Basic Guidelines for LED Lamp Package Design
Song Jae Lee
Electronics Engineering Department, Chungnam National University, Daejeon, Korea
To cite this article:
Song Jae Lee. Basic Guidelines for LED Lamp Package Design. Basic Guidelines for LED Lamp Package Design. Vol. 4, No. 5, 2015, pp. 187-194. doi: 10.11648/j.ijrse.20150405.13
Abstract: Even though significant amount of researches has been done to develop LED lamp packages for improved performance especially in terms of output power, it is believed that no standard theories or guidelines have been established yet for designing LED lamp packages. In this paper, both the InGaN/Sapphire LED chip structure and its Epi-Up or Epi-Down chip-mounting scheme have been analyzed by using Monte Carlo photon simulation method. Based on the analysis, we have established guidelines for designing LED lamp packages.
Keywords: LED, LED Lamp, LED Lamp Package Design
Even though visible light-emitting diodes (LEDs) have been commercially very successful, it is believed that the basic principles or guidelines for designing LED lamp packages may not have been well established yet.
Conventional LED lamps in general may fall into either the leaded type or SMD (surface mount design) type as schematically described in Fig. 1. The leaded type has relatively long leads that are often to be inserted through the holes made in the printed circuit board (PCB) and soldered at the bottom. On the other hand, the SMD type has structures that are more suited to be attached directly on the top surface of the PCB. Even though the leaded type LED lamps are still used extensively, for instance, in applications such as traffic lights and outdoor displays, the general trend seems to favor ever more the SMD type LED lamps.
A serious problem in the SMD structure shown in Fig. 1(b) may be that the heat generated in the LED chip is not dissipated easily. In general, both the molding compounds encapsulating both the LED chip and the dielectric chip mount have relatively poor thermal conductivities. Thus, in order to prevent severe heating in the junction of the chip, the LED driving current should be kept relatively small leading to very limited light output. In general, the more the junction temperature rises, the more likely the injected electrons and holes recombine nonradiatively or overflow to the carrier confinement layers and lead to significantly decreased internal quantum efficiency.1 In addition, the rise of the junction temperature would in general shrink the bandgap energy of the active layer, which in turn shifts the spectrum to the red and changes the color of the light output.
The key component of the LED lamp is obviously the LED chip. Thus it would be a first task in LED lamp design to accurately analyze LED chips. Fig. 2 shows a typical structure of the InGaN/Sapphire LED chip, which currently enjoys a commanding position in visible LEDs. They have excellent reliability and brightness. Furthermore their emission spectrum from ultraviolet to amber is much broader than that of the AlInGaP/GaAs system that emits mostly in the red spectral region.2 It is noted also that the blue emission from InGaN/Sapphire LEDs is exploited to implement the white LEDs that are essential for the general lighting. As well known, blue LEDs can also be fabricated by using the InGaN/SiC system. However, they may hardly compete with the InGaN/Sapphire LEDs especially in terms of maximum light output and reliability.
InGaN/Sapphire LEDs are grown on the non-conducting sapphire substrate and, as a result, their chip structures are quite unique in that the two electrodes both formed on the epitaxial side are displaced from each other in the lateral direction. Another important feature of the electrode design is that a very thin semi‑transparent p‑ohmic material is deposited on the top surface of the p‑GaN carrier confinement layer outside the p-electrode pad. The thin ohmic material is to compensate the very low conductivity of the p-GaN carrier confinement layer and prevents the driving voltage from rising too much. The deposition of the thin ohmic material outside the p‑electrode would make the isolation distance between two electrodes as small as typically 10~20. It is noted that this small separation between the two electrodes would make it crucial to align the chip very accurately when the LED chip is to be attached on the chip mount in the Epi‑down (flip chip) mode. We will discuss this problem in more detail later.
In this paper we first analyze both the InGaN/Sapphire LED chip structure and chip‑mounting schemes of either Epi-up or Epi-down. Then, based on the analysis, the basic guidelines for designing LED lamp packages would be made.
