Where can I get 200 lumen/watt LEDs?

[jtr1962]

CRI is great and all, but there is another aspect of led bulbs that not many are aware of. some bulbs do not have properly designed circuit and it "strobes" , you do not notice it with a naked eye, but some feel it, after a while some complain of a headache. you can spot that "strobes" with a camera. i had a bulb like that in my kitchen, whenever it was on, my camera had lines moving up the screen. once i changed it, lines disappeared. i had no headaches from it, but plenty of people complain they do have headaches after replacing bulbs to led ones. very common issue, called high frequency flicker.
if i had to pick a 200lm/w bulb with a HFF over 100lm/w with no flicker, that 200lm/w bulb would go starlight into a garbage can

And there's actually a sort of fix for that. When I rewired the fluorescent fixture in the laundry room to accept LED tubes, I found there was flicker. An easy way to do this is turn on a fan, and see if the blade makes moving patterns when it's spinning down. The flicker didn't bother me, but it's possible a sensitive person might have gotten headaches being under that light.

It turns out the "fix" is really simple. Most LED drivers first rectify the AC and filter it with a capacitor. Some of the cheaper ones don't even both with a cap. Those flicker the most. However, even a lot of the ones with a cap don't use high enough value. Sometimes you have room to put a higher value cap. I did this with the LED tubes in my workroom but these tubes didn't have room for it. What I did was put a full-wave bridge and large filter cap on a breadboard, then I put that in the fixture. The ~165 VDC output went to the tube sockets (polarity doesn't matter because the tubes have a full-wave bridge inside anyway. Problem solved! A lot of LED lighting designed to run off the mains will happily also run off DC if the voltage is high enough. The only caveat is this trick only works for non-dimmable lighting. It might even work for some dimmable LED bulbs, but you'll lose the dimming capability.

 

[divine]

Someone on BLF posted a review of a 1x18650 Convoy S2+ with a Nichia 519A with this chart.



According to some basic math, assuming he tested all of these measurements at 4.2 volts.

Lumens / Watt = Lumens / ( Current x Voltage )
1105 / ( 5.5 x 4.2 ) = 47.8 lpw
512 / ( 1.91 x 4.2 ) = 63.8 lpw
151 / ( 0.53 x 4.2 ) = 67.8 lpw
6 / ( 0.02 x 4.2 ) = 71.4 lpw

I imagine with a little more complicated math, we could be able to figure out an estimated lumens per watt based on just the output and runtime (and battery capacity) numbers.

Maybe LED flashlights aren't designed to have very good lumens per watt.

 

[LEDphile]

Maybe LED flashlights aren't designed to have very good lumens per watt.

"Brightest output from the smallest size and the tightest beam" means driving the LED chips at very high current densities and high temperatures. Neither of those is good for efficiency.

 

[Dave_H]

Ok looks like Philips filaments are different from typical which are two in series at 115vac as I described; capacitive limiting in Philips bulb must allow this with good efficiency.

Sorry, can't stream the video on this at the moment.

To OP: in any case 75v or 100v filaments can't be powered directly from 18v lithium battery. If you dim down a bulb can you count the LEDs per filament? Multiply by 3v (approx.) and that will be the filament voltage which should confirm the 100v stated. The only other filament arrangement I am familiar with is all wired in parallel (3v) used in solar garden lighting etc. These probably does not use 200+ lumen/watt LEDs, otherwise you could wire several in series with some form of current regulation. Are you still entertaining opening bulb and removing filaments for test/use?


Dave

 

[Dave_H]

The LED datasheet interpretation can be a bit complex. For Samsung LM301B CRI=70 CCT=4000-5700K at If=65mA and vf= 2.9v, highest brightness bin has 42 lumens max. This works out to 223 lumens/watt. CRI=90 tops at 36 lumens, which gives 191 lumens/watt max. It could be a bit better with low-vf bins, but taking advantage of this "super-efficacy" restricts CRI, CCT and vf selection.

I can't tell from this datasheet what may be further improvement in efficacy below 65mA test current.


Dave

 

[Dave_H]

In a thread some time back, someone identified Bridgelux DS102 V18 Gen 7 which is COB approx. 1" square running typically at 35vdc, up to 30-40W. From datasheet, 180 lumens/watt typical can be achieved but at CRI=70, CCT=4000K, at 31.3W. Running at lower power could probably reach 200 lumens/watt. Other considerations include spacial pattern, proper driver,`and heatsinking at it's lots of power in small area; and cost/availability TBD.

You could probably find something like this which runs from (say) 12v. In the back of my mind is a 4-die LED which could be wired 4S (12v) or 2S2P (6v), which would make driving easier for 18v; lots of 12/24v "auxiliary" automotive LED products use buck drivers which work from 10-30v at up to 1A, and quite efficient.


Dave

 

[jtr1962]

The LED datasheet interpretation can be a bit complex. For Samsung LM301B CRI=70 CCT=4000-5700K at If=65mA and vf= 2.9v, highest brightness bin has 42 lumens max. This works out to 223 lumens/watt. CRI=90 tops at 36 lumens, which gives 191 lumens/watt max. It could be a bit better with low-vf bins, but taking advantage of this "super-efficacy" restricts CRI, CCT and vf selection.

I can't tell from this datasheet what may be further improvement in efficacy below 65mA test current.

