Discussion:
LS:12 - White LED 3 Day 4 Report
(too old to reply)
Victor Roberts
2010-07-15 13:19:46 UTC
Permalink
Raw Message
The day started with an Invited Lecture on white LEDs by K. Bando from
Nichia. (Invited Lectures are 30 minute lectures. There are also 15
minute Landmark Lectures and then Posters.)

The abstract for the Nichia talk mentioned a 200 lm/W LED. This turned
out to be a 20 ma laboratory LED that produced light with an efficacy of
197 lm/W at a color temperature of 5500K. The CRI was not given. No
data was given for lower color temperatures.

We were also shown a 6.9 watt LED-based lamp that we were told is a
replacement for a 12 watt CFL, yet the LED-based lamp produced only 380
lumens compared to 810 lumens for the CFL it was replacing.

Near the end of the talk, we were shown an LED array composed of 56 LED
chips, and that produces 10,000 lumens at 105 lm/W at 4000K. No data was
given for lower color temperatures. This lamp operates at 96 watts.
Questions were raised about dissipating the heat from this 96-watt
array, and it became apparent that Nichia has not yet done any life
tests on this high power array. (It's also not clear that it has been
run continuously at 96 watts.)

During the Q&A I asked for confirmation that all this output and power
data was obtained with pulse current measurements at a junction
temperature of 25C. That was confirmed. I then asked what the efficacy
loss would be at a typical junction operating temperature, such as 100C.
I'm not sure the speaker understood the question, because he was first
confused, and then said 20% loss, which seems too much.

During the poster session, I went back to talk to the person with the
amazing magnetic ballast I discussed in the Day 2 report. Here is a
summary of his data:

Lamp: Philips F28T5 (CCT unknown)

Ballast Input Power Lamp Output System
(watts) Power (lumens) Efficacy

His magnetic 32 28.766 2415.5 75.38

Philips 30.95 26.30 2188.1 70.70
EB-S128
TL5 230

Osram 30.90 27.62 2263.8 73.26
QT-FH
1X14-35
230240 CW

The author confirmed that he used the same lamp for all three
measurements, and made the three measurements on the same day in the
same integrating sphere, so that removed one source of error. When
asked how he explained his unusual results, he said that the 10% high
frequency efficacy gain had had only been confirmed in T12 and T8 lamps,
so, perhaps it did not exist in T5 lamps. (He later said that he had
another paper with 40 lamps that showed that the HF gain in T5 was 5% TO
6%, not 10%.)

I told him that it was possible, but unlikely, that he had discovered
something, but when you are disputing 40 years of data, you have an
extra obligation to make sure you do not have any errors.

For example, based on his data, the Philips F28T5 lamp he was using, and
which is rated by Philis for 2900 lumens at 32 watts (103 lm/W) was
operating at only 82 to 83 lm/W in his experiment. That, by itself, is
enough to raise questions about his measurement setup, and was enough
for me to decide that this experiment is fatally flawed.

Vic
Victor Roberts
2010-07-16 14:31:52 UTC
Permalink
Raw Message
Post by Victor Roberts
The day started with an Invited Lecture on white LEDs by K. Bando from
Nichia. (Invited Lectures are 30 minute lectures. There are also 15
minute Landmark Lectures and then Posters.)
The abstract for the Nichia talk mentioned a 200 lm/W LED. This turned
out to be a 20 ma laboratory LED that produced light with an efficacy of
197 lm/W at a color temperature of 5500K. The CRI was not given. No
data was given for lower color temperatures.
We were also shown a 6.9 watt LED-based lamp that we were told is a
replacement for a 12 watt CFL, yet the LED-based lamp produced only 380
lumens compared to 810 lumens for the CFL it was replacing.
Near the end of the talk, we were shown an LED array composed of 56 LED
chips, and that produces 10,000 lumens at 105 lm/W at 4000K. No data was
given for lower color temperatures. This lamp operates at 96 watts.
Questions were raised about dissipating the heat from this 96-watt
array, and it became apparent that Nichia has not yet done any life
tests on this high power array. (It's also not clear that it has been
run continuously at 96 watts.)
During the Q&A I asked for confirmation that all this output and power
data was obtained with pulse current measurements at a junction
temperature of 25C. That was confirmed. I then asked what the efficacy
loss would be at a typical junction operating temperature, such as 100C.
I'm not sure the speaker understood the question, because he was first
confused, and then said 20% loss, which seems too much.
During the poster session, I went back to talk to the person with the
amazing magnetic ballast I discussed in the Day 2 report. Here is a
Lamp: Philips F28T5 (CCT unknown)
Ballast Input Power Lamp Output System
(watts) Power (lumens) Efficacy
His magnetic 32 28.766 2415.5 75.38
Philips 30.95 26.30 2188.1 70.70
EB-S128
TL5 230
Osram 30.90 27.62 2263.8 73.26
QT-FH
1X14-35
230240 CW
The author confirmed that he used the same lamp for all three
measurements, and made the three measurements on the same day in the
same integrating sphere, so that removed one source of error. When
asked how he explained his unusual results, he said that the 10% high
frequency efficacy gain had had only been confirmed in T12 and T8 lamps,
so, perhaps it did not exist in T5 lamps. (He later said that he had
another paper with 40 lamps that showed that the HF gain in T5 was 5% TO
6%, not 10%.)
I told him that it was possible, but unlikely, that he had discovered
something, but when you are disputing 40 years of data, you have an
extra obligation to make sure you do not have any errors.
For example, based on his data, the Philips F28T5 lamp he was using, and
which is rated by Philis for 2900 lumens at 32 watts (103 lm/W) was
operating at only 82 to 83 lm/W in his experiment. That, by itself, is
enough to raise questions about his measurement setup, and was enough
for me to decide that this experiment is fatally flawed.
Vic
I should add a bit of editorial content for anyone not familiar with the
lighting and LED industries.

