This article originally appeared in the June 2012 issue of SMT Magazine. Like the rest of the industry, I was swept up by the high-powered marketing campaigns promoting the performance benefits of cutting-edge dielectrics for LED PCBs. Not to mention I thought it was kind of cool making a PCB from a sheet of metal. Appealing to total cost of ownership, the wave informed us that the dielectric’s reliability performance justified the higher (purchase) cost. The wave had seemingly short-circuited the part of my brain responsible for basic physics. As a result, when the current bragged about a thermal conductivity of 2-3 W/m K, I was inhibited from realizing that this was less than 1/100 that of copper’s thermal conductivity. (400 W/m K).
In 2009, I consulted a thermal management guru who began teaching me the ways of the (thermal) force. Now, like Sid the Science Kid, I just have to know: Are brand-name thermal dielectrics on metal core printed circuit boards (MCPCBs) more effective than traditional FR-4 PCBs with Cu-plated vias?
The mechanism by which a thermally conductive PCB material works is fairly simple. The base of the LED component is soldered to a center pad of the component footprint. This center pad is not electrically connected to any of the other features. Its sole purpose is to provide a conduit for heat to be drawn away from the LED, increasing the light output as well as the life of the product. MCPCBs in their simplest form consist of a circuit layer, a thermally conductive dielectric layer, and a metal substrate (most commonly aluminum). Once the PCB is assembled it is typically mounted to a heatsink with a thermal interface material (TIM).
MCPCB materials have numerous issues associated with them. First and foremost is the lack of an industry specification. As our past webinars have discussed, this results in a wide array of potential issues to the designer—primarily incorrect specifications that leave the end customer open to failures. Further, the MCPCB material can run to 10 times the cost of standard FR-4, eroding the full potential savings of moving from incandescent to LED lighting.
This is where the lightning bolt struck me: by drilling and plating multiple vias into the heatsink pad of the LED component footprint and using traditional FR-4 two-layer plated-through-technology, could I create a PCB that rivals the thermal performance of metal-clad materials that cost up to 10 times that of standard FR-4 materials?
Being a novice in thermal calculations, I called on my trusted advisor, Clemens Lasance, who is principal scientist emeritus at Philips Research. A good portion of this career was dedicated to solving the challenge of thermal management electronic parts and systems. From the very beginning, Clemens was skeptical of vendors’ claims pertaining to thermal conductivity requirements for the MCPCB in relation to thermal performance. In fact, he claimed that the MCPCB is the last element to attack in a chain that includes TIMs, heat sinks, convections, and the LED itself.
His reply was quick and simple: When heat spreading is at stake, analytical solutions and approximate equations can only be used in limited cases where the designer is dealing with one- or two-layer problems with one-sided heat transfer and one source. Fortunately, many practical cases fall into this category, such as an FR-4 board with a copper top layer or metal core PCBs with dielectric and a metal layer. For all other situations—for example when dealing with double-sided heat transfer, or multiple sources, or more than two layers, or when vias are applied, or when layers consist of more than one material, or for which the boundary conditions cannot be considered uniform—the problem becomes intractable from an approximate analytical point of view and we have to rely on computer codes.
It should be stressed that using analytical solutions, including the 1D series resistance network, has its main merits in getting insight, hence is second-to-none from an educational point of view. However, when accuracy is at stake in the final design stages, the recommended approach for solving real-life problems is in using a 3D conduction solver.
In principle, all finite volume/finite element, etc., codes can be used that solve the 3D heat diffusion equation. In practice, only those user-friendly codes are recommended that enable a designer to get results in couple hours or so. Some popular CFD codes dedicated to the thermal management of electronic systems (FloTHERM, 6SigmaET, Icepak) used in conduction-only mode are examples of such a code.
