Many people have noticed that some LEDs, including most of the multi-LED moving display signs and most red LED digital readouts, are fed pulsed current instead of steady DC.
There is a common belief that this increases apparent brightness by taking advantage of nonlinearities in human vision. It is widely believed that apparent brightness of a pulsed LED is more determined by peak actual brightness rather than average actual brightness even if the pulse rate is fast enough for the LED to appear steadily lit. There is only a small grain of truth to this belief, and there is a different explanation why LEDs in red 7-segment readouts and many moving display signs work more efficiently when pulsed.
I have done lots of experimentation with 555 oscillators and various LEDs and solar cells in various room lighting conditions. Regardless of frequency and duty cycle, a pulsed LED that gives the same solar cell response as a continuously operated one of the same model will almost always match the visually apparent brightness of the continuously operated one. I have sometimes seen slight gains - less than 10 percent when I get any gain at all - in apparent brightness if the pulse frequency is barely fast enough to make the LED not visibly flicker. This frequency is usually around 60 Hz.
One thing many people don't realize is that most LEDs are not linear as light emitters. I have even seen one book stating falsely that they are linear as light emitters. Although photovoltaic detection is usually reliably linear, photovoltaic emission is a different story. Note the low quantum efficiency for emission and high quantum efficiency for detection (in a favored band) by many LEDs.
Human vision is nonlinear, but that nonlinearity is after a surprisingly accurate time-integration process. When a light is flashing rapidly enough to appear continuously on without flicker, what you see has a good correlation (although nonlinear) with average brightness and is surprisingly independent of peak brightness.
Most LEDs have maximum efficiency at currents near or somewhat over 20 mA. There are some exceptions - gallium phosphide red (low current red or "697 nm" red), old type silicon carbide blue, and most InGaN/GaN blue, blue-green, green, and white LEDs. These specific types are most efficient at lower currents.
So back to why pulsing an LED can increase its efficiency?
Those red 7-segment LED digital displays have lots of LED dice (chips) and
often a low total "current budget" - especially in battery operated devices.
The entire display in a digital watch or digital clock has an average of
about 14-15 segments on at a time, each with at least 3 dice for an
average of maybe 44 or more dice on at a time. The small cheap transformer
in a usual digital clock or clock radio may be good for about 300 mA of
total DC output typically at 5-6 volts or whatever. That leaves maybe an
average of 7 mA or less per LED die, less when it is 12:08 (54 or more
dice), minus current for two more dice in the colon, minus current for
electronic components other than LEDs, etc. The actual "current budget" is
probably 5 mA per LED die or less, maybe under 2 mA per LED die.
But those old fashioned deep red gallium arsenide phosphide LED dice have efficiency varying with current and have efficiency maximized somewhere around 40-100 mA per die. The LED die will make a lot more light if it gets 50 mA pulsed at a 4 percent duty cycle than with steady DC of 2 mA.
The story is similar in moving display signs, although the main reason LEDs in them are pulsed is the way moving display signs of one common technology work:
From Kirk Kohnen:
One aspect to consider though: Those display signs have several hundred LEDs. In order to manage the number of wires, they put the LEDs in two dimensional arrays. The cathodes of the LEDs in a row are tied in common, and the anodes of the LEDs in a column are also tied together (ignoring the dropping resistors). Then, each column is displayed by setting the common anode for that column to a high voltage (say 5 Volts). The LEDs in that column are lit by drawing current through the common row pins to ground.
By strobing the LEDs, you can control N^2 LEDs with 2N signal lines in a NxN array. You also only need N current limiting resistors. It might be useful to point out to folk who notice these signs flickering that they aren't strobed for luminous efficiency, but because it makes the design of the sign that much easier.
(End of portion of e-mail from Kirk Kohnen)
The LEDs in moving display signs are usually of types that have maximum efficiency at 20 mA or more. Since displayed letters are usually moving, the individual LED lamps will be pulsed anyway so it pays to use LEDs whose nonlinearity is of a kind favoring higher efficiency at higher instantaneous current. These LEDs are usually GaP, GaAsP on a GaP substrate, or GaAlP or GaAlAsP on a GaP substrate. With the exception of "low current red", these types have maximum efficiency at currents of 15 mA or more.
UPDATE 6/19/2001 - This applies mainly for older/cheaper technology moving display signs, mostly with one or no more than two lines of text, and no blue nor non-yellowish green nor white LEDs, and usually not bright enough to use outdoors in daylight. There are more advanced ones using a different technology to individually control each LED.
So what do you do if you are building an LED flashlight or bicycle taillight?
Note that most of the modern ultrabright LEDs have close enough to maximum efficiency at currents near or below their maximum rated current. If your average current through each LED is around or over 20 mA, use steady DC. It does not pay to pulse the LEDs when the average current is around or over 20 mA.
Furthermore, the usual ultrabright white LEDs, as well as gallium nitride or indium gallium nitride green, blue-green and blue LEDs have maximum efficiency at lower currents. It works against you to pulse these. Same for "low current red"/GaP red/"697 nm" red LEDs.
Pulsing is in your favor if the LED is a type that has maximum efficiency at higher current and the average current must be low. LEDs that have efficiency maximized at currents 15 mA or more are GaP other than low current red, GaAsP, GaAlAsP ("super bright" and "ultrabright" red) and Hewlett Packard's similar "T.S. AlGaAs", and InGaAlP. Colors of these types are red through yellowish green, excluding gallium phosphide red.
Please read my Copyright and authorship info.
Please read my Disclaimer.