(a) Lead type
(b) SMD type
(a) Side view
(b) Top view
2. Different Types of InGaN/Sapphire LED Chips
The first task in the LED lamp package design may be to determine the type of the chip to be employed in the lamp. In the viewpoint of LED lamp design, the InGaN/Sapphire LED chip may be classified into the three types illustrated in Fig. 3. The structure in Fig. 3(a) is a regular chip, which is commonly used in practical LED lamps and has a size of typically . With such standard‑sized chips, the maximum light output power achievable would be seriously limited. Thus many researchers have been tempted, for the purpose of enhancing the maximum light output per unit LED lamp, to try large‑sized chips as in Fig. 3(b). The structure in Fig. 3(c) is the so‑called vertical structure, in which the sapphire substrate with a relatively poor thermal conductivity is removed and the p‑electrode and n‑electrode formed on top and bottom, respectively are displaced in the vertical direction.
The above rather arbitrary classification of the InGaN/Sapphire LED chips may be valid if each structure shows its own distinctive characteristics. For instance, the vertical structure in Fig. 3(c) has a very small thermal resistance as resulting from the removal of the sapphire substrate of a poor thermal conductivity. Another important feature that should be taken into account in choosing the LED chip would be the photon extraction efficiency (or photon output coupling efficiency) . The external quantum efficiency , which is the most important parameter to determine the overall light‑emitting efficiency of LED, is related to the photon extraction efficiency as
where is the internal quantum efficiency. It is noted that the photon extraction efficiency is determined mostly by the detailed structure of the chip, whereas the internal quantum efficiency is determined mostly by the quality of the epitaxial layers in the chip. Thus, it is crucial in lamp package design to maximize the photon extraction efficiency .
In this work the photon extraction efficiency was calculated by using the Monte Carlo photon simulation method that has been described elsewhere and used extensively to analyze the LED chips and lamps.3,4 In the calculation we have assumed that all the chips are mounted on the surface with the photon reflectivity of 80% and encapsulated by the molding compound with reflective index of 1.50. Other important chip simulation parameters are basically the same as those used in the previous works.3, 4
Fig. 4 shows the photon extraction efficiency calculated as a function of the chip size. As the chip size increases, the average generated photons have to travel a longer distance to escape the chip. And the probability for the photons to be absorbed increases exponentially with the traversal distance. Consequently, the photon extraction efficiency should decrease with the chip size.
In order to compensate the degraded photon extraction efficiency in enlarged chips, the driving current should be increased by more than proportionally to the increased chip area. Consequently, the heating per unit chip area would be in general more severe in enlarged chips than in small area chips. Without significantly improved heat sinking, the junction temperature rise would be larger in enlarged chips than in small area chips. As discussed already, the internal quantum efficiency decreases in general with the junction temperature. Taking into account all these effects, it would be reasoned that the external quantum efficiency in general degrades more rapidly with the chip size than the photon extraction efficiency .
Based on the above reasoning, we may reach an agreement that the approach to increase the chip size too much in order to increase the per‑unit light output would seriously compromise one of the most important advantages of LEDs, i.e., the high efficiency. Thus, a rather practical approach to increase per‑unit light output may be to combine an LED chip of relatively small size with a good heat sink and thereby increase both the driving current and light output.
One of the important issues regarding the vertical structure in Fig. 3(c) is how the substrate removal affects the photon extraction efficiency . Even though the removal of the sapphire substrate is certainly beneficial in terms of heat dissipation, it would in general degrade the photon extraction efficiency. In other words, the sapphire substrate in the regular structure helps to increase photon extraction efficiency. The refractive index of the sapphire substrate is about 1.77 and is between 2.48 and 1.50 that are approximate refractive indices for the GaN carrier confinement layer and the molding compound, respectively. Such index‑matched substrate is to help enhance the photon extraction efficiency. It is noted, however, that the index-matching effect of the substrate occurs only to the photons transmitted into the substrate and reach the sidewall of the substrate either directly or by way of reflection off the bottom of the substrate. The thinner the substrate, the larger fraction of the photons that are transmitted into the substrate would be reflected off the bottom of the substrate and transmitted back into the GaN confinement layer without touching the sidewall of the substrate. As a result, as shown in Fig. (5), the photon extraction efficiency in the regular InGaN/Sapphire structure decreases as the thickness of the sapphire substrate decreases.
