LED Apparent Brightness

The LED lamps here in Houston seem to all run just fine. Granted every time we have an electrical storm the entire signal goes out, and we have an electrical storm about every other week, but that's pretty uniform across all of the signals, silicon or filament.

There is only one good reason to pulse a light source to get brighter illumination when multiplexing is not needed... if you are matching to a shuttered imager. I had an application that I was developing, in which we were going to use an IR filter in front of a standard CMOS imager, adn then use a pulsed IR source with a short pulsewidth matched to an equally short shutter time on the camera. By sync'ing these together, the IR illuminator could be the predominate light source, overpowering even daylight, while still maintaining eye safety.

Charlie

Too bad LCDs are so slow. The ideal backlight would be sequentially pulsed R-G-B led's.

John

by

no

25% more power in, 1% less efficient, 24% gain, no more battery power consumed ? Magic batteries, magic leds? I suspect magic mushrooms.

Another only good reason might be if the material has a threshold below it does not light up, but at the same time the above threshold current is too much to be applied all the time. This is the case for some types of GaAs Laser diodes. Details:

See figure 7 to get the idea.

Joop

But, um, is it seriously unrealistic to contemplate a panel of RGB LEDs, say, 640 x 480, all by themselves, being pixels? Like a micro-jumbotron? Frankly, I've been wondering about that ever since I started reading about the various technologies that have been studied for making flat-screen displays. Is it just that it would be so prohibitively expensive?

Thanks, Rich

Hi Rich, Actually, I think heat and power are the main reason. If you have a very bright LED, they tend to generate enough heat to need a little bit of heat sinking. As you try to get that many of them together, you don't have much room for the heat sink. At least the jumbotrons have room at the back for some fins, or even a little plumbing...

Also, adding up all those 20mA LEDs, that's a considerable amount of power!

Charlie

I think that's right. I've got some 16x16 LED panels used in making large TV-like displays, with each of the LED elements having three LED dies inside (RGB, tri-color.) So the panel is really 3x16x16 in terms of numbers of LEDs.

I've got some separated by 5mm, some by 4mm. Can't recall any separated by 3mm, yet. But all are backed by a large, thick, metal panel for heat removal. And these units accept three separate supplies, one for each color, in order to help minimize wasted heat.

It's a lot of power. Perhaps as much as some 80 watts per 16x16 panel, if memory serves. And that's just 256 color pixels. 640x480 is more than a 1000 times as much. 80-100kW would be a LOT OF POWER to dissipate in a small display. You need lots of AREA to get rid of that much heat.

Jon

Also it would be one huge piece of silicon (crystalline?) Or an enormouse amount of interconnected small pieces of silicon.

Um, I suggested using sequential RGB led's as a *backlight* for an LCD... didn't I? Not many led's, lots of lcd pixels.

Instead of having a pixel per color, each lcd pixel would handle all three colors, sequentially, so you get a 3x improvement in pixel density for free. And there would be no color filters, so light usage efficiency goes up at least 3:1. The only bug is that most lcd's are too slow to switch at the required speed, say 150 Hz or so.

There's an article in EDN (?) this month; some projection displays are indeed using fast lcd's illuminated through spinning color wheels. Still wastes light.

John

LEDs are at least as efficient as CRTs and more efficient than a ccfl backlight being pushed through lcd pixel polarizers and color filters. So for a given brightness, an led panel should run cooler than a crt or a color lcd of equal display area.

640x480x3 = 921,600 led's. If you ran all of them at 20 mA simultaneously, that would be about 36 kilowatts, making enough light to blind everybody in the room and toast the guy in the chair.

The reason nobody makes small vga-level displays from led's is cost.

John

--
Really? 

How about some numbers?

