability was undertaken by the ASSIST programme at the Lighting Research Centre in New York State which proposed a lumen maintenance of 70%, corresponding to a 30% reduction in initial light output, as the end of useful life for general lighting which has now been adopted by the relevant standards.

Unfortunately an LED system has many components that need to work together in order to enable light to be emitted from the LED, these include:

1.    The LED emitter
2.    An LED driver (not required if using AC LEDs)
3.    Mechanical and thermal management components for long operation

Even if today’s LED emitters can operate beyond 100,000 hours the total LED systems reliability is dependant upon the weakest component within the system. For example, you could have an LED that is reliable for more than 1 million hours in standard operating conditions but if the thermal management components of the LED system are poor it could only last for a few hundred hours despite having a great LED!

Therefore, in order to determine LED systems reliability the system designer must consider the possible failure modes for each component.
It is generally known that one of the weakest parts of an LED system is the LED driver due to the number and types of components they contains and so the this article will deal with discussing LED drivers and their reliability.

What do we mean by an LED driver?
The main issue with defining problems is that of what language or terminology to start off with and LED drivers are no exception as a driver means different things to different people.

The definition of LED driver I will refer to in this article is that which describes a complete LED driver subsystem enabling a user to attach high voltage AC mains in and is capable of driving a series of LEDs on one or more output channels using a variety of control techniques to obtain a desired lighting effect as shown in Figure 2. In other words it is more than just an LED driver Integrated Circuit.

The inside of a typical LED driver looks very similar to that shown in Figure 3 with high voltage components such as transformers, capacitors, MOSFETs and inductors.

There are many types of LED IC’s that can be used within the LED control system and many of them use different power circuit topologies mainly dependant upon the input and output voltage requirements of the LED drivers as shown in Figure 4. Table 3 outlines the advantages and disadvantages for each type of driving topology however for applications that use more than two LEDs, and/or there is a large difference between the input and output voltages a switching topology is more appropriate to use and is a more efficient way of handling varying system circumstances.

Two topologies offer a simple implementation: the buck topology and the boost topology. Both offer high efficiency over a wide range of electrical parameters.

The two topologies achieve their effect in different ways, and these differences affect the efficiency with which they operate. In the buck topology, the LED is placed in series with an inductor. This means that only a fraction of the energy has to pass through the inductor, the component which contributes a significant part of the losses. In this topology, the average current through the inductor is no more than the current through the LED.

In the boost configuration, all the energy passes through the inductor, as it is charged to ground and then discharged through the LED, averaged by a capacitor. In addition, the current through the inductor is higher than the current through the LED; losses increase with the square of the current. In practice, in a circuit with comparable power output, a typical boost regulator will deliver around 85% efficiency versus around 95% for a typical buck converter.

Unfortunately there is not one best way for driving LEDs. However the majority of LED drivers use current mode switching techniques due to the high efficiencies, good accuracy of LED current regulation and wide system flexibility.

New LED driver standard IEC 61347-2-13
From the 1st July 2009 all LED drivers need to adhere to the IEC 61347-2-13 safety standard which specifies particular safety requirements for electronic control gear for use on d.c. supplies up to 250 V and a.c. supplies up to 1 000 V at 50 Hz or 60 Hz and at an output frequency which can deviate from the supply frequency, associated with LED modules.

There are many improved requirements that this new standards brings to LED drivers however the most noticeable one will be the change required in the output connectors that connect the driver to the LED fixture. The majority of LED drivers use the RJ45 connector originally designed as network card connector in PC’s and these will need to be changed to double insulated type connectors if the maximum output voltage for a constant current driver is greater than 25Vdc.
It will be essential that all new LED drivers meet this standard so ask your suppliers to provide you with the test reports to ensure your installations are safe.

Driver efficiency can affect reliability

The previous section discussed different driver topologies and their relative efficiencies however the overall driver efficiency is important for system reliability as shown by Table 4. Here a matrix of the overall driver efficiency is created by looking at the efficiency of the AC to DC and DC to DC stages of an LED driver. As shown the overall efficiency of the LED driver significantly falls as both stages become less efficient with anything less than 90% efficiencies in both stages resulting in significant losses in total driver efficiency eg <81%.

