Let's talk LED layout lighting

[peteski]

I'm surprised that the lamp's CRI (Color Rendering Index) has not yet been discussed in this thread.  For explanation see http://en.wikipedia.org/wiki/Color_rendering_index This is very important with any sort of fluorescent and LED lights.  DKS' explanation of the "peaks" and color temperature is IMO a bit awkward but it is not incorrect.   However, besides the color temperature, the CRI is also important for true reproduction of the full range of colors visible to the human eye.

Most white LEDs are actually blue LEDs where the actual light emitting element is placed under a dome of phosphor (or the LED case is molded of epoxy/phosphor mixture.  WHen the LED is emitting light, some of the blue light passes through the phosphor layer while the rest of the blue light is converted to other colors (green , red and colors in between) by the phosphor.  The CRI of white LEDs depends on how ell the phosphor covers the visible spectrum of light.

Fluorescent tubes (including CFLs) work on a similar principle.  The ionized gas within the tube emits UV light and the phosphor coating on the  inside of the glass tube converts the UV light to visible spectrum.  Again, the CRI depends on the type of phosphor used.

A simple visual demonstration can be rigged by covering the light source with a piece of cardboard with a small home in it. A good prism substitute is the surface of a ordinary music CD  or a pressed (not burned) data CD.  Place the light source with the cardboard apperture in a dark room and view the light emitted through the aperture reflected of the surface of the CD. That will show the spectrum of light emitted by the light source.

Most current white LEDs have fairly complete light output over the visible light spectrum. Most fluorescent lights notoriously have very distinctive bands of emitted color, with dark bands in between them.   Those usually have a poor CRI.  As a result, when using them for illumination, they can change apparent colors of objects.

The best way to make sure to get lights which will show true colors is to find out their CRI values. I prefer using lamps with CRI value of at least 85.

[robert3985]

This is a very interesting thread.  I decided to Google about to see what the dif was between actual sunlight, incandescents, LEDs and CFLs as far as their Color Rendering Index (CRI), Spectral Power Distribution curves, and "perception of quality of light" by humans.  I discovered this article by Popular Mechanics that showed graphically the difference in the way different light technologies render their colors...or perceived colors.  Go here: http://www.popularmechanics.com/technology/gadgets/tests/incandescent-vs-compact-fluorescent-vs-led-ultimate-light-bulb-test-9#slide-1

When it's all said and done, I'm not sure what any of it proves other than go with what you like!  Each light source has its advantages and disadvantages, but the clear loser is the incandescent, especially considering the governmental obsolescence being imposed upon them, the heat they generate and their obvious inefficiency.  However, their CRI and SPD curves are smoothest, whereas both LED's and especially fluorescents have "spikes" and "troughs" which means the color temperatures of both of these technologies are "correlated" or "averaged" temperatures, and because of the sharp spikes and large troughs, their CRI's are correspondingly lower than incandescents or true sunlight.

Which means, for me, that the very best technology for taking photos would be incandescents...but...I am not going back to them because of their heat, short life, obsolescence and power consumption.

Second best would be LED's, but the price per lumen is too steep.  Interestingly, even though their SPD curves are smoother than CFL's, the perception of humans on them was more mixed.  I'm sure it's the "delivery" system, or the optics needed to get them to be more similar to incandescents.  I am sure as the technology matures, that will be worked out and the price will come down.

As far as Peteski's comments about CRI...I am not sure how a person would find out what the CRI for a particular light source is, since the rendering of the color of light between manufacturers is notoriously inconsistent.  That means each manufacturer would have to post the CRI of a particular LED "bulb" or CFL somewhere (website?) and allow the consumer to pick which light source they prefer...cheap and low CFL, or more expensive and better CRI. 

Since you (Peteski) buy lamps that have a CRI of 85 or above, how/where do you find it?  Or, do you test each batch of bulbs you purchase?

Yep DKS, the total visible spectrum IS present on all lamps, but...CFL's do have big troughs, no matter what their color temp, and I am now certain that my contention that 5,000K CFL's have more of it is incorrect. 

However, I am not certain what effect a low CRI would have on photography, since digital photography is becoming extremely competent...especially in the more professional-leaning DSLR's.  Guess I'll just keep doing what I'm doing, since that seems to work pretty well for me and my Nikons!

