DIY LED specialists:Spectrum shift of LEDs controlled by PWM


New member
I am designing a DIY LED system for my 150 (65 x 24 x24) and I emailed Rick (I didn't email him specifically, just sent a query to contact us) at to ask them why they did not sell the Mean WEll ELN 60-48P, but only the Mean Well ELN 60-48D.

His response was
"We actually don't carry PWM drivers due to the spectrum shift they cause in LEDs. We worked too hard in getting the spectrums we carry to have them shift under pwm."

He has a point. This is a true yet not well known fact about LEDs. They change spectrum (wavelength) as a function of both the amperage driving them or the Pulse length driving them. At full current or 100% PWM they act as advertised, but at lower currents or low % PWN they change output frequency.

Here is a link to a short discussion of the subject.

There is also very significant scientific literature supporting these observations.

The question is whether these effects are worth concerning oneself about.

I argue no. The wavelength at 20% PWM is 456 nm peak and 459 nm peak at 100% peak. For a green LED is changes about 10 nM for peak. Usually the PWN leads to decrease in peak frequency as you lower power. By contrast if yo use modulate current (i.d a ELN-40-48D) to dim the LEDs the frequency increases slightly as you lower power. These are relatively minor changes. So I don' think it is something to worry about

Has anyone else thought carefully or done research on the issue?

I think a discussion is warranted since DIY LED is pretty big these days.

Spaced Cowboy

South Bay Reefer
Can't see it being important myself. The emission profile is an area-under-the-curve property, not a delta function, and the peak shift is minor anyway. You were always getting light emitted at 456nm; after the shift you're just getting slightly more at 456nm and slightly less at 459nm. No biggie.

Personally I don't see the justification for the quoted remarks. Dimming by reducing the current also produces a shift in the peak output wavelength, but in the opposite direction, and therefore alters the spectrum too. Why is that shift not important If the PWM one is ?



New member
the eln-60-48p and eln-60-48d provide absolutely identical output current. the only difference is the way the current is controlled. just because one driver uses a pwm dimming input does not mean it utilizes pwm to dim the LED's. which means rick@reefledlights is not very knowledgeable and this topic is a moot point.

unless i am wrong? although everything i saw in the meanwell data sheets suggests that the P version creates a standard constant current circuit, not a pwm.

im following this topic, hopefully someone who knows if i am right can chime in.
Spectral shift is more or less moot in our applications but his comments are also kinda backwards from the industry at large. Analog dimming typically has the reputation for causing significant spectral shift. This is even used as a marketing gimmick by manufacturers who make the chips inside drivers like the meanwells. Applications that are actually spectrally critical (reef lighting isn't even close, in the grand scheme) are probably more likely to use pwm drivers.

Plus, as rwb pointed out, it is important to separate the format of the control signal fom the format of the output powering the LEDs. It is totally possible to have a driver that takes a PWM signal but has completely analog output. The LM3409 driver designed by terahz and discussed in the DIY driver thread does this. I have never measured the output of an ELN with a scope so I cannot say for sure, but IMHO it isn't significant anyways.

Finally, given the numbers quoted above, it is a totally moot point. The shift is completely withing the specified range for the given LED's color bin, so it is within the variation you will have ANYWAYS.
do the corals notice a difference between PWM and a steady light source dimmed with a pot?

Again, it is extremely important to delineate between dimming signal format vs output signal format. There are definitely drivers that have different formats. It is totally possible to have a driver "dimmed by a pot" that has PWM'd output.

That said, I have put a fair amount of effort into looking for research that answers your question and haven't found anything meaningful. Anecdotal evidence does not really suggest a solid answer, either. My approach is to build drivers that operate at the output level I ant for "steady state" and the use dimming for effects, like sunrise. I try not to build systems too powerful such that dimming is needed 100% of the time to get a proper light output.