Following the above reasoning, the photon extraction efficiency in the vertical structure with the sapphire substrate completely removed should similarly be poorer than in the regular structure with the substrate remained intact. For instance, the photon extraction efficiency in the vertical structure is estimated to be about 0.334 and is quite smaller than the photon extraction efficiency 0.564 estimated of the regular structure with the sapphire substrate of thickness 100 . A rather interesting point to note here may be that the photon extraction efficiency in the vertical structure is even smaller than the photon extraction efficiency 0.435 estimated of the regular chip with the sapphire substrate completely removed as in the vertical structure. The poor photon extraction efficiency of the vertical structure compared to the substrate‑removed regular structure is due to the difference in the electrode design. In the case of the vertical structure, the photons would be generated more likely in the central region of the actively layer and they have to travel a longer distance to reach the edge of the chip. On the other hand, in the regular structure, the photons would be generated more likely in the region near the edge of the chip and they will couple out of the chip more easily. Even though the photon extraction efficiency in the vertical structure could be improved somewhat, for instance, by optimizing the electrode pattern design5, it may not easily surpass that of the regular structure with the substrate completely removed. Based on the above reasoning we may reach an agreement that when the vertical structure is chosen instead of the regular structure, the benefit of its very low thermal resistance should be more than enough to compensate the degraded photon extraction efficiency.
3. Chip Mounting Scheme
(a) Standard-sized regular chip
(b) Large-sized regular chip
(c) Vertical structure
(a) Epi-Up on metallic chip mount
(b) Epi-Down on dielectric chip mount
(c) Epi-Down on metallic chip mount
(d) Epi-Down on Si-submount
Once the type of the LED chip is chosen to be employed in the lamp packaging, the next important task may be to decide how to attach the chip on the chip mount. In general the performance of the chip, in terms of both the heat dissipation and the photon extraction efficiency , would depend significantly on the chip mounting scheme. Fig. 6 shows the regular chips attached on the chip mount either the epitaxial side up or down. It is noted that, in the case of the vertical chip with the sapphire substrate completely removed, the distinction between up and down will not have much meaning.
In the Epi-up mounting scheme, in which the insulating sapphire substrate of the chip is made contact directly with the chip mount, a plate of bare metal of high thermal conductivity can be utilized as the chip mount as shown in Fig. 6(a). In this configuration, the two electrodes of the chip will be connected by the bond wire to the current leads that would be formed separately somewhere else in the package. In the case of Epi‑down chip mounting scheme, in which the epitaxial side of the chip is made contact with the chip mount, the chip mount should be equipped on top with two electrically isolated current leads that are to be soldered to the respective electrodes of the chip. The electrically isolated current leads may be formed either on a bare dielectric plate as in Fig. 6(b) or on a dielectric layer deposited on a metallic plate as in Fig 6(c). It is noted that either the dielectric plate or layer in the Epi‑down chip mount would seriously hamper the heat dissipation from the LED chip junction to the chip mount, which would also function as a heat sink, and thus somewhat compromise the thermal advantage unique in the Epi‑down scheme.
However, the most critical disadvantage of the Epi‑down scheme may be that it can hardly be applied to mass production, due to the difficulties in attaching the LED chip on the chip mount. As mentioned in the previous section, the lateral separation between the two electrodes in the regular InGaN/Sapphire LED chip is as small as typically 10~20, and thus it is critical to align the chip on the chip mount very accurately so that the two electrodes on the chip are in proper contact with the respective current leads on the chip mount. It is noted, however, that the margin of positioning error of typical die bonders that are often employed in the LED manufacturing industry is usually much larger than 10~20, the typical separation between two electrodes.
One of the approaches to overcome the alignment problem in Epi‑down scheme may be, for instance, to attach the LED epitaxy wafer, before being diced into separate chips, on a Si wafer that has current lead patterns formed on top, by using a mask aligner that is widely used in semiconductor device fabrication. The resulting epitaxy‑Si integrated wafer is then diced into separate chips as shown in Fig. 6(d). It is noted, however, that in chip dicing process a portion of the current lead patterns on the Si wafer should be exposed for the bond wire connection. The overall processes may be too difficult or expensive for them to be practically employed in the LED industry.
