Chasing Performance in LED Product Designs – Manufacturer Data vs. Realized Output

Posted: April 18, 2018 in General Commentary
Tags: ,
Follow along, and see if this sounds familiar…

In the process of designing a new light, you begin by collecting manufacturer data sheets. You rifle through the LED data to find lumen output for LEDs and select one with a rated lumen output of 390@700mA, 3000K CCT / 90CRI (test current), and 188 lumens per LED at 350mA (calculated value), @ 3.4 Vf, for 158 lumens per watt. Nice!

You calculate what you need to make the target 1000 output lumens, and design a product around the data and calculations, use 7 LEDs operated at 350mA, to include 30% over the target to compensate for optical losses and temperature per the manufacturer data sheets. Using first article parts to build a prototype, you send it off to the photometric lab, expecting to see results very close to what you calculated. You can live with a minimum of 900 lumens, but hope to see better than 1000, as the data provided by the component providers indicates this should be the case.

Your expectations are based on the following:

  • The LED data sheet says the LEDs can produce 188 lumens at 50mA
  • 188 lumens x 7 LEDs = 1316 lumens, providing 30% more than the desired end result
  • The optical data sheet claims an efficiency of 92% (8% loss)
  • The driver data sheet shows 97% efficiency at 350mA / at a voltage range of 18 to 34Vf, with a rating of 12W (covering the string voltage of 23.8 Vf – with a calculated watts load of 9.8W)
  • CCT selected is 3000K by LED data sheet, 90 CRI
  • Your LED case temperature tested at 40°C after 20 minutes, which the manufacturer data sheet shows minimal lumen loss, well within expectations and over-design of 1504 lumens.
  • Calculated and expected results >1000 lumens, 9.8W, 102 lm/W.

The results you are expecting

  • Luminaire total lumens 1100+
  • Watts with driver loss 10 +/-
  • Lm/W = 108+
  • CCT = 3000 / CRI = >90

The lab results come in, and you anxiously open the file and discover the following:

  • Luminaire total lumens 620.6
  • Watts load 8.7
  • Lm/W = 71
  • CCT = 2850 / CRI = 88.5
  • Measured LED case temperature 66°C
  • LED ambient temperature under the optic is 40°C
What happened?

There are numerous variables that come into play that will trim effective luminaire output from expectations built around manufacturer data sheets. Here are a few that this example will have suffered from:

  1. CCT is a generalized term. Actual LED CCT’s vary. Further, optical systems can impart a warming of LED color by as much as 200CCT. Operating LEDs at reduced current will also cause warming of CCT.
  2. CRI is an averaged value, and will vary somewhat by LED bin or production run.
  3. The actual LED lumens from purchased reals of product produce 167.32 lumens at 25°C Tj – a loss of 11% based on bin group purchased.
  4. The LED manufacturer data is based on tests of cherry picked LEDs strictly held to a Tj of 25°C, in an ambient temperature of 25°C, for a duration of 20 milliseconds – The actual LED case temperature of the LEDs was 66°C, with an ambient of 40°C, resulting in a junction temperature of 77°C – Result is a de-rated LED output of -15%.
  5. The LED manufacturer data is based on tests of LEDs that have never been populated onto a board, so have not experienced the thermal cycling processes involved there. One thermal cycle reduces the measured LED output by 7%.
  6. The driver manufacturer data assumes an ambient temperature of 25°C, operation at exactly 120VAC, operating at the maximum Vf. – Actual driver operating conditions are 42°C ambient – resulting in an efficiency loss of 3%
  7. The actual driver output is not 350mA, it is actually 340.4 mA, which is within the +/- 5% tolerance, for a current and LED lumen loss of 3%.
  8. The actual LED string voltage measures 22.4, indicating the LED Vf is 3.2, not 3.4 as shown on the data sheet – for a loss of 6% energy through the LEDs reducing lumen output
  9. The driver actually delivers 21.97Vf when connected to an LED string presenting a voltage drop of 22.4Vf while maintaining the output of 340.4mA, for a loss of LED lumens of 2%.
  10. The driver efficiency rating of 97% is at full load. The load connected in the design is 62%, resulting in a measured actual efficiency of 89%.
  11. The optic used has a measured efficiency of 97%, but that is based on simulated data derived from the manufacturers design software. Actual optical efficiency measures 88%, for a loss of 9%.

Add this all up and you get:

  • Lumen loss from actual LED lumens vs. manufacturer data = 33% (188 x .67 = 125.96)
  • Lumen loss from thermal conditions at the LED from manufacturer data sheet to actual applied conditions = 15% (125.96 x .85 = 107.066)
  • Lumen loss from driver under-current / under-voltage conditions = 9% (107.066 x .91 = 97.43)
  • Lumen loss due to actual optical efficiency = 9% (97.43 x .91 = 88.661)
  • Loss of driver efficiency due to ambient condition = 3% (.97 x .97 = .949)
  • Loss of driver efficiency due to low load condition = 8% (.949 x .92 = .873)

Total actual lumens per LED = 88.661
Total luminaire lumens = 620.63
Actual driver LED load = 7.625
Total driver efficiency at actual load = .873
Actual driver watts at 120VAC = 8.733
System lumens per watt = 71.07 lm/W

 What now?

