Must it be red? Must it be the size of the one in your link? How do you know what current it takes to get the display intensity you need? Can you PWM the current to the display? Are you multiplexing the segments?
Blue LED + phosphor will give "white" light well above 100 lm/W.
While a blue LED might give the best power conversion [W/W] the problem is the human visual system, with a very bad sensitivity for blue light, so the phosphor efficiency really makes the difference.
At high illumination levels pulsing doesn't make sense, but for low illumination levels pulsating might make sense. The OP should include a ambient light level detector to adjust the LED current depending on the ambient light level.
I would drive it from an I/O port of a TI MSP430G2553 processor, which of course would allow for pulsing the output to save power.
And the thought occurs to me that it might be possible to flash the selected segments one at a time, instead of altogether, so that the maximum current draw would just be what's need to drive one segment. And, I could get away with using one resistor instead of seven. :-)
If each segment flashed overall at, say, 60 Hz., then the question is how long each segment would have to stay on to convince the eye that nothing is blinking. If 2.38 msec would be enough, then that would work. I bet it would, possibly with an even shorter pulse duration. Anyway, certainly worth experimenting with.
It doesn't- the visual brightness goes down with the average current. The LED voltage goes up so you're actually reducing the efficiency (but it doesn't matter much since the power is usually wasted in a resistor- it just degrades the uniformity of segment brightness somewhat).
It will work, sort, of, but will **not** save power. It also runs into limits (LED peak current in cases where decent brightness is required or probably output drive current in your case) at only a few digits. There is **no free lunch**.
Above the flicker fusion frequency the eye just averages the brightness and the brightness is pretty much proportional to current , so there is no gain if there is no flicker. But have fun experimenting.
Highest efficiency is probably with super-red GaAlAs run on a 3.3V supply voltage for the MSP430. Super-red is orangey-red in color (I forget how many nm wavelength offhand).
These three are only low power if you don't update them. When updated they use *large* amounts of power for as much as a second. While they offer high contrast compared to LCDs, they are very temperature sensitive.
The only one I've seen in public use was a full color model in a supermarket for the checkout. It was on a support post rather than between the cashier an myself so that I didn't realize it was the cash register display. Out of the corner of my eye it looked like a full color poster.
The price displays they use in the local Kohl's are LCD.
It is not more efficient and the total power is not reduced for the same perceived brightness.
This somewhat common myth may have had some basis in fact 40-50 years ago when LEDs were crude and had a kind of threshold current of 1mA or whatevr before they started to give off much light, but it's not true with modern LEDs.
That might be old information. Early LEDs had lots of parallel leakage, as if there were a resistor shunt. At low currents the resistor dominated, the LED didn't turn ON. Modern LEDs are more nearly ideal: brightness is proportional to average current, no advantage to pulses of high current
But isn't there some aspect of human visual perception that gives us an effective free lunch, particularly on very short pulses of light?
In another project that I'm familiar with, he flashes a high-efficiency individual LED for 4 msec once a second. It's 3.6V going through a 1K resistor, and the LED is very much visible, and rather bright, even though it's only on
0.4% of the time. Now you could use a 2K resistor for 8 msec to preserve the same average current draw per second, or 4K for 16 msec, or 8K for 32 msec, but at some point pretty quickly the LED isn't going to be visually detectable at all. So in that case it seems that pulsing provides an advantage.
What I was thinking was using the 7-segment both to display a value seemingly continuously for a while, but also as an indicator blinking light by flashing a horizontal segment individually when no value is being displayed, pretty much as described above. It just seems that pulsing would let me do either or both of those functions at the lowest power consumption.