Another important issue regarding the chip mounting scheme is how it affects the photon extraction efficiency . Fig. 7 shows the photon extraction efficiency calculated as a function of the chip size both in the Epi‑up and Epi‑down chip mounting schemes. A rather intuitive prediction may be that the photon extraction efficiency would be generally higher in the Epi‑down scheme in which the sapphire substrate is completely cleared of the interference from the chip mount. However, in our calculation in Fig. 7, when the width of the chip is relatively small, i.e., below about 260 , the photon extraction efficiency in the Epi‑down scheme is actually poorer than in the Epi‑up scheme.4
In order to properly understand how the chip mounting scheme really affects the photon extraction efficiency, it is necessary to consider the effect of photon trapping inside the epitaxial region as a result of the total internal reflection off the sapphire substrate interface. As discussed in the previous section, the refractive index of the sapphire substrate is about 177 and is considerably smaller than the refractive index of the GaN carrier confinement layer, about 2.48. Thus a significant fraction of the emitted photons, i.e. the ones travelling roughly parallel to the epitaxial layers, would be reflected total‑internally off the sapphire substrate boundary and be trapped inside the epitaxial region. And the rest of the emitted photons, i.e. the ones travelling rather perpendicularly to the epitaxial layers would avoid the total internal reflection and mostly transmit into the substrate region.
An important point is that the two categories of the photons escape the chip through different windows in each chip mounting scheme. First, for the photons trapped inside the epitaxial region, only the sidewall is opened in the Epi‑down scheme, whereas in the Epi‑up scheme the top surface of the chip where the thin semitransparent p‑ohmic material is deposited is also partially opened in addition to the sidewall. Thus, for the photons trapped inside the epitaxial region to escape the chip, the Epi‑up scheme is preferred over the Epi‑down scheme. Next, for the photons transmitted into the sapphire substrate, only the substrate sidewall is opened in the Epi‑up scheme, whereas in the Epi‑down scheme both the substrate sidewall and substrate bottom surface that is now directed upward and therefore cleared of the chip mount are opened. Thus, for the photons transmitted into the sapphire substrate to escape the chip, the Epi‑down scheme is preferred over the Epi‑up scheme.
Each chip mounting scheme, as reasoned above, having its own unique advantage but enjoyed by only a fraction of the photons, i.e., the photons trapped inside the epitaxial region or the photons transmitted into the sapphire substrate, it is not simple to judge which chip mounting scheme is preferred for the average photons to couple out of the chip. However, our calculation in Fig. 7 shows that when the chip size is relatively large the Epi‑down scheme has higher photon extraction efficiency than the Epi‑up scheme. It is noted that, in this regime of large area chip, most of the output photons would have passed through the upward windows, which have much larger area than the sidewall windows. Consequently, in the Epi‑up scheme of this regime, the photon extraction efficiency would be determined largely by the amount of the photons trapped inside the epitaxial region as well as the transmittance of the thin ohmic material deposited on the top surface of the chip. Similarly, in the Epi‑down scheme the photon extraction efficiency will be determined largely by the amount of the photons transmitted into the substrate. It is noted that the approach to grow the epitaxial layers on the roughened surface of the sapphire substrate would increase significantly the number of photons transmitted into the substrate and thereby improve significantly the photon extraction efficiency in the Epi‑up scheme.4
Another important point in Fig. 7 is that when the chip size is below about 260 the Epi‑up scheme has higher photon extraction efficiency than the Epi‑down scheme. It is noted that, in the Epi‑up scheme of this regime of small area chip, even the photons transmitted into the substrate are able to couple out of the chip through the widely opened substrate sidewalls.