A ccfl is maybe 30% efficient, DC-to-light. An lcd pixel with color filter can't theoretically be better than 16% (33% for the color filter, 50% for the polarizer) and in real life is maybe half that at best. So my Dell lcd screen is at best 5% efficient, not counting the electronics. Probably less. And the 5% is when all pixels are full-on, screen white, which few actually are; its dissipation does *not* go down even when pixels are black... the light is just wasted.

I've seen the assumption of 5% panel transmission used for lcd's. Multiply that by 30% for the ccfl, you get 1.5% overall.

Good LEDs are, what, 20% efficient? And when you dim an led pixel, it uses proportionally less power, unlike the lcd. Zero when black.

I don't have numbers on CRTs, except to note that at equal brightness they dissipate a *lot* more heat than an lcd screen.

Really.

See above.

John

I think that with 307,200 of each color, you will get enough light with a fraction of a milliamp apiece, less than 1/10 of a milliamp for InGaN green and blue pixels.

- Don Klipstein ( snipped-for-privacy@misty.com)

The "pixels" in the LED display *are* fairly bright and you are right that not nearly as much power would be required for a readable display as is used in these "outdoor" display systems.

I'd be interested to see a real examination of efficiency comparisons, though.

I've stood in front of these wall-sized displays as they placed NTSC pictures on them and they don't toast you. They *are* bright! But it appears to me that far and away most of the energy is wasted as heat, not emitted as visible light.

Well, it would be expensive. These units I have were, in largish quantities, some $70-90 each. I don't know how much of that was bonding and how much was heat-sinking, so if the light levels were reduced perhaps a fair percent of the cost could be stripped out, but it still would be pricey. They used 6 DACs in each panel, PWM within that for % brightness control, and each little die was carefully placed and bonded behind an individual plastic lens.

I have to believe the expense would remain fairly high for bonding a million dies into anything, regardless of the power issues. Not to mention trying to support controlling apparent color of each pixel at fairly high frame rates.

Jon

Call it 2mW per pixel, roughly. 307k of them would be about 615 watts, with all of them lit. Probably have to add a meaningful percentage for driver overheads, though.

Anyone got a wire-bonder, a fair length of gold wire, and some spare time?

Jon

--
With a white backlight running CW and three planes of filters in front
of it, all you have to do is attenuate what you don\'t want to come
through.  The other way, with three LEDs per pixel, you\'ve got to send
enough current into each of the LEDs comprising the triad to get it to
radiate just exactly the right hue at just exactly the right time.

That means that, in order to push current through each LED in the
triad in order to get the pixel to radiate the proper color, turning
the LEDs in that triad on for short periods of time would be
necessary.  Because of that, the array would radiate.  That\'s not a
good thing, and using nano (or pico) amps to do do the color selecton
through attenution seems to make a lot more sense.

That's doing the math backwards. Assume 3 watts of actual light (that's probably generous) and 10% led efficiency (which is pessimistic) you get 30 watts total input power.

John

Someone built one -- a large one -- the NASDAQ sign in Times Square.

Blue LEDs, digital TV bring daylight-bright signs to masses

Saco Smartvision built the overall sign from 320-mm-sq modular panels, which each contain an array of 16=D716 pixels, totaling 256 pixels. Each

30=D730-mm pixel, in turn, comprises eight surface-mounted LEDs: two blue ones in the middle and two greens and a red on either side (Figure 1). The side-by-side arrangement of 8200 panels yields a 1600-pixel-wide, 2000-pixel-high screen. If you do the math, you'll see that this arrangement yields nearly 2.1 million pixels and 16.8 million LEDs. The sign measures 36.6m (120 ft, or eight stories) tall and 27.4m (90 ft) wide. If you go further with the math, you'll see a discrepancy between the area of the sign and the stated number of panels the vendor used to build it. This situation is not a case of playing with the specifications; it occurs because the sign's surface is not an unbroken solid but instead has 30 cutouts for the building's windows. .=2E. Power consumption of the Nasdaq sign is 60W/ft2 (540W/m2). .=2E.