Is mcu pwm a decent way to control a power led current?

That's under-driven. A BANG and a plume, if driven properly.

Easily.

You can also set a PWM and measure the gross, average LED current. That's more than good enough, and a lot less demanding.

I made very careful measurements of many, many LEDs some years ago. My setup toggled the D.U.T. between PWM and constant current, testing the visual effect.

Generally--but not in every case--d.c. was better.

Cheers, James Arthur

Power leds like constant current and are unhappy about anything else (read "unhappy" as have short life and may even smoke)

You need to use an inductor. A picaxe is too slow to control when to switch of the current directly but could still be used if you measured the average current (a small R with a C across that) and use the picaxe to change switching frequency.

"David Eather"

PWM control is standard practice.

There are many LED driver ICs that offer PWM control of brightness.

It is far more efficient and keeps the colour constant.

.... Phil

Thanks David. Any idea how to size the inductor? A scope possible. The picaxe will do 10 khz, I'm guessing.

jb

Dear Sir, Is there any reason why one would need a mcu, pwm etc., to control a single LED ? There are a lot of simpler, intuitive ways of achieving the same goal without any of the complexities -- what is the added advantage of mcu, pwm ?

Pretend it's a buck regulator. The PicAxe has no timers?

Efficiency, for one.

"Ralph Barone"

The context was the use of PWM current with LEDs.

The OP is an idiot and wants the impossible.

.... Phil

You really want to control the current in a LED. Unless you really know what you are doing, I don't suggest setting up a continuous conduction switch mode regulator using a microcontroller. However you can do a discontinous conduction. I can't say I ever did this for a LED, but I have designed discontinous conduction switchers for situations where simplicity is favored over quality.

For discontinuous conduction, you always assume the inductor has been discharged. That is key. Your current versus time plot will be a triangle. Give the apriori knowledge of the inductor value, you pick a time that gives you a desired peak current. You charge the inductor then dump it into the LED.. Now for a voltage regulator, computing how long it takes to dump the current is simple. For a LED load, I guess the on-voltage of the LED comes into play.

If it were me, and it has been me, just get a chip dedicated to the task, of which there are many. I used some cheap Micrel part to do this.

I haven't made any serious studies - just a bit of simple testing, along with information I have read. The relationship between duty cycle, average current and perceived brightness is going to be complex, and will probably depend on the type of LED, the application, the switching frequency, and how hard you are driving the LED. Certainly in many cases, a PWM cycle will give higher perceived brightness per watt - but I can well believe that in other circumstances the opposite is true.

The OP will just have to do a bit of testing himself with the LEDs he wants, and the brightnesses he wants.

On a decent sized heatsink they last a little while, but their output no longer increases much at all with the current. It used to be a great trick with the early ones to dunk in LN2 to stiffen the crystal lattice and vastly improve the QE would last about a dozen goes before expiring.

Old metal can transistors used to go off with a really good loud pop.

These days the only things that go bang when the magic smelly smoke comes out are capacitors. Very distinctive smell it has too..

We once had a batch of light emitting EPROMs - they didn't work too good after programming either.

Given that a constant current drive is relatively easy to do.

You have to be very careful, what you are comparing, e.g. "white" LED vs. monochrome, are you measuring the threshold of detectability or illuminating effects at say 300-1000 lx, in the middle of field of view or peripheral vision etc.

When our ancestors were common food for many predators, it helped a lot, if you could detect any movement in the peripheral vision, even if the image was not very accurate or couldn't detect the color.

Thus you need to keep the PWM frequency sufficiently high e.g. in indirect illumination.

The theory about advantages of PWM compared to constant illumination seems to be that there must be a minimum of photons in say 1 ms to create a nerve impulse. A constant flow of photons for 1 s might not generate a nerve impulse, while the same amount of photons in a bunch will generate a single nerve impulse.

For photopic viewing, the peak is at 555 nm (yellowish green) at 686 lm/W, while the dark adapted eye scotopic curve peak is at 507 nm (greenish blue) with 1700 lm/W, thus LEDs with different colors might give contradictorily results depending of the degree of dark adaptation.

One should not forget the different pulse frequency rates for different kind of cells. I would not be too surprised, if "PWM is better" might apply for scotopic viewing, while it might not hold for photopic viewing.

I found that the heftier red LEDs can pretty much stand up against a pair of half spent AA batteries which makes it a lot easier for amateur astronomers to convert small torches to pure red light. They ignore the instructions to put a resistor in series even if it is given...

Some LED illuminated street furniture like crossings use too low a frequency and in peripheral vision when you move your eyes the object leaves a stream of stationary copies on your retina. Some LED car brake lights have similar obvious flicker problems too. I can't see why since a higher frequency ~1kHz ought to mean cheaper components.

It is well above the danger zone for flicker epilepsy but slow enough to be obviously flashing when you are moving past it.

Please let me know the simpler ways!

mcu gives you switching of multiple channels to control wavelength by using RGB leds, plus ability to flash, plus cooling control, timing, etc. It's the golden hammer I know.

jb

Hmm, Well I know nothing about LED phosphor degradation. But I would have guessed it's the temperature and not the light intensity that is the important parameter. (maybe some exp(activation energy/kT) type of thing.)

If anyone knows differently I'd like to hear it.

George H.

I suspect it is a bit of both coupled with the total flux interacting with any slight darkening or manufacturing defects that might be present. The surface flux off the latest generation LEDs is broadly comparable with that of the suns photosphere - not good to look at!

Damaged LEDs phosphors are visibly darkened over the LED die.

The tail of most energetic photons will probably do the most damage which is why colour prints always fade red/magenta first and blue last.

I ran an ordinary 5mm red LED on 5V PC powersupply for about an hour before the LED failed open-circuit, probably due to the resin expanding. the current was around 100 mA.

At 3V they'd probably last as well as torch bulbs do.

Unless you are aiming for very large annual productions (millions, at least 100,000s units), I would suggest using some of the available chips, accepting a low voltage, generating up to 40 V capable of driving 10-12 LEDs in series with a dimmer control input.

With three such controller chips and an MCU to handle the user interface and controlling the PWM chips should be good starting point.

After gaining experience with such systems, you might be able to designing a product around a single (higher performance) MCU, if you are really aiming for the million units/year market.

I wonder if this is a deliberate strategy because flickering brake lights at the correct frequency (above the fusion frequency when directly viewed, but below that in peripheral vision) could result in a safer car.

Okay, here is some evidence that I may be right:

They call it "conspicuity flicker".

--sp