[GaryHinshaw]

I completely agree that the best advice about choosing the spectrum of your lighting is to try it in as-close-to-layout conditions as you can and see if you like it.  You should especially gather locos and cars of different colors and see how they look to you.  When I first tried my daylight bulbs, I was quite surprised how vivid all the blues were, for example.  Effective temperature and CRI are just two numbers that attempt to characterize what is typically a very "rich" spectrum.  By the way, CRI values are often listed on the box for fluorescent tubes; I imagine the same is true for CFL's and LED's, but I haven't researched that.

For those of you interested, I thought I'd mention a little of the background physics.  My apologies if it's pompous, or if graphs give you a headache.  Please skip it in that case.  The first thing to ask is: what is the spectrum of sunlight?  This Wikipedia plot of brightness vs. wavelength summarizes the that information pretty well:



The smooth curve labelled '5250 C Blackbody Spectrum' is the spectrum that would be emitted by a perfectly absorbing ("black") body at a physical temperature of 5250 Centigrade.  The shape of this spectrum is governed by quantum statistical mechanics and was first predicted by Max Planck near the turn of the 20th century.  In fact, Planck's explanation of this spectrum invoked the idea that light consisted of quantized particles ("photons") rather than continuous waves, and was one of the first lines of reasoning that led to the quantum revolution in the following decades.  It was a remarkable intellectual triumph!

The surface of the Sun matches the blackbody conditions pretty well.  (Of course the Sun is not "black", the term means that any light the reaches the Sun would be absorbed, or "thermalized", by the hot gas of its atmosphere, and eventually re-emitted with this spectrum.)  The actual spectrum of sunlight measured in space is shown by the yellow curve.  While there are a number of fine absorption lines in the solar spectrum, the overwhelming character of it is it smoothness and its peak at a wavelength of ~550 nm (greenish-yellow light).   Once sunlight reaches the atmosphere, a good deal of it is absorbed or scattered by atmospheric gases.  A typical spectrum measured on the ground looks like the red curve above, but this will depend on time of day, particulate content, and elevation.  In the visible range (~390-750 nm), the spectrum mostly just looks like a dimmer version of the primary spectrum.

The effective temperature of an artificial light is an attempt to identify which blackbody curve is closest to the actual spectrum produced by the lamp.  To see what this means, it helps to see how the blackbody spectra change as you change the temperature of the emitter:



The peak of the spectrum shifts to longer wavelengths as the temperature decreases.  This is a direct result of the fact that photon energy is proportional to its frequency, nu, specifically, E= h*nu, where h is (appropriately enough) Planck's constant.  As the temperature drops, there is not enough energy in the system to populate the high energy (high frequency) modes, so the high frequency side of the spectrum rolls off more and more, which means less blue, more red.  If you have a lamp with a measured spectrum, you can mathematically 'fit' a blackbody spectrum to it and find which one it most resembles; that is the effective temperature (or 'correlated color temperature') of the lamp.

As noted earlier, a fluorescent lamp typically has a feature-rich spectrum, something like this:



Describing this spectrum by a blackbody curve (and temperature) is a crude approximation at best, but it is well-defined, and it does give a surprisingly good indication of what the light looks like to a human.  (Part of the reason for this is that the human eye responds logarithmically, so that we can cope with a large dynamic range of light.  This has the side effect that features in a spectrum tend to be de-emphasized by the eye.)   Roughly speaking, the color render index is a measure of how much structure is left in the spectrum after the best-fit blackbody portion is subtracted from it.  It is defined as 100 if the spectrum is already a blackbody, and the values drop from 100 as more and more structure remains in the subtracted spectrum.  The index tells you very little about the nature of the structure (e.g., which wavelengths are enhanced or diminished), and the actual recipe for measuring CRI is rather technical.

In a digital camera, the CCD sensor is typically divided into separate red, green, and blue pixels, in a Bayer pattern:



The R, G, and B filters are broad and centered on their respective colors.  The only thing that matters to a camera is how each of the three filters responds to the spectrum presented to it. The goal of color correction is to adjust the relative RGB response to what it would be if the camera were presented a 5000 K blackbody spectrum.  As long as each R, G, and B sensor detects some light, color correction is possible, though it can get tricky in practice.

Sorry to ramble... I love physics.

-gfh

[Philip H]

Gary,
Who let the rocket scientist in?   

[Dave Schneider]

Outstanding Gary! This has to be a Railwire first: The on topic presentation of the Planck curve and Wien's displacement law. I appreciate the figures being presented as wavelength instead of frequency as you physicists are apt to do! I haven't been so happy with a post since Dave Vollmer posted a geologic map in his building thread.  I wish I still had access to a visible wavelength spectrometer, but I now work in the thermal infrared and microwave region of the spectrum. The fluorescent light spectrum is a very useful figure. Interesting how little blue (450-495 nm) there is in this spectrum compared to green (495-570 nm) and red (620-750 nm). Thanks for digging that up.

