The last few years have seen the development of larger, higher power, higher luminescence light-emitting diode (LED) chips. Prior to this, LEDs were typically 0.020" x 0.020", came in various colors and were used primarily as optical indicators. This new breed of LEDs come in various colors, including white and UV, are typically 0.040" x 0.040" and are designed to provide high intensity light for illumination applications. Individually packaged versions of these LEDs are being used in arrays and clusters to replace standard incandescent and compact fluorescent light bulbs and are much more efficient at converting electrical power into usable optical power.
There are applications in the medical field where very high intensity, densely packaged arrays are required for illumination and curing applications. Examples of these are disinfection lamps, dental curing, body scanning equipment, blood analysis, neonatal jaundice treatment, and even lighting for surgery.
There are various challenges in packaging high density, high luminescence arrays of chips — increased electrical power dissipation density and the resulting need for efficient thermal management. The lower the temperature the LED chips are subjected to, the longer will be the expected lifetime of the product.
LED arrays are usually put together in a configuration shown in Figure 1. The die is attached to a carrier substrate using a die-attach adhesive. High power LEDs typically have a metallized backside and, since this is the cathode, it must be electrically connected to a conductive pad on the substrate. Materials used for this purpose can be conductive epoxies or solders. Eutectic bonding could also be used but the high temperatures required (>300° C) and the time required to connect all elements in the array exceed the time/ temperature withstanding capabilities of most LED chips, typically 280° C for 10 seconds.
The metallized carrier substrate must be capable of efficiently transferring the heat through and across the substrate and providing electrical conductors of low enough resistance to carry the large electrical currents required to drive the LED chips.
The heat sink compound is required to transfer heat from the back of the substrate carrier to the metal heat sink. These materials should ideally compensate for surface imperfections in the heat sink and substrate and for differences in their camber. These materials are typically thermally conductive grease or epoxies.
The heat sink itself spreads the heat over its surface so that the heat may be removed by air or liquid flow. Heat sinks are usually made of highly thermally conductive metals, such as copper or aluminum. The physical design is done to maximize surface area.
Many such designs are possible. The experimental model used for this study is shown in Figure 2. Sixteen blue, 1 mm square, LED chips are connected in a 4 x 4 array (4 parallel chains of 4 LEDs in series). There are four 0201 thermistor chips attached to the assembly — one on top of a centrally located and one on an outer LED chip, one centrally located on the substrate and one at the periphery of the substrate. Each LED chip has two 0.001" gold thermo-sonically bonded wires connecting the anode surface to the substrate conductors. The aluminum heat sink was attached to the substrates by two retaining screws.
The intent of this study was to determine the relative merits of different substrates, different die attach materials, and different heat sink compounds. Vendors of these elements were asked to recommend their most suitable materials.
A Taguchi “design of experiment” was performed using:
- 96 percent Alumina with thick film conductor traces.
- Insulated Metal Substrate (IMS) with a copper core and 5 mil nickel/gold plated copper traces.
- Aluminum Nitride Substrate (ALN) with 5 mil nickel/gold plated copper traces. (Note that Beryllia (BeO) ceramic is an exceptionally high thermally conductive substrate material but, because of its toxicity issues, was not considered for this study).
- Die Attach Materials:
- A conductive epoxy (Vendor A).
- A conductive epoxy (Vendor B).
- SAC305 solder paste.
- Heat Sink Compounds
- A thermal grease (Vendor C).
- A thermal grease (Vendor D).
- A thermally conductive epoxy (Vendor E).
Two sets of tests were done on the experimental models. In the first set, the test samples were held vertically (leads down) in free air and powered by applying a constant current of 400 mA to each array. The four thermistors were measured by automated test equipment every three seconds for 12 minutes (until the temperature stabilized). At this point, the power was cut and the thermistors recorded for a further 2.5 minutes during the cool down phase.
In the second, an infrared (IR) camera recorded the heat emissions from the assembly during ramp-up and ramp-down so that the isotherms on the rear of each substrate type could be studied. For this test only, the heat sink was not attached.
Results and Conclusions
While a large amount of data was collected and analyzed, some general conclusions became apparent:
- ALN is by far the best substrate material of the three. IMS is second and Alumina third (See Figure 3).
- SAC 305 solder for die-attach is superior to both conductive epoxies used even though these were designated as “best in class” by their manufacturers and superior to solder.
- Although there are performance gains given by thermal greases as heat sink compounds over thermally-conductive epoxies, this is not as major a contributor to the thermal performance as the substrate or die attach selections.
- The data derived from the thermistor measurements and from the IR camera pictures shows that the transverse heat flow across ALN and IMS significantly spreads the heat more evenly across the substrate surface reducing the centrally located “hot-spot” seen on the Alumina substrate. (See Figure 4)
While the three substrate types tested have far superior thermal conductivities to glass/epoxy laminates, often used to interconnect LEDs, there are cost implications and these have to be weighed against the required light intensity output and thermal performance of any actual product.
This article was written by Andy Proudfoot, Vice President of Technology at TASK Micro-Electronics, Inc., in Montreal, Canada. For more information, Click Here