Maker Pro
Maker Pro

San Jose LED Streetlamps

M

Martin Brown

Jan 1, 1970
0
This is continuing from a discussion from an astronomy forum where the
"massive" energy savings of 80% were claimed by San Jose for changing
from astronomer friendly low pressure sodium SOX lights to white LEDs.
Citing the following article in the Mr Roadshow column of the Mercury
News. But the numbers just do not add up. Anyone care to comment?

http://www.mercurynews.com/mr-roadshow/ci_14521828

I can't see any obvious reason why European experience of these lights
does not apply in the USA.

Low pressure sodium gives something like 160 lumens/watt and almost 200
l/W with optimised solid state ballast. They are mandated near world
class observatories (in this case Lick) because they are almost
monochromatic narrowband emitters and can be easily filtered out.

Best power LEDs are close to 100 lumens/watt now.

I looked up the PE&G report advocating this change and found that they
reckon to power a 55W SOX tube requires 92.5W so I guess they used a
half dead antidiluvian magnetic choke ballast as their baseline. I
reckon a typical modern ballast would be more like 10W at most and could
be easily half that.

http://www.etcc-ca.com/component/co...street-lighting-and-network-controls-san-jose

The legal disclaimer on page 1 rings alarm bells for me.

They go on to compare this bogus highly inefficient baseline SOX setup
with the luminous efficiency of the bare LED units without DC control
gear in the first table I. And so claim spuriously large energy savings
from making the conversion. The estimates further into the document are
slightly more realistic and by table XVII they do include control gear
for the LED systems (with some optimistic assumptions).

Nowhere do the numbers approach the 80% in the news article though.

Their numbers in W are
SOX 92.5 (extremely high for a 55W SOX tube)
LED 100% 75
LED 75% 52.4
LED 50% 34.9

My estimate is
SOX* 65 (Philips electromagnetic ballast)

And so by the time you add in the overhead of 5-10% for control gear I
reckon the 100% LEDs will use almost as much energy as the original
installation is allegedly using at present. I reckon they are in for a
nasty surprise after they finish spending $50M on these new
streetlights. Retrofitting decent ballasts would be a far better choice
with a saving of 30% power for a $20 new ballast component retrofit.

Is there some obvious reason why modern SOX ballasts do not work well on
US mains or is this an attempt to make the old lamps look like they are
very inefficient so as to create momentum for a change to LEDs?

I am aware that some residents of San Jose hate the low pressure sodium
lights on largely cosmetic grounds of poor colour rendering.

There is undoubtedly a maintenance saving since in theory at least since
LED based luminaires should last 3-4x longer than SOX tubes.

Regards,
Martin Brown
 
D

Don Klipstein

Jan 1, 1970
0
This is continuing from a discussion from an astronomy forum where the
"massive" energy savings of 80% were claimed by San Jose for changing
from astronomer friendly low pressure sodium SOX lights to white LEDs.
Citing the following article in the Mr Roadshow column of the Mercury
News. But the numbers just do not add up. Anyone care to comment?

http://www.mercurynews.com/mr-roadshow/ci_14521828

I can't see any obvious reason why European experience of these lights
does not apply in the USA.

Low pressure sodium gives something like 160 lumens/watt and almost 200
l/W with optimised solid state ballast. They are mandated near world
class observatories (in this case Lick) because they are almost
monochromatic narrowband emitters and can be easily filtered out.

Best power LEDs are close to 100 lumens/watt now.

I looked up the PE&G report advocating this change and found that they
reckon to power a 55W SOX tube requires 92.5W so I guess they used a
half dead antidiluvian magnetic choke ballast as their baseline. I
reckon a typical modern ballast would be more like 10W at most and could
be easily half that.

<SNIP from here to edit for space>

I do agree that the advantage of the LEDs is overstated as is often the
case. However, the low pressure sodium is not quite 160 lumens/watt, but
145 in new condition. As the lamp ages, light output is largely constant,
and its power consumption increases slightly.

One significant disadvantage of sodium lighting is low scotopic/photopic
ratio. At streetlighting illumination levels of a few to several lux,
human vision is usually mesopic, which means both photopic and scotopic
vision are significantly functioning.

Low pressure sodium has an s/p ratio of .23, high pressure sodium has
s/p ratio usually around .62-.65, and higher luminous efficacy white LEDs
have s/p ratio usually 1.7-2.1. At streetlighting illumination levels,
light sources with higher s/p ratio can effectively provide a given degree
of illumination with less illumination according to the usual photometric
units such as lumens, candela and lux.

However, I seriously agree with doubting the figure for 80% energy
consumption reduction, even with higher s/p ratio light and improved
ability to direct where the light goes.

However, in the area where light pollution would affect the observatory
(mentioned somewhere in what I snipped out), I would keep the low pressure
sodium.

- Don Klipstein ([email protected])
 
T

Tim Williams

Jan 1, 1970
0
Don Klipstein said:
One significant disadvantage of sodium lighting is low scotopic/photopic
ratio. At streetlighting illumination levels of a few to several lux,
human vision is usually mesopic, which means both photopic and scotopic
vision are significantly functioning.