This has been a very interesting discussion and I have learned quite a bit. In the end, like all things it comes down to tradeoffs. Incandescent lights are out for me due to heat and power requirements. As Robert points out, there are times when they have their place, but not for lighting the entire railroad. I have decided to stay away from LED light strips at present, but may pick up some shorter lengths for testing. The LED fixtures mentioned by DKS look great, but I am still hesitant to do that way at present (worried about valance construction). My plan at present is to experiment with T-8  fluorescent tubes, augmented by some CFLs were needed. The UV warning is really appreciated. I really appreciate all of the input and encourage others to weigh in.

Best wishes, Dave

[MVW]

You guys are a bunch of nerds. 

As opposed to those of us who just play with toy trains. 

Jim

[peteski]

Gary is a nerd who plays with trains!    Seriously though, Gary, that was an excellent explanation!

Bob, I read your other post in the Crew Lounge and I'm surprised that you (as a serious photographer) were no aware of the CRI specifications for lamps.  I'm not trying to put you down, I'm just surprised, period.  I'm an amateur photographer and I have been aware of CRI lamp ratings for several years.

As gary eloquently explained, Digital cameras use sensor elements which use RGB filters over the light detectors.  As he also explained, the high CRI rating od a lamp doesn't guarantee a good performance with a digital camera. If the high-CRI rated lamp has color peaks in between the colors used in the camera's sensor, this will results with poor color rendition.  My logic dictates to me that the best illumination for a digital camera would be a white lamp in which the emissions peak at red, green and blue (at the same wavelengths as the camera sensor's filters).

Such pairing of light emitter/detectors exists in flatbed scanners which use cold-cathode fluorescent lamps and light sensors with red, green and blue filters. Those lamps are designed with specially blended  tri-phosphor coating.  They appear white to the human eye, but when the light is viewed split through a prism (or reflected off a CD) they clearly show that they are emitting red, green and blue light and nothing in between those colors.

As far as finding CRI rating of lamps, more and more often manufacturers include this raring either on the packaging or in online technical specs.

The Wikipedia article about CRI to which I linked in my last post shows a photo of such package.


Also, PHILIPS fluorescent tubes I buy (at Home Depot) have CRI specified on them.  For example one of the bulbs I use is "Soft White", COlor temperature 3000K, CRI=85.

Here is a chart of fluorescent lamp specs, including CRI (although it is very dated)!
http://www.thekrib.com/Lights/fluorescent-table.html

[GaryHinshaw]

Gary is a nerd who plays with trains!    Seriously though, Gary, that was an excellent explanation!

Guilty as charged!  Thanks for the kind words though.  Having spent some time with Dave in Anchorage this summer, I knew he did remote sensing, and your post on CRI made we want to fill in a bit of the back story.  So I thought I'd have a little fun with it, and thought Dave would get a kick out of it.

[Dave Schneider]

Thanks Gary. I really did enjoyed your post. The plot of the solar spectrum also explains why we see light in the 450-750 nm region...because that is where it peaks.

Best wishes, Dave

[robert3985]

WOW!  Love this thread!  Although light behaves as both particle and wave (until observed), a fact that Schroedinger admitted, but made fun of, modern technology uses this fact and many more in the quantum mechanical toolbox without explaining the "why" of it all...an obvious omission that engineers and practical physicists hardly worry about when designing machines to use these principles for our enjoyment and benefit.

Gary...no way would I consider your explanations "pompous".  The graphs and charts make what your explaining much clearer.

Peteski...I'm not offended whatsoever by your curiosity as to why I've never heard of CRI before.  Actually, I'm a little curious as to why I haven't heard of it too.  It has not been a serious omission in my photographic efforts, as I am pretty happy with my results.  When I think about it, it's probably because I use professional lighting equipment, and available light...both of which I trust to give me consistent results.
It's only recently that I've been using my layout lighting as my primary lighting source for my model train photos (like in the last two years).

In all my university photographic classes and on-the-job apprenticeships over the years, not once have I heard anything about a lamp's CRI as something I needed to worry about as far as photographic quality is concerned.  When I wanted a lamp to replicate daylight,  I'd just buy them at my local photography store...250W incandescents of course...and they'd last about six hours.  I had half a dozen of 'em for my copy table before I started using strobes. 

