Query on pulsing LEDs: power vs. apparent brightness

The human retina is pretty close to a perfect integrator, AFAICT.

I've used a variety of pulsing schemes. The eye doesn't care, but the LEDs do, hence LED efficiency dictates the choice of schemes.

Most LEDs are less efficient at high pulse powers, YMMV. White 5mm LEDs lose efficiency badly; IRLEDs like it, ISTR.

That information conflicts with my own extensive testing, however, it might be due to their choice of LEDs.

I ran identical LEDs (various colors) side-by-side, one on a pulse generator, and its twin on equivalent d.c. power. No visible difference, over a very wide range of duty cycles.

Another test had a computer switching between two PWM setups, alternating twice a second or so through the same LED. For equal-power patterns, as long as it didn't flicker, no difference in brightness was detectable.

Hear hear!

-- Cheers, James Arthur

Hi, James.

Thanks for resp> >> I'm working on a personal project which will involve pulsing a

[...]

Ooooo-kay. But what are the integration parameters? I realize that this probably falls under... um, "bio-neurology"? (Yes, I just made that up. I think)... but I haven't come up with the right search keyaords to point me at it.

If the eye doesn't care (or even if it does), any increased efficiency in the current-to-lumens part of the question that I can incorporate into my project will reduce its power requirements, yielding brighter LEDs and/or longer periods of activity. Will be A Good Thing, in other words.

Ah. Thank you. Something I wasn't aware of.

[...]

Just so I don't misunderstand... using the "magic constants" from the article, that would be something like this?

LED1: 10mA steady current vs. LED2: 10mA pulsed at 60Hz/5%duty

Sigh. No "free lunch"?

It did seem that a 95% power savings would have been too good for some LED driver chip manufacturer to overlook it (and the company's competitors would have picked up on it in about three seconds ).

Anyway, thank you for the comments. I may still try playing with pulse rates and duty cycles a bit, but I won't spend too much time on it; if I find anything as dramatic as a 95% savings I'll post it back here.

Frank

--
  "...[T]hough they saw the issues clearly and perceived the remedies,
   the sense of public duty tended to fade when the outlook was
   depressing or the political necessities distasteful."
                         -- Barbara Tuchman / The March of Folly

Photons hitting the retina produce photochemical reactions essentially instantly, if that answers your question.

I don't know what the decay rate is, but infer from flicker perception that 1/60Hz is several time constants.

Pulsed 5mm white LEDs lose to resistive losses--i^2*r. For that reason, equal d.c. power is better than any pulsing scheme.

No. LED1: 1mA d.c., LED2: 20mA pulsed, 60-20KHz, 5% duty. No visible intensity difference.

Exactly. This ground has been plowed.

The only advantage I've gotten is from pulsing certain yellow LEDs. For some reason they're horribly inefficient at low current, and much better at higher. I s'pose they're either leaky, or have some sort of threshold.

--=20 Cheers, James Arthur

Ummm- no way. The claim was the /effect/ of 100 lm/W with a 50 lm/W output and not a 20:1 energy savings.

Frnak McKenney schrieb:

Hello,

if you want to use very little power, you should flash each LED with short pulses with low repetition rate. If you try to get the ilusion of a continuosly lighting LED, you would need much more power.

Bye

In the late 70's during development of Medical Instrumentation and the efforts to save power to the LED; we empirically found that the lowest we could go was to a duty cycle of around 1 in 5, thus the average power was 20% what it would have been. We found the brightness perceived by the user was still a function of the peak power, in this case peak power during the ON cycle. There is variation between people so if you make this for others, you must consider that. But if for yourself, just set up an experiment and adjust down to minimum. But, when you perform the test; do it early morning, be rested, no smoking for 24 hours, and have no caffeine and, of course, use your peripheral vision to check for flicker, not direct view.

Again, the period frequency determined perceived flicker and with less duty cycle than 20% the LED started being perceived as dimmer. Be sure to use an LED that has low capacitance, else you'll start losing power there, too.

That's really interesting. I was the only observer, so maybe my eyes aren't the same as everyone else's.

James

I'm impressed with your test setup of using the SAME LED in two different manners. You should get better than 0.1% match doing that. Ignroingany temperature shifting.

I was not clear on your test using two different PWM, but supplying the same power each cycle.

Did you mean: the computer cycled between one bit of time using steady power, followed by a bit of time using PWM [switching on and off] with the exactly the same average power? and in that test you saw no difference?

More explicitly, example, 50mW for 1 second followed by 100 cycles for

1 second with a 25% duty cycle using a 200mW peak? and repeating this pattern. You saw NO difference between the two techniques? You should have seen a substantial brightness increase.