Low pressure sodium has an s/p ratio of .23, high pressure sodium has
s/p ratio usually around .62-.65, and higher luminous efficacy white LEDs
have s/p ratio usually 1.7-2.1. At streetlighting illumination levels,
light sources with higher s/p ratio can effectively provide a given degree
of illumination with less illumination according to the usual photometric
units such as lumens, candela and lux.

Obviously then, they should install loads of gallium phosphide. Green is
supposed to be the brightest percieved color.

Maybe they should start in Sweden.. they'd be more comfortable with the
"Borg" look, right? ;-)
However, in the area where light pollution would affect the observatory
(mentioned somewhere in what I snipped out), I would keep the low pressure
sodium.

Hmm, monochromatic LEDs are still pretty easy to filter. GaP linewidth
might be 20nm, more than the laserlike sodium doublet, but not as bad as
phosphor anything (white LEDs, fluorescents, etc.).

Tim
 
D

Dirk Bruere at NeoPax

Jan 1, 1970
0
This is continuing from a discussion from an astronomy forum where the
"massive" energy savings of 80% were claimed by San Jose for changing
from astronomer friendly low pressure sodium SOX lights to white LEDs.
Citing the following article in the Mr Roadshow column of the Mercury
News. But the numbers just do not add up. Anyone care to comment?

http://www.mercurynews.com/mr-roadshow/ci_14521828

I can't see any obvious reason why European experience of these lights
does not apply in the USA.

Low pressure sodium gives something like 160 lumens/watt and almost 200
l/W with optimised solid state ballast. They are mandated near world
class observatories (in this case Lick) because they are almost
monochromatic narrowband emitters and can be easily filtered out.

Best power LEDs are close to 100 lumens/watt now.

I looked up the PE&G report advocating this change and found that they
reckon to power a 55W SOX tube requires 92.5W so I guess they used a
half dead antidiluvian magnetic choke ballast as their baseline. I
reckon a typical modern ballast would be more like 10W at most and could
be easily half that.

http://www.etcc-ca.com/component/co...street-lighting-and-network-controls-san-jose


The legal disclaimer on page 1 rings alarm bells for me.

They go on to compare this bogus highly inefficient baseline SOX setup
with the luminous efficiency of the bare LED units without DC control
gear in the first table I. And so claim spuriously large energy savings
from making the conversion. The estimates further into the document are
slightly more realistic and by table XVII they do include control gear
for the LED systems (with some optimistic assumptions).

Nowhere do the numbers approach the 80% in the news article though.

Their numbers in W are
SOX 92.5 (extremely high for a 55W SOX tube)
LED 100% 75
LED 75% 52.4
LED 50% 34.9

My estimate is
SOX* 65 (Philips electromagnetic ballast)

And so by the time you add in the overhead of 5-10% for control gear I
reckon the 100% LEDs will use almost as much energy as the original
installation is allegedly using at present. I reckon they are in for a
nasty surprise after they finish spending $50M on these new
streetlights. Retrofitting decent ballasts would be a far better choice
with a saving of 30% power for a $20 new ballast component retrofit.

Is there some obvious reason why modern SOX ballasts do not work well on
US mains or is this an attempt to make the old lamps look like they are
very inefficient so as to create momentum for a change to LEDs?

I am aware that some residents of San Jose hate the low pressure sodium
lights on largely cosmetic grounds of poor colour rendering.

There is undoubtedly a maintenance saving since in theory at least since
LED based luminaires should last 3-4x longer than SOX tubes.

Regards,
Martin Brown

Also factor in lifetime and cost of replacement.
How much does it cost to replace one streetlamp?
How often?
It's going to take at least 2 guys, a special truck and traffic
management for each instance. Also factor in the cost of disruption of
traffic.
 
P

Paul Keinanen

Jan 1, 1970
0
One potential for truly massive energy savings is actually real - if
implemented.
Lights with patterns defined at installation time.
Imagine an array of LEDs, each with a 8 degree optic, pointed so as
they overlap slightly. These are configured at install time to only
light the required elements.

Current street light fixtures are really bad, outputting light in
placed, in which it is not needed, some even output directly into the
sky causing light pollution. Others output a lot of light close to the
horizontal plain, falling directly into the eyes of pedestrians and
drivers. This will destroys the night adaption of the eye and higher
illumination levels on the ground is required.

Unfortunately, the current street infrastructure with light poles at
every 30 m is quite suboptimal. Either the distance between light
poles should be similar to the pole height or a horizontal bar should
be suspended above the street center line and the light fixtures can
then be installed every few meters on the bar. In such a situation,
LEDs would work quite effectively.

The life expectance for various LEDs used for illumination is strongly
dependent on the thermal conditions of the LED itself. To reach such
huge numbers quoted would require connecting it to a good heatsink
which is cooled by ice cold water :). With a realistic heat sink in a
light fixture, the life expectance is much shorter.
 
M

Martin Brown

Jan 1, 1970
0
Tim said:
Martin Brown wrote:

How much government money will PE&G get for changing to white LEDs?

Roughly $50M/62k = $800 per unit.
How much government money will PE&G get for swapping out ballasts?

