In my opinion, it has to be the lineup of Philips Ultra Efficient LED bulbs. Sure, they don't have the CRI of 95 that the Philips Ultra Definition bulbs have, but the efficiency is much better and absolutly incredible for a CRI of 90, and the lifespan is rated at 50,000 hours. Now, of course, almost all modern LED bulbs do not have great quality, but Philips is at least a lot better, in my experience, than most other brands.

TL;DR: What's your opinion on the best LED lightbulb?

None of them, except maybe the very earliest ones that were of reasonable quality.

In what parameter ?

Light quality (for anything that resembles white light) - None of them

Longevity and reliability - Something DIY with big derating for all LEDs and driving components. (It does not have to be complex, even "capacitive dropper" can be made to last for decades with correctly chosen components)

If not DIY, for ordinary screw base lamps, the runner up's for reliability would be probably something that looks like the oldest LED lamps with the big exposed heatsinks, or low power (<=2W) filament lamps, or generally anything that looks abnormally large for its power rating regardless of shape

But the driver failures (for non DIY lamps) are harder to predict and can happen anytime in any lamp

Both Philips and Ledvance's >90 CRI lines have great light quality. Osram/Ledvance also offer 2400 or 2500k extra warm white, >90 CRI lamps which make dimly lit rooms extra cozy.
All are quite efficient, but practically all leds are these days anyway - i don't care whether a lamp uses 5w or 10w because either is negligible on my electricity bill, compared to the cost of running a washing machine.

Avoid those 210lm/w 80cri filament lamps (aka dubai lamps). Those have a very bad color shift to greenish yellow.

Can't say much about longevity. That's lab test stuff, you'd need to test hundreds if not thousands of lamps to get a realistic impression of failure rates. So far, i have only had 1 or 2 LED lamps die in about 10 years of using various ones.

CRI is only one parameter, even not the most important one

The most important thing is plain suitability of the light for the purpose (which may or may not require CRI), and whether the light achieves the purpose without too much adverse effects - Glare, suppression of night vision, eye damage, overheating of the surroundings by IR, either lack or excess of contrast vs. some other light source, effects on wildlife - to name a few



I have not seen the Dubai lamps in person, but besides the spectrum (i guess not different from any other common 2700K LEDs), one thing i would really question is whether they do flicker. There are many types of LED drivers (including ones used in screw base lamps). Some of them - and especially filament ones where there is little space for a driver - have no capacitor for energy storage, and have quite bad flicker

From low power filament lamps made by Harmony (China), sold here under the brand "Eurolux", i noticed that the "non dimmable" ones have relatively low flicker, while the "dimmable" ones are bad. This again is explained by the lack of capacitor, in this case due to a capacitor being incompatible with how a standard dimmer works, so no being used in dimmable lamps

The greenish color is actually the one thing that would not bother me at all



Lab tests for longevity have some issues :

 - Lab does not simulate all conditions. Sometimes some condition is not tested at all, sometimes it may be tested to the standard requirements but then they may still differ from real use conditions

 - Lab may use accelerated aging test, but the only significant controllable accelerated test variables are temperature and humidity, and for a "dry location" lamp humidity is probably not even tested. The lamp may have a failure mechanism which is not affected by the controlled variable, so only show up in a real full life length test (that's 3 years for a 25K hour rated lamp). But then, by the time there is a result for the test, the lamp will be obsolete

If you actually know the design of the driver used in the lamp, the ratings (so the design margins) of the components used, and the quality of the components, you may have a rough estimate, which might point you to things that would never show in a lab test

The philips lamp i'm referring to, does not flicker. It appears to have some generous amount of smoothing since it takes a visible amount of time to fade to black when turned off.

The spectrum is radically different from other lamps, the color is very visibly shifted towards greenish yellow, away from the black body locus. This can also be verified with the color coordinates from the data sheet.

Domestic use, I presume.

I value 4000K and 3000K versions with high CRI and high R9 index, do not mind reduced efficiency.

Interestingly, some not very much loved corn cob lamps have very long life and good reliability due to excellent cooling achieved! Poor CRI usually, though.

Any LEDs except the heavily filtered ones (~2200K and below) are inadequate for domestic use just because of disproportionate peak of Blue light in the spectrum exactly at 450nm



The Dubai lamps long afterglow indeed may point towards presence of capacitor, which is then a significant advantage

The greenish color may point towards a presence of an additional phosphor in the mix, which if anything, would improve the light quality, not reduce it. Though they still would contain lots of Blue (no other way to achieve the high Lm/W with LEDs)



The reason the cobs last is because in many of them the driver is just a capacitor in series with the line. Switching semiconductors not present, a whole lot of failure modes not possible...

