A successful LED system design must be able to transfer the active device’s heat efficiently from its own PN junction to the ambient. This path includes the printed circuit board on which the LED is mounted and the enclosure and/or heatsink. The designer must confirm that housings and shrouds participate efficiently in maintaining a good operating temperature.
Depending on the application, LEDs may overheat or the lighting system may not operate correctly if the airflow suddenly short-circuits because air that is sucked in by the fan and heated when passing through the heatsink accidently bypasses it, or is guided back to the intake of the fan by other features in the assembly. Alternatively, LEDs may overheat if they are positioned such that they are in stagnant air — a common problem with ceiling downlights. In some applications, the heat generated by the LEDs may be useful. De-condensation or deicing of an automotive headlight’s lens often is done by forcing the warm air that results from cooling the LEDs to blow onto the lens to heat it (Figure 1).
Predicting Hot Lumens
The new Mentor Graphics FloEFD™ thermal simulation tool with its photo-thermal LED compact model capability provides a solution to these design challenges. The compact model approach is based on using extensive measurements on the LED itself using the transient testing systems T3Ster® and TeraLED®[1-5]. These more accurate compact models based on measurements allow engineers who are creating automotive lighting, back lighting, street lighting, and general LED lighting applications to uniquely, and more quickly, calculate the forward voltage, heating power, and luminous flux that result from the given electrical current value for LEDs and LED arrays in situ. Designers can run transient simulations just by inputting what is typically known — the forward current.
The LED compact model calculates power consumption of the LED as a function of temperature, as well as what proportion of that power goes to heat and what is emitted as light. Post-processing capability allows designers to not only see how hot the LEDs become but also how much actual heat is generated and the lumen output of each LED. From this information, design engineers can confirm quickly that the lighting product meets its design goal. Without this new capability, engineers would have to define a thermal resistance model for the LED and apply an estimated heat dissipation rate based on an expected temperature for the LED without actually knowing its exact magnitude because the voltage and optical power depend on the as yet unknown temperature of the LED for that specific forward current (Figure 2).
With the LED compact model in FloEFD, designers can define the current and then calculate the temperature, which is a result of the thermal characteristic of the LED that was obtained from T3Ster, and the efficiency of the product at cooling the LED and its environment, both of which are modeled in FloEFD. From this, the luminous flux or “hot lumens” and the heat generation rate the LED has at this junction temperature and current can be obtained. The temperature of the LEDs varies based on the different currents at which they are running. These different currents result in different LED temperatures and luminous fluxes.
Faster and Better Thermal Management
A concurrent computational fluid dynamics (CFD) approach with a tool like FloEFD enables engineers to shorten the design cycle by including accurate thermal simulation for every design iteration. Unlike upfront CFD, which relies on the export of the CAD model for import into the CFD system, concurrent CFD is fully embedded within the mechanical computer-aided design (MCAD) environment, thereby eliminating the need to transfer the model with a neutral file format such as STEP or IGES which loses any parametric definition present in the original CAD model. Parametrically defined geometry aids simulations involving design variants.
Meshing and other technologies enable use of CFD technology with just the understanding of the product and its behavior. Simulation times, and particularly mesh generation, which is traditionally the longest step in the process, is reduced to a minimum.
CFD analysis is far more flexible than hardware prototyping for early design proposals and just as effective. Once a virtual model’s behavior is well-characterized, hardware prototyping can begin with a good grasp of what works and what doesn’t. In some cases, it can eliminate the need for physical prototyping.
In the past, CFD simulations were done by analysts with specialized skills in advanced math and fluid dynamics, plus a command of the complex modeling tools required for CFD work. But advancements have brought CFD technology to the engineer’s desktop, dramatically simplifying and accelerating analysis without suffering a loss in accuracy. Concurrent CFD applications automate the most demanding steps associated with preparing and running a simulation. The CFD tool embedded in a standard MCAD environment such as CATIA V5 allows designers to develop models and test the heat dispersion properties of an emerging luminaire design. The design engineer can use the dimensions and physical characteristics of the proposed design stored within the MCAD application by directly using the MCAD geometry. The CFD software detects and assigns grids to the solids and flow spaces, creating an optimized computing mesh. It aids the designer in setting boundary conditions and automatically provides solution-control settings that ensure convergence when the solver runs.