4. Combinations of LED Chip and Chip Mount
(a) Regular chip on metallic chip mount
(b) Vertical chip on metallic chip mount
(c) Regular chip on dielectric chip mount
(d) Regular chip on dielectric chip mount
Based on the previous discussions regarding both the type of LED chip and chip mounting scheme, we may reach the following agreements. First, the large area chip in Fig. 3(b) with the photon extraction efficiency seriously degraded may not be used better, except for special applications requiring large light output of short duration as in flash lamps for digital cameras. On the other hand, the vertical chip in Fig. 3(b) also with the photon extraction efficiency seriously degraded has a very low thermal resistance and thus it would be possible, when combined with a good heat sink, to increase significantly the driving current and thereby the light output. Regarding the chip mounting methods, the Epi‑down scheme requiring too complicated chip attaching process may hardly be adopted in the LED manufacturing industry. Furthermore, the Epi‑down scheme may not be preferred much over the Epi‑up scheme even in terms of the photon extraction efficiency, especially when the chip size is relatively small as is the case in most practical LED lamps. Thus, both the large area chip and the Epi‑down chip mounting scheme not being allowed in LED lamp package design, most practical LED lamp structures would then be obtained by attaching, in Epi‑up mode, the two types of the chip, i.e., the regular chip or the vertical chip on the two types of the chip mount, i.e., the metallic mount or the dielectric mount. The resultant four types of LED lamp structures based on chip‑chip mount combination are schematically described in Fig. 8.
One of the important output characteristics of LED lamps may be their light out power vs. current curve. In general the output power of an LED lamp would depend very complicatedly on various parameters of the chip and chip mount employed, and thus it is very difficult to generalize the current‑output power characteristics for various LED lamp structures.6 Thus, here we may better present the schematic current‑output power curves that are expected typically of the four types of LED lamp structures in Fig. 8.
The LED lamp structure in Fig. 8(a) consists of the regular chip attached on the metallic chip mount. In this case, the overall thermal resistance form the junction to the bottom of the chip mount would be relatively large, due to the relatively thick sapphire substrate with relatively poor thermal conductance in the heat dissipation path. Thus, when the driving current is relatively low and therefore the heating in the junction of the chip is not very severe, the output power would increase approximately linearly with the current. However, when the driving current is increased enough, the large amount of heat generated would not be dissipated easily and as a result the output power would start to saturate with the current.
In the LED lamp structure in Fig. 8(b), which consists of the vertical chip attached on the metallic chip mount, the overall thermal resistance form the junction to the bottom of the chip mount would be very small. Thus, the saturation of the output power would occur at a much larger current level than in the structure in Fig. 8(a). It is noted, however, that the output power level in the linear region is actually smaller than in the structure in Fig. 8(a), due to the photon extraction efficiency seriously degraded in the chosen vertical structure. The structure in Fig. 8(b), however, has a significantly extended linear region compared to the structure in Fig. 8(a), and as a result, a power‑crossover point is often observed between the two structures.
The LED lamp structure in Fig. 8(c) consists of the regular chip attached on the dielectric chip mount. In this case, the overall thermal resistance form the junction to the bottom of the chip mount would be extremely large due to the added thermal resistance coming from the dielectric chip mount. As a consequence, the output power would start to saturate at an even smaller current level than in the structure in Fig. 8(a). Another important point is that the output power level is also usually smaller than in the structure in Fig. 8(a). In general the dielectric chip mount has poor photon reflectivity compared to metallic chip mount, and thus the more output photons incident on the chip mount surface would be absorbed leading to reduced output power level.
Lastly, the LED lamp structure in Fig. 8(d) consists of the vertical chip attached on the dielectric chip mount. In this case, even though the sapphire substrate is eliminated in the heat dissipation path, the overall thermal resistance form the junction to the bottom of the chip mount would be still very large. Thus, the output power, which is already compromised as result of the degraded photon extraction efficiency in the chosen vertical structure, would start to saturate at almost the same current level as in the structure in Fig. 8(c). Thus, the power‑crossover point would in general not be observed between the two structures.
5. About Chip-On-Board Packaging
In real applications of LED lamps, they would be installed often on a special printed circuit board (PCB) called the metal core printed circuit board (MCPCB). The MCPCB with improved thermal characteristics would help not only increase per‑unit output power but also stabilize the various LED output characteristics, such as the spectrum distribution and reliability. Fig. 10(a) shows a unit LED lamp put on the MCPCB. Considering only the thermal characteristics, however, it would be much more helpful to install LED chips directly on PCB (COB: chip on board) as shown in Fig. 10(b). Since the PCBs on which LED chips are installed are usually much thicker and wider than the chip mounts in unit LED lamps and thus have much larger heat capacity compared to the chip mount. Thus the direct attachment of the LED chip on PCBs would lead to drastically improved output power and stabilized output characteristics.
