Single vs dual LED light efficiency

Sane

Newly Enlightened
Joined
Jan 31, 2023
Messages
2
Location
EU
LUMONITE Compass headlight has a single Cree XM-L2, with 3.6V 3500mAh battery (12.6Wh) and produces 600lm @ 3h 35min. Peak luminance 1250lm (measured 1367lm).

Lupine Piko X4 headlight has dual Cree XHP50, with 7.2V 3500mAh battery (25Wh) and produces 650lm (6W) @ 4h. Peak luminance 2100lm.

So, how come the efficiency of Lupine is only 60% of Lumonite ?
 

LRJ88

Enlightened
Joined
May 4, 2014
Messages
647
There's a whole lot of factors involved here, but first and foremost i have to ask; what do you mean by efficiency in this?
 
Joined
Apr 13, 2020
Messages
336
LUMONITE Compass headlight has a single Cree XM-L2, with 3.6V 3500mAh battery (12.6Wh) and produces 600lm @ 3h 35min. Peak luminance 1250lm (measured 1367lm).

Lupine Piko X4 headlight has dual Cree XHP50, with 7.2V 3500mAh battery (25Wh) and produces 650lm (6W) @ 4h. Peak luminance 2100lm.

So, how come the efficiency of Lupine is only 60% of Lumonite ?

Because the Lumonite with a single XM-L2, even G2, is probably not getting more than about 100 lumens/watt at 600 lumens so they are drawing 6W. 3-35h = 21.5W which obviously exceeds the battery capacity. I am going to guess it dims considerably as it discharges.
 

sirpetr

Newly Enlightened
Joined
Sep 7, 2012
Messages
96
Location
Czech Republic, Prague
Hi Sane, easy calculations but long to describe the process. I describe it on Lumonite Compass, XP-L2, 600lm and 12.6Wh battery. There are two things you must estimate - efficiency of the driver and optics losses.

1) First is to find out how many W do you need to produce 600 lumens with that type of LED
When you know LED type (XM-L,XP-L2 etc) and bin (T6, V6, W2...) then you can calculate how much current and voltage you need to have to shine that amount of light (lumens). Current multiplied by voltage give you W value - needed power. Usually manufacturer specifies that voltage and current in 25° or 85° LED die temperature so rather choose 85°C (=less efficient). Also each manufacturer states some range of light output because no two LEDs are same so they fit them into ranges described with MIN/MAX light ouput. Its up to you what number you take, but we always use MIN to be sure that at least that amount of lumens is produced. Taking just MAX value doesnt make sense. In this example XM-L2 with U2 bin in 85°C produce 589lm when supplied with 1500mA and 2.83V (taken from Cree datasheet: https://assets.cree-led.com/a/ds/x/XLamp-XML2.pdf ). Power = 1.5A * 2.83V = 4.245W. Round it to 4.32W to have 600lm.

2) Then you must account for voltage conversion losses
Drivers can be generally 70-99% efficient. Efficiency up to 90% is rare but possible. FIY we make our drivers with 98% efficiency but It took a long time to develop and its not in the whole voltage range and currents. But take 90% as an example, so, to supply 4.32W to LEDs you need approx 4.8W from battery.

3) Then you must account for optics losses
Light from LEDs goes through TIR optics or reflector, both have optic losses. Reflectors could have higher losses. Best and large TIR optics have 90-92% efficiency. So at least another 10% of light is lost there. When you need to have 600lm on output, LED must produce at least 660lm. So you should make previous calculations again for 660lm to be accurate. But lets make it easier and just simply add 10% more power to account for that, so 4.8W * 1.1 = 5.28W.

4) Calculate true (best estimated) runtime
You know battery Wh rating (12.6Wh). With constant regulated light you have runtime = 12.6Wh/5.28W = 2.39h = 2h23min. So its not constant regulated all the time. Its okay as long as the manufacturer shows you runtime graph so you know what to expect. Chinese brands often use runtime graphs with logarithmic or pseudo scale to impress customers that the step down is not so bad - like this runtime graph


You see its not complicated. Previously Cree had pct.cree.com page where you could easily dial all LED parameters but its not there anymore after Cree renamed to Wolfspeed. So its more difficult to go through datasheets to read voltages and currents from there.

Generally two LED lamp should have higher efficiency = runtime. Piko X4 in my knowledge is also using XM-L2 LEDs but higher U4 bin, thats approx 8% brighter (or lets say more efficient). What I personally do not like is their 6500K color temperature.
 
Joined
Apr 13, 2020
Messages
336
Previously Cree had pct.cree.com page where you could easily dial all LED parameters but its not there anymore after Cree renamed to Wolfspeed. So its more difficult to go through datasheets to read voltages and currents from there.

The LED part of the Cree business was sold to SMART Global Holdings in March 2021 and is still known as CreeLED. Wolfspeed is a separate company that was the remaining business of the original Cree Inc. and change to Wolfspeed to rebrand around silicon carbide non-LED devices which the company was based around.

To your points, you my achieve 98%, but likely over a very narrow range where Vf is very close to Vbat. A linear regulator can achieve that too :) .... not a sales thread.

I am going to guess my 100lpw is closer, in practice, that your 113, but either way, still comes down to it can't sustain that level at the run times states and will have to reduce in brightness.
 

kerneldrop

Flashlight Enthusiast
Joined
Apr 24, 2021
Messages
2,333
Location
South
2) Then you must account for voltage conversion losses
Drivers can be generally 70-99% efficient. Efficiency up to 90% is rare but possible. FIY we make our drivers with 98% efficiency but It took a long time to develop and its not in the whole voltage range and currents. But take 90% as an example, so, to supply 4.32W to LEDs you need approx 4.8W from battery.

