This document serves as a write-up of the design progression as of January 28, 2013. The approach taken is to look at the design options from a functional requirements perspective, assessing the engineering requirements to build a functioning fixture. This document outlines several design options for a liquid cooled fixture. The performance and size characteristics of a given design are assessed, and then compared against air cooled design options of the same size constraints providing equivalent or better light output and beam profile. The final conclusion of this analysis is that the only situations where liquid cooling can be justified are for high brightness flood fixtures and high brightness linear fixtures with tight beam control in 1D and a very broad beam in the other direction. Alternate cooling technologies such as heat pipes, air jet impingement, etc. have not been considered as they have been interpreted as out of the scope of the IRAP project mandate as per previous discussions.
It is preferable to use the most robust design option (fewest failure points and longest MTBF) possible for a given form factor. If the reliability of two designs is expected to be similar, the cheapest design option will be preferred. Wherever possible, modular designs will be used to enable multiple configuration options (e.g. cooling type, optical beam profiles, power supply options, etc.) from a single light engine. The electrical design and circuit layout should be designed to be fault tolerant wherever possible, with graceful failure modes designed in. Examples of this include temperature feedback to throttle the power delivered if the cooling capacity is degraded by pump or fan failure, and a circuit design such that blowing one LED will not cause the entire array to go dark.
- Manufacturing and installation of a liquid cooled system will be more complicated and time consuming compared to traditional CRS Electronics fixtures. There are numerous additional fittings, hoses, and cables to be connected and routed, and each system will need to be pressure tested (typically systems are pressure tested over a multiple hour period) and filled with coolant in house before delivery. In order to approach the temperature ranges specified (-50°C to +45°C), a 50-50 mix of ethylene glycol (toxic to animal and plant life) and distilled water is necessary.
- Forced air convection, with one or more fans as the only moving parts, is inherently more reliable than liquid cooling which requires pumps, radiators, fans, and multiple fluid fittings providing many more possible failure points. Additionally, a forced-air heat sink would still function moderately well if one or more fans failed, allowing a graceful throttling back on light output. If the pump on a liquid cooled system fails the heat transfer capability without coolant flow is substantially worse. Thus, if a forced air cooled system can be designed in the same space constraints as the liquid cooling option the forced air design is preferred, cost permitting.
- Heat pipes require only simple fans to function, and as such should be inherently more reliable than liquid cooling. If there is no functional requirement for remote cooling further than 6” or 12” away from the heat source, heat pipes are preferred over liquid cooling. Given the slight increase in complexity over forced air designs, air cooled solutions are preferred over heat pipe systems if they can be constructed in the same space and cost constraints.
- Optical control of an LED array is difficult as a fundamental law of optics limits the beam angle from a distributed source. The only viable option for efficient beam shaping of a distributed source or LED array is a parabolic or compound parabolic reflector.
- To collimate a light source using a lens system, the lens array will need to be one focal length away from the array. In this configuration the beam angle from the system is given by (See figure 1, below). In order to collect the majority of the light from the array we require >60° (for a 120°FWHM LED output) which means and also requires a lens assembly with a diameter to focal length ratio > 1 ( ) which is only achievable with a multiple element lens assembly (expensive, and additional air-glass interfaces will add significant optical losses to an already inefficient system.) Thus, collimating a distributed light source with a single lens system to get long projection distances can be done only with very low optical efficiency. Where high system efficacy is required a lens-based system is not viable.
- For a simple parabola the focal length, rim diameter, and depth are related by . While the optimal design would likely be a compound parabola or a defocused simple parabola to get a better beam profile (optical simulations would be required to optimize the reflector shape), the size requirements will be similar to the simple parabolic reflector and we can the simple case for estimating the size requirements. In the case where we are looking to collimate an LED array the array should be located close to the focal point. Setting the radius of the reflector at the focal plane equal to the radius of the LED array, the focal length of the parabola is . The nominal beam angle for the reflector is then given by . Solving for the reflector radius for any given beam angle and LED array radius, and then using this to determine the reflector height is straightforward. It is clear that a reflector for a large LED array will need to be prohibitively large to attain a tight beam spread.