For LED driver PSUs with low wattages then the overall power efficiency of a driver may be managed however for larger wattages the inefficiencies are generated as heat within the driver and may become significant as shown in table 5 for a 40W LED driver.

As shown in table 5, even if the AC to DC stage of an LED driver is at 90% efficiency and the DC to DC stage of the 40W driver is at 90% efficiency then more than 7.6 Watts of heat is generated meaning that the system has to be able to dissipate this heat to remain reliable in operation. The issue becomes significantly worse as the power of the LEDs to be driven increases. For high efficiency low wattage LED drivers (<20W) plastic cases can be used however for LED drivers >20W it is recommended that the driver case is constructed from aluminium to enable improved heat dissipation and significantly improve reliability.

Obviously the design of an LED driver becomes critical and the higher the efficiency of both stages the better however when one designs an LED driver the LED characteristics and input voltages varying during operation thus resulting in a difference between driver efficiency from design to operation. This variation becomes greater if the number of LEDs (or the forward voltage of the LED load) being driven changes significantly from te parameters used in the circuit design.

Figure 5 demonstrates what can happen when the output voltage of the LED driver varies between that chosen for the theoretical design and what may happen in practical systems. Here a typical LED driver DC to DC stage has been designed for an output voltage of 36V and using a standard IC chip can achieve a maximum efficiency of 92% within a reasonable tolerance of components and input voltage. However, if the forward voltage of the LEDs connected to the circuit (due to thermal heating or increased forward current settings) or the number of LEDs connected to the driver changes significantly the efficiency of the driver could change significantly and become much less efficient. Some driver systems can quickly degrade from 90% efficiency down to 70% or less with changes on the output of more than +/- 20% load variation resulting in excess heating of the driver and possible early failure.

The reduction in driver efficiency has a major impact on operating temperature of the LED driver and as such reduces its reliability especially for current carrying devices such as MOSFETs or electrolytic capacitors.

Electrolytic Capacitors – LED drivers Achilles Heal

The majority of high power (>15W) LED drivers employ electrolytic capacitors either on the input AC stage to enable filtering of noise or on the output channel DC stage of the driver.

It is well known that electrolytic capacitors are one of the weakest elements of a driver circuit and frequently result in failure, especially at elevated temperatures

Environmental factors affecting the service life of an aluminum electrolytic capacitor include temperature, humidity and vibration (environment), as well as electrical factors, applied voltage, ripple current and charging/discharging conditions. In capacitors for mid-to-high-voltage filters, temperature and applied voltage are the most important controlling factors. The estimated service life may be calculated based on the core temperature of the capacitor and the applied voltage.

In general a capacitance change or tangent change for loss angle indicates that the product life has been affected by temperature. Generally, as the ambient temperature (neighboring temperature of the capacitor) increases, capacitance decreases and tangent change for loss angle takes place more rapidly. This is mainly because electrolytic solution generates gas due to electrode reaction and diffuses it outside via a sealing rubber.

The Electrolytic capacitors’ properties at “hot” temperatures are well-explored thanks to the Arrhenius Law:

L = L0 × 2(Tm –T)/10 × (Vm/V)2.5

Where, L0  is the manufacturer-rated endurance at maximum temperature Tm, hours
T is the operating temperature of the capacitor, C
Vm is the maximum manufacturer-rated capacitor voltage, V
V is the operating voltage, V

Therefore, the lower the core temperature of the capacitor during actual use, the longer the estimated service life is. The core temperature of a capacitor may be lowered by lowering either the ambient temperature or the load current (operating conditions), or by either boosting capacitance or lowering internal resistance.

The useful life of an electrolytic capacitor decreases exponentially as the capacitor body temperature increases therefore it is vital that a high temperature rated capacitor is used within the LED driver and that the maximum operating temperature is well below the temperature rating.

A simple method to test the quality of an LED driver would be to measure the temperature of each of its electrolytic capacitors and determine how close the measured temperature is to the temperature rating displayed on the side of each capacitor. If the two temperatures are close then the likelihood is that the LED driver will fail very quickly however if the operating temperature is small compared to the capcitors temperature rating then the electrolytic capacitor should last a long time and increase reliability.