(a) Unit LED lamp on PCB
(b)LED chip on PCB (Chip-On-Board)
It is noted, however, that the COB scheme may not be practical enough to be adopted extensively in the LED industry for various reasons. First the PCB on which LED chips are installed, possibly with other electrical or mechanical components, will have a shape, size, or design that may in general be widely different depending on the particular application. As a consequence, it would be extremely difficult or expensive to automatize, on this type of nonstandard PCBs, the various LED lamp manufacturing processes, such as chip attachment and encapsulation with molding compound. Similarly the testing of the each LED lamp for the important output characteristics, such as the output power, beam pattern, spectrum, and reliability cannot be easily automatized on the nonstandard PCB. Furthermore, if any of the LED lamps on the PCB fails to meet any of the various specifications required in the particular application, it would be very difficult or nearly impossible to repair or replace the failed one that has been molded strongly on the PCB.
On the other hand, in the case of the unit LED lamp installed on the PCB as shown in Fig. 10(a), all the manufacturing processes would be performed on standard lead frames, whose shape, size, or design would be determined independently of the particular applications, and would easily be automatized. The resultant uniform unit LED lamps of the same shape and design would then be tested using the standard and automatized testing procedures. Lastly only the unit LED lamps meeting all the required specifications would be installed on the PCB and be replaced quite easily, if any problem occurs after installment.
(a) Chip mount without isolating inter-walls
(b) Chip mount with isolating inter-walls
Using Monte Carlo photon simulation method, the performance of the regular chip, vertical chip, and large area chip are analyzed and compared in terms of the photon extraction efficiency. In addition, the Epi-Up and Epi-Down chip-mounting schemes are analyzed and compared in terms of both the photon extraction efficiency and the difficulties of implementation. Based on the results of the analysis, we have established the basic guidelines for the LED lamp package design as follows:
• LED chips with chip area considerably larger than about have the photon extraction efficiency seriously degraded. Increasing the driving current by more than proportionally to increased chip area in order to compensate degraded photon extraction efficiency, would inevitably lead to junction temperature rise, which in turn would tend to decrease even the internal quantum efficiency too. Thus, the large area chips are in general not desired, except for special applications requiring large output power for a short duration of time.
• The approach to increase the output power or photon flux density by integrating multiple chips in one package as shown in Fig. 11(a) would also suffer from the degradation of the photon extraction efficiency as a result of the significant optical coupling between neighboring chips. In order to eliminate the in‑between optical coupling, the chip mounts with isolating inter‑walls as shown in Fig. 11(b) should be used.
• The regular chip has higher photon extraction efficiency than the vertical chip, as a compensation for the increased thermal resistance. It is always preferred over the vertical chip in applications requiring low output power that is to be achieved with minimal heating.
• The vertical chip has a much lower thermal resistance than the regular chip, as a compensation for the degraded photon extraction efficiency. In order to fully exploit its thermal potential for increasing the driving current and thereby output power, it should be mounted on metallic chip mounts with improved heat sinking capacity.
• Dielectric chip mounts with a larger thermal resistance also have usually a poor photon reflectivity leading to significant optical loss, and therefore they would not in general be suited for applications requiring large output power. Especially it is not very sensible to attach the vertical chip on the dielectric chip mount since the approach leads to both poor efficiency and poor thermal behavior.
• The Epi‑down chip mounting scheme requires a chip attachment process that is so delicate that it can hardly be adopted in the LED industry. Furthermore, it is not preferred over the more conventional Epi‑up scheme even in terms of the photon extraction efficiency, when the LED chip has a smaller area than about .
• The COB scheme, despite the advantages of increasing output power and better stabilizing output characteristics, cannot be used extensively in the LED industry due to the difficulties both in implementing it and in replacing the failed LED lamps.