The next step is to adjust current to the LEDs to push output up to 500mA, and get the following:

Total actual lumens per LED = 120.243
Total luminaire lumens = 841.70
Actual driver LED load = 10.9
Total driver efficiency at actual load = .903
Actual driver watts at 120VAC = 12.07
System lumens per watt = 69.73 lm/W

Now what?

The following is what happened:

  • The higher current increased LED output, but increased heat as well, so lumen increase was less than the increase in current supplied. The result is less than expected output increase and lower efficacy.

So, you push the LED to its maximum test current of 700mA and get:

Total actual lumens per LED = 159.59
Total luminaire lumens = 1117.13
Actual driver LED load = 16.3
Total driver efficiency at actual load = .88
Actual driver watts at 120VAC = 18.53
System lumens per watt = 60.28 lm/W

Now what happened?
  • More current increased heat even more, in both LED points and ambient inside the luminaire
  • The higher current increased the driver LED load beyond the original driver selected, requiring a change from a 12W capacity driver to an 18W capacity, which has a lower efficiency when loaded at 16.3W.
  • The composite of heat and driver selection compounded to reduce efficacy, but increased lumen output to exceed the design target.
The solution?

There are several:

  1. Lower expectations. The original target is obviously pushing the limits of the LED configuration selected. There is no solution to resolving the difference between manufacturer data and actual performance, this is simply the reality of this technology.
  2. Search for higher efficacy LEDs, more efficient optics, and higher efficiency drivers. An improvement of 10% at the driver, 5% in the optics and 10% at the LED will produce the following results:

Total actual lumens per LED = 144.3 (@ 500mA)
Total luminaire lumens = 1010.1
Actual driver LED load = 12.6
Total driver efficiency at actual load = .93
Actual driver watts at 120VAC = 13.55
System lumens per watt = 74.55 lm/W

  1. Further improvements in thermal design may also produce additional gains in efficacy by increasing lumen output without adding any additional power.

This is a common issue with designing LED products. Data sheet values simply do not stack up as they might seem and represent values that are not representative of application conditions. This applies to LEDs and drivers alike and will vary by product lot as well.

Then what happens is…

So, you dig around and find even better LEDs and drivers, have a custom optic made that reduces power and produces the best total package. Perhaps not at the 100lm/W you wanted, but close. Customers love it and you can make a profit from it, so off to market you go!

About this time, you will receive a notification from the LED manufacturer that their new version of the LED you are using will be discontinued at the end of the year, to be replaced with a Gen MCMVLXX product that will produce higher lumens by 10%, and operate at 3.1 Vf, while operating at a higher Tj temperature with less lumen loss.  The driver manufacturer will also send notification that it is out of stock on the driver selected, with no firm date when their Chinese vendor will deliver new inventory. The optic manufacturer will then announce that the optic you chose is not compatible with the Gen MCMVLXX LED update, but are working on it. This all means re-testing and re-evaluating all of the decisions made before, and potentially necessitating revisions to UL listing and investigations that could lead to re-testing there.

Welcome to the world of LED product design and development! 

Seriously, while this exploration is fictitious, it does represent the variables that make designing around component manufacturer data unpredictable, if not completely unreliable. There are some take-away’s from this:

  • Never assume you will get the lumens out of an LED that are shown in data sheets. Read all of the data and make corrections that more closely resemble actual application conditions – then subtract another 15% to be safe.
  • Never assume that drivers will deliver exactly the name-plate current and Vf. When in doubt, test the selected product under the conditions it will be used under. Also note that efficiency numbers are generalizations that rarely match actual application conditions. The only way to know what the exact efficiency is, is to test load the driver in question with the intended LED / LED array.
  • Never assume that optical manufacturer efficiency data is correct, or even based on actual test data. The difference in realized optical efficiency and manufacturer data can be significant.
  • Thermal conditions, for the driver as well as LED have a large impact on lumen production and system efficiency. You cannot have a system operate at too low a temperature. Also, remember that data provided is rarely realistic to actual application conditions, so this alone will have a significant impact on system performance.
  • A last side note: Verify system performance under all line voltages anticipated. Just because a driver functions at 120VAC on a bench for 5 minutes, does not mean it will work at 277V after 12 hours continuous operation. Dimming issues are far more common at high voltage (277V) than they are at 120VAC, so test at both, through the entire dimming range.
  • Oh, yeah… dimming driver are notoriously bad at holding efficiency, even at full brightness – and fall off as the product is dimmed from there.
  1. Jim Cuppini says:

    I tried to search for this type of info for the last week, including the DOE, but could not find anything close to this practical approach to designing products…and what to expect

Leave a Reply