Furthermore, the service life of an aluminum electrolytic capacitor for mid- to high-voltage filters is affected by the applied voltage. If the applied voltage is between 60% and 100% of the rated voltage, the estimated lifetime can be extended by lowering the applied voltage below the rated voltage.

Unfortunately, there is no current definition for an electrolytic capacitors end-of-life scenario however manufacturers tend to use:

  • 10% to 20% decrease in  capacitance (measured at 120Hz) to define end-of-life
  • 200% increase in Equivalent Series Resistance (ESR) (measured at 120Hz)

Initial studies of electrolytic capacitors undertaken by Han Lei at the Lighting Research Centre demonstrates that capacitance decreases while ESR increases with time as shown by figures 7(a) and (b) for a 330μF, 5000 hour lifetime at 105oC 35V max rated capacitor.
It is important to note that the LED driver output current will increase as the capacitance decreases and the ESR increases and therefore the higher the operating temperature is the faster the output current peak-to-peak value will increase.
It is being proposed that the end-of-life of an LED driver occurs when the output current peak-to-peak value reaches 200% of the initial value enabling simple LED lifetime predictions to be extracted as shown in figure 8 for a single channel 350mA LED driver.

An LED drivers’ reliability depends upon:

  • The number and quality of components used within the driver design
  • The rated wattage of the LED driver and the maximum         operating temperature of the electrolytic capacitors
  • The overall efficiency of the AC-DC and DC-DC stages of the driver.
  • A suitable thermal management system for the driver such as an aluminium case or forced air cooling fanif appropriate.
  • Good driver design where component placement is determined by both safety, EMC and thermal considerations.
  • The ambient operating temperature where the driver is used.

Methods are being developed to enable reliability testing of LED drivers by predicting the lifetime of electrolytic capacitors used within LED drivers. Therefore, it will soon be possible to determine the good quality LED drivers from the bad.

In the meantime, I would suggest you take your LED driver and determine if it contains electrolytic capacitors and assess their respective maximum rated temperature values against the operating temperatures measured and if they are close look for another LED driver with higher rated electrolytic capacitors.

High temperature electrolytic capacitors are rated from 105oC up to 125oC and I would not recommend the use of LED drivers that do not employ such components.

Next issue
In the next issue I will look to explain what MTBF means and how it is possible to calculate a value for complex electronic products such as LED drivers. In addition, I will discuss how to practically measure the efficiency of a typical LED driver along with why the PFC of drivers is so important for overall LED fixture efficiency.

If you have any questions or comments that you would like answering then do not hesitate to contact the editor of mondo*arc or myself.



Figure 1: The high performance Cool, Neutral and Warm white LED arrays from Bridgelux in Star and Square package formats.

  • LEDs

    Table 1: Bridgelux LED Array flux characteristics showing minimum “cold” lumen output as well the Typical “Hot”
    lumen output for each product.

  • LEDs

    Figure 2: Simple block diagram of an LED Driver system

  • LEDs

    Table 2: Bridgelux arrays now compare favourably with conventional lighting technologies in terms of raw lumen output.

  • LEDs

    Figure 3: The inside PCB of a typical LED driver high voltage section.

  • LEDs

    Figure 4: The type of LED driver topologies used dependant upon input and output voltages of the system

  • LEDs

    Table 3: The types of LED driver topologies that are used in LED drivers.

  • LEDs

    Table 5: The heat dissipation for a 40W LED driver dependant upon varying AC-DC and DC-DC stages

  • LEDs

    Figure 5: LED driver efficiency fall off from a theoretical design limit

  • LEDs

    Figure 7(a) and (b) showing the decrease in capacitance and the increase in ESR of an electrolytic capacitor with increased operating temperature.

  • LEDs

    Figure 8: A typical LED driver time to failure prediction curve verses operating temperature with real data points measured.

Related Articles


Follow us on…

Follow Mondo Arc Magazine on Twitter Follow Mondo Arc Magazine on Facebook Follow Mondo Arc Magazine on Linked In

mondo arc india

darc awards DWLF IALD PLDC LRO