XM-L at low drive currents

[Tobias Bossert]

LEDs normally show a efficiency peak at relatively low drive currents. For older types of LEDs (up to Cree XP-G) I can find measured efficiencies in this forum, but not for XM-L. Can someone point me to the corresponding thread where I can find efficiency of XM-L in the range 10 to 350 mA? The diagrams in the original datasheet doesn't allow to read exact values at low currents.
Thanks

[idleprocess]

Cree is not characterizing the performance of their power LED's below 100 or 150mA.

If you want to see Cree's more granular test data, it's available at http://pct.cree.com/.

Sample data for T4:
If = current
Rel = relative output (1=100%)
Lm = Lumens
Lm/W = Lumens per Watt
VF = Forward Voltage
W = Watts

If	Rel	Lm	Lm/W	VF	W
0.15	0.22	53.17	132.26	2.68	0.40
0.16	0.24	56.77	132.02	2.69	0.43
0.17	0.25	60.37	132.10	2.69	0.46
0.18	0.27	63.97	131.90	2.70	0.49
0.19	0.28	67.48	131.53	2.70	0.51
0.20	0.30	71.08	131.38	2.70	0.54
0.21	0.31	74.58	131.08	2.71	0.57
0.22	0.33	78.18	130.96	2.71	0.60
0.23	0.34	81.69	130.71	2.72	0.63
0.24	0.36	85.29	130.62	2.72	0.65
0.25	0.37	88.80	130.40	2.73	0.68
0.26	0.38	92.31	130.01	2.73	0.71
0.27	0.40	95.82	129.83	2.73	0.74
0.28	0.41	99.32	129.66	2.74	0.77
0.29	0.43	102.83	129.35	2.74	0.80
0.30	0.44	106.34	129.05	2.75	0.82
0.35	0.52	123.69	127.78	2.77	0.97

Assumptions are based on a generous 25C die temperature; real-world conditions will derate that by a bit.

Using a 2-order polynomial (thanks, Excel!) trendline over the entire chart, I get the following equation for relative output:

y = -0.1531x^2 + 1.5455x - 0.0069
x = If
y = Relative output

... which lines up with the output well enough when graphed.

So for a T4, we guesstimate the following:

If	Rel	Lm
0.01	0.01	2.05
0.02	0.02	5.75
0.03	0.04	9.44
0.04	0.05	13.12
0.05	0.07	16.80
0.06	0.09	20.47
0.07	0.10	24.13
0.08	0.12	27.78
0.09	0.13	31.43
0.1	0.15	35.07
0.11	0.16	38.70
0.12	0.18	42.33
0.13	0.19	45.94
0.14	0.21	49.55
0.15	0.22	53.16
0.16	0.24	56.75
0.17	0.25	60.34
0.18	0.27	63.92
0.19	0.28	67.49
0.2	0.30	71.06
0.21	0.31	74.62
0.22	0.33	78.17
0.23	0.34	81.71
0.24	0.36	85.25
0.25	0.37	88.78
0.26	0.38	92.30
0.27	0.40	95.81
0.28	0.41	99.32
0.29	0.43	102.82
0.3	0.44	106.31
0.31	0.46	109.80
0.32	0.47	113.28
0.33	0.49	116.75
0.34	0.50	120.21
0.35	0.52	123.66

Vf calculates to be the following if you want to figure out watts and lumens per watt:

y = -0.0725x^2 + 0.4581x + 2.6148
y = Vf
x = If

[Tobias Bossert]

Thank you very much
The polynom interpolates the nodes given by the cree tool very well. This should be true for near extrapolations too. For example to estimate the values down to half the lowest known node. For XM-L this would be down to 75 mA.
But I have no idea how accurate an extrapolation could be, when you go down to 10 mA, that is 15 times below the lowest node known.
In case the polynome is valid down to 10 mA, efficiency would be down to about 87 lm/W for Bin T6. At 20 mA it is 127 lm/W.
Maximum efficiency seems to be achieved just at 150 mA with about 155 lm/W for Bin T6.
In case the extrapolation would be valid on the upper end of the scale too, we could hope that XM-L T6 can put out about 1030 lm at 5A (for short time only). Not sure if that's true too.
I have no sample in hand at the moment. As soon as I have one, I will make a simple scale test to validate the extrapolation down to 10 mA.