The simplest, most stable, and fastest mesh for a CFD solver is a Cartesian-based mesh. The errors caused by the skewness of cells are eliminated because of the rectangular cells but there is the general problem of mesh realization at the model’s surfaces. Commonly, the surface in these cases is either meshed with a tetrahedral cell type (hybrid mesh) or the mesh creates a step-like surface, which is not practical for free-form surfaces. One promising solution is the immersedbody mesh with a special “partial cell” technology separating the fluid from the solid part within a cell and creating subgrid control volumes. This partial cell approach recognizes the geometry with an excellent degree of detail and little to no additional effort is required on the part of the engineer to create an optimal mesh even for the most complex geometry because it can be fully automated.
Partial cells enable a complex surface to be realized by a simple Boolean operation that cuts the cell intersected by the solid surface into multiple sub-cell control volumes, each of which can have either different phases (for example, liquid-solid- gas) or same phase but different properties (such as steel-copper- aluminum) — all contained within one partial cell in the mesh. With this technology, it is possible to drastically reduce the amount of cells needed to represent accurately a complex geometry, maintain the solver stability, and reduce numerical error. Additional technologies and engineering models, such as a porous media definition for filter and other pressure-loss-causing geometries that need not be resolved in detail, improve the usability of this solution approach and further reduce the cell count. In this way, simulation is brought toward the design engineer with increased ease-of-use, but also with high accuracy.
An LED lighting application consists of many parts, from small to large, from simple shapes such as a screw or sheet of metal or plastic to complex surfaces such as reflectors and the housing with a lot of faceted surfaces or pins, ribs, and clips for final assembly. Using a traditional CFD tool to mesh these geometries would take days. But with a tool that implements the partial-cell methodology, it can be meshed within a few hours automatically. For example, in two to three runs of meshing, finding the optimal settings for the mesh in all areas of a headlight can take one to two days — not work days but calendar days — when running meshing during breaks or overnight. With one to two days for the simulation and a few hours to set up a model, it takes approximately three to four days for a complete simulation from loading the geometry the first time to a converged result. And once that is done, every change in the geometry, such as a different heatsink or fan or boundary condition, takes 30 minutes to an hour to apply the changes to the model and then run the simulation again; in one to two days, the results are done. Four to seven simulations of different scenarios can be done in the same amount of time that it would take to do one simulation with traditional CFD.
For more information, Click Here
- JESD51-14 “Transient Dual Interface Test Method for the Measurement of the Thermal Resistance Junction to Case of Semiconductor Devices with Heat Flow through a Single Path,” http://www.jedec.org/sites/default/files/docs/JESD51-14_1.pdf.
- András Poppe, Gábor Farkas, Gábor Molnár, Balázs Katona, Tamás Temesvölgyi, Jimmy-Weikun He, “Emerging standards for thermal testing of power LEDs and its possible implementation,” SPIE Proceedings 7784: 10th International Conference on Solid-State Lighting, San Diego, USA, 01/08/2010-05/08/2010, p. 778414, Paper 7784-38 (ISBN: 9780819482808).
- András Poppe, Mentor Graphics, “Thermal Characterization Confirms Real-World LED Performance,” http://www.mentor.com/products/mechanical/techpubs/request/thermal-characterizationconfirms-real-world-led-performance-59097
- András Poppe, Gábor Molnár, Péter Csuti, Ferenc Szabó, János Schanda, “Ageing of LEDs: A comprehensive study based on the LM80 standard and thermal transient measurements”, CIE 27th Session-Proceedings, CIE 197:2011: (Volume 1, Part 1–2), Sun City, South Africa, 10/07/2011–15/07/2011, Vienna: CIE, pp. 467–477, Paper OP57 (ISBN: 978 3 901906 99 2)
- Mentor Graphics, “Enhanced Turbulence Modeling in FloEFD," http://www.mentor.com/products/mechanical/techpubs/request/enhanced-turbulence-modeling-in-floefd-65206, 2011.