I would love to learn more about drivers and efficiencies. There's not a lot out there for the layman to just google and read.
So many people say Zebra is super efficient, but I don't think they take into account lumen output when comparing run times.

Convoy just came out with drivers they claim are 90% efficient.

Folks say boost drivers are most efficient. I wonder if that's true.
 

sirpetr

Newly Enlightened
Joined
Sep 7, 2012
Messages
96
Location
Czech Republic, Prague
The LED part of the Cree business was sold to SMART Global Holdings in March 2021 and is still known as CreeLED. Wolfspeed is a separate company that was the remaining business of the original Cree Inc. and change to Wolfspeed to rebrand around silicon carbide non-LED devices which the company was based around.
The thing is when I go to pct.cree.com (Product Characterization Tool) its redirected to wolfspeed.com/pct and its 404! Its a shame such useful tool is no longer available.

To your points, you my achieve 98%, but likely over a very narrow range where Vf is very close to Vbat. A linear regulator can achieve that too :) .... not a sales thread.
Yes, agree, thats what I wrote in my post. Its hard to give one efficiency number for all voltages and currents. But switching converter will almost everytime be better than linear.

Folks say boost drivers are most efficient. I wonder if that's true.
Its not true. Generally buck (synchronous) driver is more efficient because of its topology and current flowing through the coil just one time. In synchronous boost, its more difficult and they have greater losses on coil´s resistance. But when designed well both Buck or Boost can be 92-98% efficient. In our products, some of them are Bucks but more Boosts.

I would love to learn more about drivers and efficiencies. There's not a lot out there for the layman to just google and read.
See e.g. these resources:
https://en.wikipedia.org/wiki/Buck_converter
https://www.instructables.com/Make-a-microcontroller-based-boost-converter/
https://www.st.com/resource/en/appl...r-for-battery-chargers-stmicroelectronics.pdf

Information is there already, just start googling. Every IC manufacturer has application notes with all informations. Lot of forum projects, threads. Its only about you If you really want to dig into that and spend lot of hours with it. I started 13yrs ago without any knowledge, but after 1yr I had first working non(synchronous) buck converter
 

Dave_H

Flashlight Enthusiast
Joined
Nov 3, 2009
Messages
1,375
Location
Ottawa Ont. Canada
Achieving 98% conversion efficiency with any sort of ordinary non-isolated low-voltage buck converter is highly unlikely, even if well-design and everything optimized (which means a narrow set of conditions). There is switching/conduction loss in the transistor(s), inductor loss, and diode loss if design is not synchronous (some still are not).

Ones I've seen or worked with are typically 85-90%+ in the middle of current range, low at low current where impact is usually minimal; and dropping off at higher current.

I recall a report of switching converter using Gallium Nitride in place of silicon, which reached up to 97-98% efficiency, but highly optimized and under specific conditions.

Condition of very high efficiency when Vbatt is just above Vf is possible. I saw a solar charge controller which appeared to use this sort of reasoning, makes for good marketing. But for LEDs, how is that likely to occur with vf around 3v at which Li-ion cell at that voltage is essentially dead. Even if Vf is say 3.4v, cell is end of charge and operation will not persist for very long.

As pointed out, even a linear regulator is not necessarily low. Take vf=3.0v and vbatt=3.8v, which gives 79%.

Dave
 

sirpetr

Newly Enlightened
Joined
Sep 7, 2012
Messages
96
Location
Czech Republic, Prague
Ones I've seen or worked with are typically 85-90%+ in the middle of current range, low at low current where impact is usually minimal; and dropping off at higher current.

As pointed out, even a linear regulator is not necessarily low. Take vf=3.0v and vbatt=3.8v, which gives 79%.

Dave
Yes, I agree, many buck/boost converters are in range 80-90%. But difference between 79% and 90% is high in my opinion. You loose 10% runtime and generate 10% more heat!


Achieving 98% conversion efficiency with any sort of ordinary non-isolated low-voltage buck converter is highly unlikely, even if well-design and everything optimized (which means a narrow set of conditions). There is switching/conduction loss in the transistor(s), inductor loss, and diode loss if design is not synchronous (some still are not).
Maybe for you unlikely but we have them. At least one of them (we have 8 different in 8 headlamp models) has 98% efficiency in some operating range (voltage, current), its synchronous, low frequency one. Mosfets wth low Rsdon and low Qt, fast transition of states. High quality and large coil, best ceramic capacitors. Measured with high quality intruments. Confirmed with runtime calculations/measurements. I design these DC/DC converters for 12yrs, I have to know something about it.

Low currents and high currents have lower efficiencies. The closer the voltages, the higher efficiency.
 
Last edited:
Joined
Apr 13, 2020
Messages
336
Maybe for you unlikely but we have them. At least one of them (we have 8 different in 8 headlamp models) have 98% efficiency in some operating range (voltage, current), its synchronous, low frequency one. Mosfets wth low Rsdon and low Qt, fast transition of states. High quality and large coil, best ceramic capacitors. Measured with high quality intruments. Confirmed with runtime calculations/measurements. I design these DC/DC converters for 12yrs, I have to know something about it.

Low currents and high currents have lower efficiencies. The closer the voltages, the higher efficiency.

Let me guess, when Vf is just under Vbatt and essentially running as a linear regulator :)

If you have a high efficiency architecture that is achieving 98% with true switching, it should stay close to that. Certainly not dropping below low-mid 90's for a buck.
 

KITROBASKIN

Flashlight Enthusiast
Joined
Mar 28, 2013
Messages
5,437
Location
New Mexico, USA
Thanks for the interesting reading!

For an LED with a Vf around 3, a low internal resistance lithium ferrous phosphate cell hums along quite nicely at 3.2V. While charging, should max out less than 3.6. This might make for an efficient matchup with quality components?
 
Last edited:

Latest posts

Top