From the data sheet output up to 65 mA appears to be proportional to drive current (i.e. no droop). Basically then the only improvement in efficiency driving at less than 65 mA would come from the lower Vf. For example, if you drive at 20 mA Vf drops from about 2.73V to roughly 2.63V. That gives an efficiency improvement of about 4% relative to 65 mA. So you'll get about 199 lm/W with the CRI 90 version, and 231 lm/W with the CRI 70 version. Bottom line, not enough improvement to really be worthwhile since you'll need over 3 times as many LEDs for the same output.

 

[JustAnOldFashionedLEDGuy]

"A lot of LEDs do"
Then why is everybody having so much trouble coming up with examples? Why are philips 60/100w replacement bulbs the only things that take advantage of this?

Because most people don't have the knowledge / experience to slice and dice a datasheet + relate that to binning information. Downloading the Vero 8 data sheet will show test conditions up to 210 LPW, it even spells it out, no interpretation needed. You could use the Citizen LED simulator to see at low currents you can get 220LPW with big arrays, and low current: https://ce.citizen.co.jp/cms/ce/lighting_led/en/products2/notice_Simulator__ver9_202109.html

You could use the Samsung small LED tool, https://www.samsung.com/led/support/tools/led-component-calculator/ , though it is not updated, but get an idea of how the LM301B behaves, see that you can buy this: https://www.digikey.com/en/products/detail/samsung-semiconductor-inc/SPMWHD32AMV5XAR3SU/16609142 and estimate that at 10mA, you are going to get between 230 - 250 LPW.

 

[jtr1962]

I got two of the 60W equivalents today. I can confirm that they stay cool. After letting one run a few hours it was around body temperature at most. That includes the base with the driver. No flicker, either. My method to test for flicker is to use a fidget spinner. While it's slowing down I'll briefly see different strobing patterns if there's flicker. I can even detect very high frequency PWM, as might be present on those multicolor LED bulbs with remote control. The strobe patterns would be very closely spaced.

On another note, we're getting to the point where we're not going to see much improvement going forward. I estimated these produce about 1.5 watts of waste heat while drawing 4 watts from the AC line. So at best we'll have 60 watt equivalents drawing about 2.5 watts. Since nothing is ever 100% efficient, that won't happen. Maybe we'll eventually see 3.5 watt versions, perhaps even a little lower.

I also bought some of these. These are the so-called "ultra definition" bulbs with a CRI of 95. The efficiency isn't bad, either, at 119 lumens per watt. CRI 95 is always roughly 20% less efficient than CRI 80. If these were CRI 80, they would be around 150 LPW. These bulbs get a little warmer than the 200 LPW ones, but not horribly so. Philips could probably make a non-dimmable 160 LPW CRI 95 bulb if they wanted to. A dimmable one might be 140 to 150 LPW.

 

[JustAnOldFashionedLEDGuy]

Philips could probably make a non-dimmable 160 LPW CRI 95 bulb if they wanted to. A dimmable one might be 140 to 150 LPW.

It takes a fair amount of parasitic power to keep a dimmer happy with an LED bulb unfortunately. Hold current for a triac is about 20-30mA for triac dimmers, and you want most of them to work. Problem is it is 20-30mA whether the voltage is 30V, or 168V on the AC waveform. You can have active circuitry for this, but you need tolerance and time for it to work, which means passive parts with some parasitic consumption. This is a 13.5W bulb making 1600 lumens. A 140lpw dimmable? ... probably. 150lpw, unlikely. At 60W, the parasitic draw necessary will drop the LPW.

 

[Dave_H]

200 LPW and above is a noble goal but it's starting to approach theoretical (and practical) limits for white LEDs, somewhat above 300 LPW. Article I read indicated greater than 400 LPW possible but with narrow limits on CCT, and rather bad CRI due in part to a green tint; might be OK for some non-critical applications but certainly not residential.

A flicker detection method which sometimes works is to move the light rapidly back and forth and observe the light trails due to eye persistence of vision. It works on LED Christmas lights and some other small bulbs.

I tried this with an E12 0.7W single filament bulb, found duty cycle about 2/3. Of course the old "wave your fingers in front of the light" and "how many fingers do I have?" also gives some indication. This filament measured about 97vac on DMM with very low dc content.

Checked local HD and no sign of the new Philips bulb yet.

100 LPW or a bit above combined with good CRI and lifetime is good for me.

Dave

 

[hugodrax]

TCP makes 200 lumen per watt bulbs.
Spec sheet below
They have an 8000 high cri model that does 8000 lumens for 40w and work in enclosures.

https://a89b8e4143ca50438f09-7c1706ba3fabeeda794725d88e4f5e57.ssl.cf2.rackcdn.com/spec_sheets/files/000/083/083/original/tcp-high-lumen-led-filament-bulbs-specs-voct22.pdf?1666973776

 

[Dave_H]

Interesting company, TCP. They also make A19 bulbs with 9W, 730 lumens, 6000h life and unspecified CRI. I tried one and it ran very hot, probably explains the short life expectancy.

TCP filament lights are interesting, efficacy range 128-200 lumens/W, but they are all high power/brightness (A21/16W at 181 lumens/watt gives 2700 lumens). One of the better OTC filament bulbs I see is 6.5W, 800 lumens which is at the low end of this range.

Also wondering about cost/availability, as they target commercial/industrial versus residential.


Dave

 

[Dave_H]

A problem with clear bulbs is in cases where you view them directly i.e. not shaded or overhead, there can be a lot of glare. A frosted version is going to give lower efficacy.


Dave