Hi Robert, Are you saying that if I run an LED at (say) 5mA DC and then pulse it with a 20% duty cycle also at 5mA, that the brightness will appear the same to me?

I may have to try it for myself.

George H.

I must have the same eyes that James does. Pulsing LEDs at 5mA are dimmer than non-pulsed... Brightness appears to increase linearly with duty cycle. (1 kHz frequency)

George H.

The computer test compared the effect of various frequencies and duty cycles, alternating, through one LED. Peak current was fixed. I detected no brightness difference between equal-power patterns. That test compares the effect of average power, but doesn't indicate whether the eye is peak-detecting or not.

I also tested two matched LEDs side-by-side, one driven by a signal generator, the other with a d.c. current, in the several-mA range IIRC. There was no apparent difference between the d.c. LED and the pulsed LED, pulsed at equivalent average power. 5% duty 20mA appeared as bright a) regardless of frequency, and b) as equal to 1mA d.c.

IOW, AFAICT the eye doesn't peak-detect. It integrates the photons it receives electro-chemically, then decays over 20-50(?)mS.

I tried a bunch of other combinations--can't remember them all. I might have notes somewhere.

-- Cheers, James Arthur

f I s

on

, y s y I e

For a quick test, try connecting the LED straight to a pulse generator with a current-limiting resistor.

If eyeballs peak-detect, varying duty-cycle should not affect perceived intensity.

To my eye, perceived intensity varied directly with duty cycle, and was independent of frequency.

--=20 Cheers, James Arthur

er

n e e

of

an

t f

As I

ods

tion

d

ye

n

em,

ncy

is

say

g

s I

be

an

5%

s.

o

re

be

e t s

This was from memory dating back to 77. I remember the conclusions of pulsing the current 'saved' power, but from your experiments not so?!

You got me, Either not done right, something changed, or, ?? Perhaps, we ran the frequency, slow enough to 'trigger' the retention. For example, too fast and the eye integrates as expected with no benefit from pulsing. BUT operate at the verge of flicker and the eye's retention keeps the LED being perceived as ON even though you stopped current through it, thus saving power.

Did you try that?

The eye DOES have

Thanks. That helps, but I'm still trying to see if this is enough to constrain the problem. How do I word this? ... Suppose hitting the human eye with X photons per second over 1.0 msec (totalling 0.001*X photons) produces apparent brightness Y for 10 msec (1/100th sec). If I continue to use current to pump X photons per econd into that same eye for 5 msec (an additional 4*0.001*X photons), aren't I wasting photons (and current)?

That's why I'm curious about the "charge/discharge" characteristics of the eye in such (ridiculous?) detail.

[...]

Ah. Thank you for the clarifcation.

Frank

--
  "Expecting the world to treat you fairly because you are good is
   like expecting the bull not to charge because you are a
   vegetarian."              -- Dennis Wholey

Hello, Uwe.

Thanks for jo>

Ah. My apologies if I was unclear. I'm looking for a way (or ways) of controlling LEDs which produce(s) apparently continuous output (probably on the order of 0.5sec to 3 sec) but accomplishes that with the least drain on a battery supply.

Frank

--
  Indeed, the first step in applying someone's "optimal" formulation
  is deciding if their "optimal" comes within the bounds of your
  "good enough".             -- Tim Wescott, in a comp.dsp post

[...]

Ah. Thank you. That gives me some specifics to work with.

As for your test constraints, though... I'll just have to approximate them. The "no smoking" part isn't a problem, but as for the rest, about the _only_ way I can do _anything_ in the early morning (before, say, 11AM) is with some form of caffeine.

Frank

--
  No man can be judged except against the background of his own
  time.  The standards of yesterday are not the standards of today,
  and the circumstances of daily life were vastly different.  Before
  one attempts to render judgement, one should consider the world in
  which the man existed, and the customs of the time and place.
                -- Louis L'Amour / The Sackett Companion

Frnak McKenney schrieb:

Hello,

you were not unclear, I noticed you want continuous output. But continous output for 3 seconds means more drain on the battery than a series of 12 pulses with a period of 250 ms and a duration of 50 ms. If you really want very low drain, think if it is possible to replace continous output with a series of short pulses without looking continuous.

Bye

Good point. The visibly flashing pattern is much more visible -> see = turn signals, modern emergency vehicle indicator lighting.

?-)

I believe LEDs have improved a great deal since then.

No. I'd expect intensity to track average power though, just the same. I've not detected any peak-detection in retinas, not mine anyhow.

--=20 Cheers, James Arthur