It can be a bit confusing that the performance of "traditional" lamps
and LED are measured under very different conditions. The output and
efficacy of traditional light sources is measured with the sources fully
warmed up and operating in under staeady state conditions. While the
performance of the lamp may change when operated in a different ambient
temperature, most notably for fluorescent lamps, the published
performance is typically what you will get from that source under actual
operation.

The published data for LEDs, on the other hand, is measured when the LED
junction is at a temperature of 25C (room temperature.) This measurement
is made using a short current pulse so that the LED junction temperature
remains essentially constant during the measurement. However, when
LEDs are used to produce light, the junction temperature rises well
above 25C and the efficiency of light production drops. Therefore, to
determine the performance of an LED-based light source you need to start
with the performance at 25C and then make any adjustments for the
temperature of the LED junction during steady state operation.

The change in light production as a function of junction temperature is
listed on data sheets provided by the major LED manufacturers. For a
typical LED the drop in efficiency is 5% to 10% at 100C relative to 25C.

There is also the pesky issue that lower CCT LEDs have lower efficacy
than higher CCT LEDs. This is why LED manufacturers tend to give the
efficacy for high CCT LEDs. If your application requires a low CCT, you
will not get the same efficacy as given in the press release.
(Traditional light sources tend to have just about the same efficacy for
all available CCTs.)

An additional issue is that when LEDs are manufactured, the performance
of individual LEDs can vary over quite a large range; sometimes as much
as 2:1. The manufacturers report the very best results, but have to
seel all the LEDs they make, or else throw the low performance units
away and charge much more for the the best ones. So, the LED performance
you see in a press release is not what is typically for sale.

The performance of traditional lamps, on the other hand, is much more
consistent. I believe the standard deviation for lumen output at rated
power is less than 5%, but I don't have data to support this. So, what
is published is basically what you get.

Vic
Kappa7
2013-09-19 16:04:07 UTC
Permalink
Raw Message
Post by Victor Roberts
The day started with an Invited Lecture on white LEDs by K. Bando from
Nichia. (Invited Lectures are 30 minute lectures. There are also 15
minute Landmark Lectures and then Posters.)
The abstract for the Nichia talk mentioned a 200 lm/W LED. This turned
out to be a 20 ma laboratory LED that produced light with an efficacy of
197 lm/W at a color temperature of 5500K. The CRI was not given. No
data was given for lower color temperatures.
We were also shown a 6.9 watt LED-based lamp that we were told is a
replacement for a 12 watt CFL, yet the LED-based lamp produced only 380
lumens compared to 810 lumens for the CFL it was replacing.
Near the end of the talk, we were shown an LED array composed of 56 LED
chips, and that produces 10,000 lumens at 105 lm/W at 4000K. No data was
given for lower color temperatures. This lamp operates at 96 watts.
Questions were raised about dissipating the heat from this 96-watt
array, and it became apparent that Nichia has not yet done any life
tests on this high power array. (It's also not clear that it has been
run continuously at 96 watts.)
During the Q&A I asked for confirmation that all this output and power
data was obtained with pulse current measurements at a junction
temperature of 25C. That was confirmed. I then asked what the efficacy
loss would be at a typical junction operating temperature, such as 100C.
I'm not sure the speaker understood the question, because he was first
confused, and then said 20% loss, which seems too much.
During the poster session, I went back to talk to the person with the
amazing magnetic ballast I discussed in the Day 2 report. Here is a
Lamp: Philips F28T5 (CCT unknown)
Ballast Input Power Lamp Output System
(watts) Power (lumens) Efficacy
His magnetic 32 28.766 2415.5 75.38
Philips 30.95 26.30 2188.1 70.70
EB-S128
TL5 230
Osram 30.90 27.62 2263.8 73.26
QT-FH
1X14-35
230240 CW
The author confirmed that he used the same lamp for all three
measurements, and made the three measurements on the same day in the
same integrating sphere, so that removed one source of error. When
asked how he explained his unusual results, he said that the 10% high
frequency efficacy gain had had only been confirmed in T12 and T8 lamps,
so, perhaps it did not exist in T5 lamps. (He later said that he had
another paper with 40 lamps that showed that the HF gain in T5 was 5% TO
6%, not 10%.)
I told him that it was possible, but unlikely, that he had discovered
something, but when you are disputing 40 years of data, you have an
extra obligation to make sure you do not have any errors.
For example, based on his data, the Philips F28T5 lamp he was using, and
which is rated by Philis for 2900 lumens at 32 watts (103 lm/W) was
operating at only 82 to 83 lm/W in his experiment. That, by itself, is
enough to raise questions about his measurement setup, and was enough
for me to decide that this experiment is fatally flawed.
Vic
I know that 3 years are passed, but just in case here you find an experiment run on 40 28W T5 lamps on magnetic L-C ballasts with interesting results:
http://hub.hku.hk/bitstream/10722/155605/1/Content.pdf

Kappa7

Loading...