How Bright is Bright?

ch

I know this one!

See table at right.

You're welcome.

Mark L. Fergerson

I know this one!

See table at right.

You're welcome.

Mark L. Fergerson

This page has some pictures of experiments involving varying the LED current and getting different perceived brightness:

The relationship between LED current and perceived brightness is VERY non-linear. I recently had to implement a PWM LED brightness controller in an FPGA. For my application I only needed 16 brightness steps, but my FPGA needs a timing resolution of at least 256 steps. The first non-zero brightness level (i.e. step 1 in the range 0-15) only needs the LED to be on with a duty cycle of 1 in 256. The last step before fully-on (i.e. brightness 14 in the range 0-15) has a duty cycle of 217 in 256. Values in between are suitably spaced in a non-linear way using a look-up-table. This gives 16 brightness steps that "appear" pretty equal.

In another recent bit of design work, I needed to pass a tiny current through an LED even when "off" to allow detection of LED failure. Even at the lowest current that I could detect with my monitoring circuit (just over

1uA) there was still a visible glow that was deemed to be too much. The final solution was to only apply this tiny test current occasionally for a very short pulse.

Regards, Simon.

Your eyes have a limited dynamic range accompanied by a slower acting offset (probably some changes in dynamic range too)...aka dark adaptation.

The apparent brightness of the led will vary with the recent history of your eyes. A led in a dark room doesn't have to be nearly as bright as one in daylight.

The purpose of the light also matters. If it's intended to be an alarm condition that requires attention, you might want it brighter than a status led.

I don't think there's any substitute for "taking a look at it" in the actual operating environment.

LED light output vs current is yet another issue.

A square law works very well. Could be implemented with trivial circuilt.

Wow. FPGA for LED control. Incredible.

Do not detect current. Detect the voltage drop across.

Wow.

VLV

that sounds like those fire control panels in building lobbies where all the LEDs are always sort of flashing, but not really as the panel idles in a self test mode.

I know, right? I did that on a frickin' Z80, with enough cycles to spare that I put in animations and a tone generator!

(This one displaying just the intensity modulation, but the "8's" can be scrolling text; meanwhile, a one-shot digital timer was used to generate chiptunes.)

The exact profile used in that animation was,

wave .byte 4, 14, 28, 64, 255, 64, 28, 14 ; waviness .byte 4, 14, 28, 64, 255, 64, 28, 14 wave2 .byte 3, 3, 10, 40, 144, 144, 40, 10 ; halfway interpolation .byte 3, 3, 10, 40, 144, 144, 40, 10

So the full sequence was 3, 4, 10, 14, 28, 40, 64, 144, 255, and back down, which as I recall, I approximated from an A*exp(sin(t) + 1) type function. Obviously, the pulse width here is 0-255 out of 255, as in Simon's case.

Tim

--
Deep Friar: a very philosophical monk. 
Website: http://seventransistorlabs.com

Display brightness is usually specified in NITs (cd/m^2). So it depends on the area that is emitting the light. Some useful information:

--
Paul Hovnanian     mailto:Paul@Hovnanian.com 
------------------------------------------------------------------ 
Dyslexics have more fnu.