[idleprocess]

I wouldn't take the exercise much further without some hands-on testing. Lacking the tools, knowledge, and time to do that sort of thing I just settle for best "back-of-the-napkin" estimates since I'm never building things where peak performance or efficiency is critical.

[MikeAusC]

Test results in http://www.candlepowerforums.com/vb/...testing/page12 show that for recent LEDs that maximum Lumens/watt occurs at around 5% of the maximum rated current.

Below that current, the L/W drops dramatically.

[Tobias Bossert]

Test results in http://www.candlepowerforums.com/vb/...testing/page12 show that for recent LEDs that maximum Lumens/watt occurs at around 5% of the maximum rated current.

Below that current, the L/W drops dramatically.

Hi all,
I recieved an XM-L U2 1A from KD (s:009989) now and tested it against low drive currents.

Yes, MikeAusC, you are right, the maximum lm/W is at about 5% of maximum rated current, for XM-L this is at 150 mA.

No, MikeAusC, you are not right, the maximum is relatively broad and the drop at the low side is not as dramatical as suspected:
Maximum is about 163 lm/W at 150 mA, but down at 5 mA it is still slightly above 100 lm/W.

I have no means to measure real flux (lm). So I measured intensity in a fixed arrangement with a lux meter and scaled the result to match with pct.cree.com at 700 mA.

You can download the results here

For optimizing the overall performace of driver and LED, it would be the best to use analog dimming as long as the average drive current is above the point of maximum lm/W.
But it is better to switch to PWM in case the average drive current falls below this point. In this case, the current during the pulses should be about 100 to 200 mA for XM-L.

[wquiles]

I recently made my own Joule Thief boards and I have been driving my own P60 LED module using an XM-L at around 120mA or so. Plenty bright even at those "low" drive currents
http://www.candlepowerforums.com/vb/...70#post3910570


Will

[Kestrel]

I recieved an XM-L U2 1A from KD (s:009989) now and tested it against low drive currents.

Yes, MikeAusC, you are right, the maximum lm/W is at about 5% of maximum rated current, for XM-L this is at 150 mA.

No, MikeAusC, you are not right, the maximum is relatively broad and the drop at the low side is not as dramatical as suspected:
Maximum is about 163 lm/W at 150 mA, but down at 5 mA it is still slightly above 100 lm/W.
[...]
You can download the results here

For optimizing the overall performace of driver and LED, it would be the best to use analog dimming as long as the average drive current is above the point of maximum lm/W.
But it is better to switch to PWM in case the average drive current falls below this point. In this case, the current during the pulses should be about 100 to 200 mA for XM-L.

This is good info & very helpful, thanks. For the new 'joule thief' P60's, folks will be curious about this data.

[GeoBruin]

Will,

I have one of your V1 Joule Thief boards that Nailbender has mated with an XP-G (R4 4000k 5B3 tint) in a P60 drop in. I'm getting about 200mA at the tail, producing about 27 lumens from a single CR123. Does that math check out?

Also, what is the expected behavior of this light? Should it maintain the same output until the battery is down to ~.8 volts and then cut out?

Thanks!

I recently made my own Joule Thief boards and I have been driving my own P60 LED module using an XM-L at around 120mA or so. Plenty bright even at those "low" drive currents
http://www.candlepowerforums.com/vb/...70#post3910570


Will

[wquiles]

Will,

I have one of your V1 Joule Thief boards that Nailbender has mated with an XP-G (R4 4000k 5B3 tint) in a P60 drop in. I'm getting about 200mA at the tail, producing about 27 lumens from a single CR123. Does that math check out?

Also, what is the expected behavior of this light? Should it maintain the same output until the battery is down to ~.8 volts and then cut out?

Thanks!

That sounds about right. Now, these are not current regulating drivers - these are boost drivers, so as the voltage goes down, the output current goes down, all the way down to about 0.8 volts when they stop working.

Will