- In sizing the overall system, the power consumption has been limited to 80% of the available power from a 30A 120V AC circuit (equivalent to two standard 15A household circuits) as this is commonly available. Sizing the power requirements to 80% or less of the rated power is dictated according to the electrical code. For a 120V 30A circuit at 80% load the usable power available is 2880W.
- In selecting the heat exchanger for the system we have two design approaches to choose from. Selecting equipment designed for industrial environments and capable of being mounted exposed to the elements will cost approximately $710 for the radiator and required fan(s). Choosing equipment designed for computer cooling applications (requiring placement in a well-protected enclosure) will cost roughly $280. In the cost estimates below, the industrial approach has been selected.
- The additional power requirements for pumps and fans will impact the efficacy of the system. The preferred AC 10” Tubeaxial Ostro fan runs at approximately 33W, and the preferred Iwalki RD-05H pump runs at approximately 18W. This accounts for a total of 51W additional power, reducing system efficacy by approximately 2% for a 200,000 lumen fixture.
- Design concepts are initially based on assessing the number of LEDs required to hit a target output level and efficacy given assumed optical and electrical efficiencies. This solution size is approximated using the maximum LED packing density possible accounting for manufacturing tolerances on the LEDs and pick and place machine clearance and precision (3.45mm chips can be centred every 4.15mm).
- Thermal requirements are roughed out to assess feasibility and whether or not it fits under the product umbrella (e.g. whether liquid cooling or heat pipe cooling is necessary). If advanced cooling technologies are not necessary write up the design option and shelve for now.
- Assess required cooling assembly size (size required for fittings, manufacturability, etc.) against break point for air vs. liquid cooling. If size requirements for the liquid cooling module are large enough that air cooling becomes effective, write up design option using air cooling technology and shelve for now.
- Rough out the optical design for general purpose beam angles (flood, and 40°-60° spot). If size requirements for the optical components are large enough that air cooling becomes effective, assess if enough room is available to space out the LED array and use individual LED optics, write up design option using air cooling and shelve for now.
- Given approximately equal pricing per Watt for power supplies comparing multi kilowatt supplies versus off-the-shelf constant current LED drivers, there is little incentive to design a custom power supply. Adjust the number of LEDs and LED layout to fit an integer unit of off-the-shelf LED drivers.
- Lay out PCB design, including thermistor feedback to avoid catastrophic failure if cooling pumps, fittings, etc. fail or system otherwise reaches an unsafe temperature. Circuit design should be done so that the system fails gracefully, throttling back to provide the max amount of light possible without over-heating the LEDs during the fault condition. If multiple strings are used per power supply, design so that if one LED string blows the remaining strings can handle the additional power without immediately self-destructing.
- Assess design for suitability to target applications and markets. If market analysis is promising, prototype and test performance, manufacturability, etc. Implement design changes if necessary
In addition to cooling technologies and optical design options, the following topics have been investigated under this project.
- Remote Phosphors for White light (using Blue or Royal Blue LEDs): Phosphors have been investigated for use in a luminaire and for the creation of custom spectra for specific customers and/or specific application spaces. From the results of this work it is clear that remote phosphors are not suitable for the project as the optical losses in the phosphor layer will cause it to heat up beyond the phosphor’s maximum rated temperature. This work is being written up as a separate document.
- Anti-Stokes/Up-Conversion Phosphors: Newly commercialized materials are available that absorb in the IR region and emit light in the visible spectrum. This is a multiple-photon process, in that the phosphor absorbs one photon to generate an excited state, which in turn absorbs another photon to jump to an even higher energy state. From this highly excited state the system undergoes a radiative transition to the ground level, emitting a single photon at a shorter wavelength than the pumping light. After discussions with one of the leading companies in this field it is clear that the process is highly inefficient, and for up-conversion to wavelengths of interest to CRS Electronics, new materials would have to be developed in a lengthy and expensive research program.
- Encapsulation materials and methods: Use of optically clear silicones and epoxies has been evaluated as a potential encapsulation method for CRS Electronics fixtures. Prototypes have been made and evaluated using the GS and a test platform. Encapsulation using either QSil 216 or Dow Sylgard 184 appears to be a viable option for encapsulation, and depending on final fixture design may be attractive for the project. This has been written up as a separate document.