Just make sure that the capacitor have adequate voltage rating, that there is SOME ballasting resistance in the circuit too, and that the circuit is not exposed to line transients (which throhg the capacitor are translated into current surges through the LEDs) too often

https://www.lighting-gallery.net/gallery/thumbnails.php?album=8799

These ultra efficient 210lm/w 80cri things have an extra phosphor in the red area, not in the green. The overall color impression of the 3000k one is vaguely reminiscent of a greening out coated HPMV lamp, but yellowgreen shifted instead of green. The color is less neutral white than from a 530 halophosphate lamp, although the color rendering is better!

Annoyingly i cannot find data about the color coordinates. I have dug into those (and even plotted them) for other LED lamps a while back on this forum.

The blue peak of high CRI lamps are there but not really noticeable by eye imho.

There is a visible peak in the green here too, so apparently there is something in there. And it makes sense : Our eyes are very sensitive to green, so adding output in the green is an easy way to boost Lm/W

The more the blue is overpowered by emissions elsewhere in the spectrum (including in blue, but not in the narrow band centered on 450nm), it is less significant

I just can not get the idea behind that blue peak hysteria except some *slight* detrimental effect on CRI, honestly.

And, already for many years >99% of filament and even corn-cob LED lamps work with some kind of electronic driver and a small electrolytic. Probably using a tiny inductor with the off-line driver IC is cheaper than a large film capacitor.

Only colored filament lamps I have seems to retain resistive or capacitive ballasting, may be to withstand running on a flasher.

I think the peak sensitivity of the human eyes is at the lime green. Otherwise, how green thallium iodide lamps are less efficient than HPS and LPS lamps?

Yes, dor the peak sensitivity of the human eye under photopic vision is at 555.5 nm. I think the reason why thallium iodide green metal halide lamps are less efficient than LPS and HPS is because thallium has very strong UV lines, just like mercury vapor, which in a lamp without a phosphor, is completely wasted.

I just can not get the idea behind that blue peak hysteria except some *slight* detrimental effect on CRI, honestly.

And, already for many years >99% of filament and even corn-cob LED lamps work with some kind of electronic driver and a small electrolytic. Probably using a tiny inductor with the off-line driver IC is cheaper than a large film capacitor.

Only colored filament lamps I have seems to retain resistive or capacitive ballasting, may be to withstand running on a flasher.

Why do you think that the hysteria around the blue peak in the spectrum of white LEDs is false?

It's not that there is no effect possible at all. But many things are way overblown, and the peak is also pretty damn small with modern lamps. Any judgment should be made with the state of the art lamps.

Furthermore, many types of lamp have significant bumps or lines in the blue spectrum. Cool white halophosphate have a significant 'shoulder' in the blue region with on top of that the 395 and 405nm spectum lines.

These ultra efficient 210lm/w 80cri things have an extra phosphor in the red area, not in the green. The overall color impression of the 3000k one is vaguely reminiscent of a greening out coated HPMV lamp, but yellowgreen shifted instead of green. The color is less neutral white than from a 530 halophosphate lamp, although the color rendering is better!

Annoyingly i cannot find data about the color coordinates. I have dug into those (and even plotted them) for other LED lamps a while back on this forum.

Found it!


The blue lines are from one of the 3000k ultra efficient Philips lamps with 8 filaments (12NC 3222 981 51991, i sadly did not photograph the EAN which is easier to search for). It is shifted away from the black body radiator line (the line that indicates where a white is as neutral as possible). The black ones also, but for a 4000k lamp of the same type.

The yellow and pink lines are >90 CRI LED lamps and spot on the black body line. I posted more information on this forum at some point, but haven't searched for it yet.



Here's a spectrum image of an Osram 94cri 2700k LED lamp, a fairly modern one. Yeah, there's blue output from the pump diode but it's not exactly a large amount. Of note is the lack of the valley between the pump diode blue and the phosphor emission that you can see in the philips ultra efficient lamp, and that Osram's lamp does not need the extra peaky red phosphor to have a good red rendering.

Finally and annoyingly - Ultra Efficient is just one of those trade mark thingies. You kinda have to be more specific and write down the 12NC product code, or the EAN. Otherwise there's no way of knowing exactly which ultra efficient lamp we're talking about.  These 210lm/w things stick out like a sore thumb in the DIY store lamp demo rack, but other Philips lamps are also called that!