{ illuma·blogspot·com }

Sunday, January 8, 2017

••◊ DIY LED Light With High CRI

I've spent the last year on a personal electronics project, purely out of an engineer's propensity for exploring.  The project finally came to life during the holiday shutdown a few weeks ago. 

Since 2006 I've been working on films and watched the dawn of (terrible) LED lights.  They started out anemic with extremely poor color rendition.  In a nutshell, it was clearly a developing technology.  Only people who didn't care about color, or didn't know better, used the early products and their cheap-o knock offs from China.  That is...until they saw their footage and realized that there was *no color* in the talent's skin or everything looked extremely green (been there myself).  LEDs quickly developed a bad reputation, much like digital audio developed a bad reputation for it's early incarnations. 

Then along came the automotive industry and government energy policies to transition consumer LED lighting products over to LED.  All of a sudden people expected LEDs to actually look good and the video industry is still riding that wave.  Also, the video industry started to develop remote phosphor panels so the product developers could finally tightly control the color rendering capabilities of their fixture.

My initiative was started when I saw the new violet based "white" LEDs from Yuji.  I specifically liked the spectrum of the VTC5730.  The problem with most (not all) blue based "white" LEDs is that they have a large spike in blue, nothing in violet, a dip in cyan, a large "hump" in yellow-green, and not enough in flesh-friendly red.  It's a compromise to get something that looks decent to our eyes, but cameras don't adjust their perception based on the general color in the room and memory.  They just record what's there.  So blue-based "white" LEDs often had a reputation for making people look green (too much green, not enough red).  That's changes with violet based LEDs.  Now there's very little of any gap in the spectrum and the LEDs output very closely matches real daylight (I used the 5600K LEDs).

Here are a couple pictures of the front and back of the fixture, which you can compare to the CAD pictures above.

Learning CREO Elements took a few weeks and a few re-do's of the CAD when I messed up (big time!).  However, that was the point of doing a project like this.  I have no plans to enter the LED lighting market.  I just wanted to exercise my brain outside the familiarities of my paying job.  Learning the new CAM computers in our machine shop was likewise a growth exercise. Now I feel like I'm pretty confident in creating new parts if I need them. What I also learned is that machining is a pain and I'd rather be designing...so for my next project I plan to buy a case.  It's worth the money in terms of my sanity.

The PCB's were all done through ExpressPCB.  That's why all the boards are 3.8" x 2.5".  They offer a special deal if you use that specific PCB size.  It's also why there are six LED "bulbs" in the design.  The one in the lower left corner has a temperature sensor installed for safety since the LEDs are only rated to 70C.  The firmware debug cable is sticking up outside the top ventilation grill.  I've just been too lazy to remove in the last week.

I originally intended the LED light to be a platform where I could drive multiple types of LED "bulbs".  The drivers can easily handle 120mA to 700mA LEDs (standard driving currents).  If I was to do this project again I would just take all six LED driver boards and combine them into one with 120mA drive only.  That would have made cable routing much easier - it's currently a mess inside and something I need to think about for future projects.

A part of the project that was necessary from a development perspective was the LCD display.  Obviously it's not need for operation, but I wanted to be able to see specific drive levels and the thermal response of the system.  I didn't have a way to simulate the thermal conduction of the light before constructing it, so it was a matter of applying educated engineering practice/guesses and crossing my fingers.  It turns out that the bulb gets up to about 61C when fully cooking.  The LED max out at 70C, so it worked out.

I also included a forced cooling ventilation system at the bottom.  This really helps with keeping the bulb temperature down.  What I learned here is that the vendors don't exactly tell you what to expect in terms of fan noise.  The first fans I bought were TERRIBLE in terms of noise.  When I bought the second batch it made a HUGE difference in terms of audible noise. However, they still aren't silent, despite vendor claims.  What I did as a compromise is only make the fans turn on when the bulb goes past 40C and the speed ramps up all the way toward 60C.  I figured if the light was that close I wouldn't need the power output and it's best to have the fans turned off for audio recording.  One of the main issues here is that I didn't have the option of a custom heat sink extrusion, like Cineo or Arri LED lights.  I had to buy an off the shelf heat sink and machine it to work, which meant compromise.  If I had the option of a custom extrusion I probably would have ditched the fans altogether.

Here's the final working light - turned way, way down.  At 5 feet, ISO 100, 24fps, I measure f4 and a quarter at full power.  That's about 240 foot-candles, which is pretty respectable output - on the order of some expensive pro-level LED fixtures.  I can clearly see a difference in a fully sun lit room at mid day. Some improvements could be made by using actual white solder mask and other optical optimization around the bulb.  I'm sure I'm loosing some light just due to the 1/4" thick plexi-glass in front of the bulbs.  I had to cover the bulb boards in white silk screen because ExpressPCB didn't give me an option of solder mask color - yet another compromise.

Overall, it was a great learning and growth opportunity. I learned to generate my own mechanical CAD in both 2D and 3D.  I learned CAM machining.  I also learned about thermal management and developed a new LED driver topology that I haven't seen elsewhere (mainly for cost reasons).  If I was to expand this project I might add a mount for barn doors and a soft box.  I might also create bi-color LED bulbs boards, which is mainly a matter of soldering together new bulb boards with a mixture of LEDs, a bit of firmware magic, and adding another knob to the UI.

For now, I'm just going to move on to the next project.  As I mentioned, there's no commercialization initiative here.  I just wanted to create something as an engineering art project.  I've learned what I'm going to learn.  Now it's time to move into the wild world of audio.

No comments:

Post a Comment

Only comments that DO NOT include links will be allowed. Any comments that include product advertisements will be deleted. Other than that, thanks for stopping by and I appreciate you taking the time to write a comment.