Of course, nobody used fluorescents in those days (we all avoided them if at all possible because they were impossible to filter for), and nobody had heard of LED's powerful enough to be used for photography.

I guess it's kinda like an artist not knowing the composition of his pencil "lead" when he's drawing a portrait...just how it feels in his hand and looks on the board.

Guess my age is showing...

[tom mann]

Some other thoughts on this:

About a year ago, I was remodeling some rooms in my house and wanted to use recessed lighting with LED bulbs to avoid heat issues.  That is the neat thing about LEDs:  they are cool to the touch even after hours of operation.

I used EcoSmart and another brand that I can't recall (Commercial Electric?).  I ended up keeping the EcoSmarts, despite a very cool color (that is a little too blue for my tastes) and a .5-1 sec. turn on delay (annoying since I have 4 in one room).  The other brand needed to be grounded (which I thought was odd) and 1/2 of them worked out of the packaging.  I returned them for Philips. 

The Philips LED bulbs are perfect - great color that is consistent with the label on the packaging.  The two that I have turn on at the same time.  However, they cost a little more.

If I was doing layout lighting, I would select a flexible track lighting fixture and populate it with these $29 bulbs (HD also sells a 4 pack for $107).  I bet the four pack would be sufficient to light 12' of a layout.

Just my thoughts.

[PGE_Modeller]

About three years ago, I started investigating LED lighting for the layout.  The thing that caught my attention was the flexible LED strips with 3 LEDs per 2" segment that could be cut into segments that were a multiple of 2" that would allow the strips to be bent in the "horizontal" plane rather than the "vertical" one.  These operate from a 12VDC source and the equivalent of a Pulse Width Modulated transistor throttle allows complete control for dimming.  The strips I experimented with have a colour temperature of 3500 K and were augmented by 12 VDC LED floodlights with a colour temperature of 4100 K.



The photograph shows the test area on my under construction layout.  In the foreground-background direction, the strip LEDs have a 120 degree illumination angle so the strips are mounted at a 45 degree angle behind a valance that mimics the shape of the front fascia (when it is installed).  A contoured horizontal baffle will restrict the light to the actual scenicked surface of the layout with little or no spill-over into the aisles.  As noted on the photograph, the illuminance of this combination at the centre of the scene, as measured by a Gossen Luna 6 Pro light meter in incident (rather than reflected) mode, is approximately 550 - 600 lux.  A brightly lit office is typically 1000 lux.  To my eye, the colours in the photograph are a good match to what I see on the layout and are also close to what I see at the workbench.  If I wanted to increase the brightness and broaden the spectral range of the lighting, I would consider adding a second strip of LEDs with a colour temperature of 2700 K.  However, with the room lights turned off, I am happy with the illuminance of the current set-up.

(Edit:  The only illumination for the photograph was from the LEDs.  The room lights were turned off.)

At the time I started the experiment, the LED strips were costing $15/foot!  The same strips are now down below the $1/foot range.  The LED floods use an MR16 socket and run in the $9 range each.  These are Canadian prices.  My plan is to use one of the floods every 2 feet.  Based on the experiment, I expect to illuminate both decks of a 15' X 17' L-shaped layout with a total power consumption of about 200 W.  In terms of room heating, this is roughly equivalent to having two extra people in the room during an operating session.

Cheers,

[avel]

Interesting read about LED life  http://hackaday.com/2012/10/23/graphing-the-efficiencies-of-led-light-strips/

Also I wouldn't think about LED lighting with those "strips",  use these as a minimum
 http://www.luxeonstar.com/Neutral-White-CoolBase-Rebel-Stars-s/347.htm

These get pretty expensive for units with all the electronics and hardware to run them.
https://www.sparkfun.com/products/8202

http://www.ledwholesalers.com/store/index.php?act=viewProd&productId=746

Anyone try those 68watt fluorescent / 300watt incandescent equivalent bulbs? I have one, its huge but I've only seen them at 2700K. I got one at HD for around $17.

Also off topic, I use 23watt/100watt equv. around the house and try to always dremel out holes/slots in the base to extend the lifespan. Only because I read somewhere they expire early from circuitry heat related failure (and from short powered on state cycles). Most of the CFL's aren't mounted with the base down anyway, so they get hot.
And no I haven't done any testing to back this up with solid data, I've planned to but forget when changing bulbs.

[NARmike]

There is a thread on the MRH forum going right now (I feel like I just admitted cheating on my wife  ) that is quite good at addressing the LED subject... specifically the strips found on eBay from Hong Kong.

http://model-railroad-hobbyist.com/node/9736