For a fair comparison between an FR-4 circuit board with copper layers at both sides provided with vias and a metal core circuit board there is no other way than to use a 3D conduction solver. Of course experiments also can do the job, but to explore a range of parameters such as interface materials, heatsinks, convection modes, a choice of vias, copper layer thickness etc., the time gained by using numerical simulations is substantial. Estimated order of magnitude: One day for numerical simulation; one month to prepare and perform the tests. [1, 2]
Okay, so now we know we’re on the right track. I asked Clemens to break out the Crayola 64-pack and draw me a picture so I could understand the concept. He did better: He created a 50,000-cell numerical model to analyze the various constructions I was discussing with him. The explanations behind these formulae and units of measurements can be found in Clemens’ paper “Two-Layer Heat Spreading Revisited,” presented at SemiTherm, March 2012, in San Jose, California [3].
We first wanted to compare the FR-4 constructions. In addition to the standard plated through vias, we thought to model vias that are filled with solder, as well as vias that are pure copper filled for comparison purposes.
Author’s Note: Clemens made his comments in regard to the following graphs:
The following graphs are based upon an analysis using FloTHERM in conduction-only mode (hence not solving the Navier-Stokes equations), applying quarter symmetry to a model consisting of both layouts with the LED slug as the power source and a heat sink base of 2mm coupled with an interface material to the PCB in question.
Almost no difference can be found in the thermal performance of a standard via versus via filled with solder or copper. This makes sense considering the very high thermal conductivity of copper. There is, however, a huge difference between an FR-4 board that has no vias and one that has plated through vias. Knowing this baseline data, we then plotted the thermal performance of the FR-4 boards against that of a traditional MCPCB that has a thermal resistance of approximately 0.09°C in 2/W.
Almost no difference can be found in the thermal performance of a PTH FR-4 board and a MCPCB across a wide range of convection methods (h).
For years we have been pushing the idea that plated vias having >100x the thermal conductivity dielectrics should be a viable alternative. We now have mathematical proof of concept.
Now that we are armed with the confidence of having our theory proven correct via formulae, it made sense to run out and purchase a thermal imaging unit and create LED PCB test vehicles.
The copper pours are comprised of 1-inch square areas of solid copper pour. Figures 2 and 3 are screen shots of both examples.
For the two-layer designs we duplicated the 1-inch top surface copper pour on the bottom layer. The intent was to have the heat from the LED base travel through the vias, and then spread across the bottom copper pour.
After allowing a one-hour burn-in to reach normal temperature, we took thermal images from a top vantage point of each LED test vehicle. The thermal imager also allows us to take temperature reading from the hottest point, namely the LED itself. The theory is that the heat measured from the top side is heat that has not been conveyed to the heatsink via the thermally conductive dielectric (in the case of the MCPCB) or the PTH (in the case of the two-layer FR-4 boards) (Figure 5).
The values in the chart represent surface temperature measurements only. However, it is more critical to calculate the temperature rise from ambient to glean the true performance of the thermal stackup, as shown in Graph 4.
When considering alternative solutions, it’s always important to know what exactly will be the benefit. For the bean counters, it boils down to money saved. If we can assume that a particular design will perform the same between a Metal Core PCB design and a plated via FR-4 design then can I save enough from a cost standpoint to compensate for additional testing and UL-related costs associated with converting my product over? Let’s see (Graph 5).
Graph 5: *Values will not calculate exactly due to rounding in the piece price.
Our test results demonstrate that FR-4 PCBs with Cu-plated vias achieve thermal transfer roughly equivalent to that of metal core PCBs with thermally-conductive material. Please note that the results are specifically for this set of constraints including number of vias, LED type, and choice of MCPCB. Any designer wishing to convert their MCPCB design to use FR-4 with plated vias should not only contact a thermal management consultant for part-specific calculations, but also perform physical tests as we have done here. Should you need, Clemens Lasance can be contacted at pcb@frankenthalerfoundation.org.
From a cost standpoint, the potential savings are substantial. The raw material savings alone more than compensate for the added processing steps required to create a two-layer plated through PCB. Further, it is difficult to find a supplier that is UL certified to particular MCPCB material; conversely, using FR-4 technology allows the end user to add a great deal of competition to the supply base…as most PCB Fabricators are already UL-certified to use FR-4 materials.