Probably none.

You can call me naive if you like, but I had assumed that PE&G were just
selling the electricity that the street lights consume. And that they
were acting as consulting engineers to the city council.

I did wonder if they had already changed to electronic ballasts and were
stiffing the city for 90W for each 55W lamp according to the original
contract whilst supplying only 60W with the new ballasts. Seems pretty
odd that a city that cannot afford to run all its installed low pressure
sodium lights can afford to spend $50M on a boondoggle like this.

At least in a city with mercury lighting there would be a prospect of
making real energy savings with LED technology.

Regards,
Martin Brown
 
M

Martin Brown

Jan 1, 1970
0
Ian said:
Not addressing all of this.

Best power LEDs are more like 140lm/W.
This is however not at their max output.

I have a torch that uses one and is highly collimated. I hadn't realised
they derate with power quite that fast and only looked at max output.
Best I had seen was 120lm/W under lab conditions and infinite heat sink.

Lumiled also seem to have announced some 140lm/W devices.

Incidentally does anyone know if the quantum efficiency of modern LEDs
still improves further when you dunk them in LN2 ?
(that is if the thermal shock doesn't kill them)
For example the Cree XP-G range can do 140lm/W at 1W, though about
80lm/W at 5W.

When they can do 140lm/W at 5-10W they will be competitive as spotlamp
replacements (a position where CFLs are truly dreadful). Pricing is
still an issue even allowing for the longer lifetime.
One potential for truly massive energy savings is actually real - if
implemented.
Lights with patterns defined at installation time.
Imagine an array of LEDs, each with a 8 degree optic, pointed so as
they overlap slightly. These are configured at install time to only
light the required elements.

The built in collimation makes them excellent for roadside signalling.
very bright at long range and not too dazzling when you get close to.

I don't doubt their day will come, but I reckon the early adopters like
San Jose are headed for a very nasty surprise.
95% efficient ballasts are available.

And I presume that also holds for solid state HF SOX ballasts too.
Certainly low voltage ones exist at 95% efficiency, but the common mains
SOX ballasts like Philips BSX355 waste about 10W (and not 45W).

Regards,
Martin Brown
 
M

Martin Brown

Jan 1, 1970
0
Robert said:
?? "Low pressure sodium (lamps) are almost monochromatic narrowband
emitters" ??

YES. Almost 95% of their visible light output is in the yellow D-lines.
The next strongest line in the visible is less than 0.5% of the total.

Someone has drawn the spectra of the elements and various light sources
with a linear scale spectral graph. You can just about see the third
brightest visible line if you know where to look (569nm).

http://ioannis.virtualcomposer2000.com/spectroscope/elements.html#sodium

There is a 10% line output in the near IR around 820nm but modern tubes
filter it with an InO window to improve thermal efficiency.
Hmmm..a number of nice yellow bands, a number of red bands, a number
of green bands, not to mention a few lonely ones here and there.
Yup...monochromatic...

In power terms almost everything goes into the 2 sodium D-lines at 589nm
and 589.6nm. There are some faint spectral lines but they are negligible
(and unlike mercury they don't sit on any astrophysically important
wavelengths). Low pressure sodium light is pure enough to demonstrate
interference effects in lab experiments requiring monochromatic light.

By any reasonable definition it is monochromatic light.

Regards,
Martin Brown
 
P

Paul Keinanen

Jan 1, 1970
0
Best power LEDs are more like 140lm/W.
This is however not at their max output.
For example the Cree XP-G range can do 140lm/W at 1W, though about
80lm/W at 5W.

At least Cree seems to specifies the characteristics at 25 C junction
temperature. In the real world, this is a valid assumption only during
a few seconds at turn on. The output drops by 20-30 % for higher
junction temperatures.

It seems to be very hard to get any reliable MTBF figures, but
assuming that the MTFB is halved for each 10 C increase in
temperature, which is typical for semiconductors, you really have to
operate the device well below the maximum junction temperature.

For long life, you will have to operate the device below 100 C
junction temperature. From the heatsink design point of view, this
resembles the problems when building audio power amplifiers using
germanium transistors a few decades ago :).

In order to obtain the claimed efficiency and long life, you really
have to operate a "5 W" device at 1 W, thus in order to get the
expected illumination, you have to use 5 times the number of devices,
thus making it even less economically attractive.

When reading specifications for power LEDs, you must be very careful
to check the actual condition in which the claims are valid. While the
marketing claims for some ordinary semiconductor part might be a bit
excessive, the marketing claims are much worse in the power LED
industry.

By the way, San Jose and other similar places close to the equator are
nasty places for LED street lights due to the long warm nights, making
it hard to dissipate the heat into the environment. While at higher
latitudes, street lights might be needed for 18-24 hours during the
winter, transferring the heat into subzero ambient temperature is much
easier. During the summer, only a few hours (if any) is needed and
even then, the heat is easily dumped into a moderate temperature air.
 
M

Martin Brown

Jan 1, 1970
0
Don said:
<SNIP from here to edit for space>

I do agree that the advantage of the LEDs is overstated as is often the
case. However, the low pressure sodium is not quite 160 lumens/watt, but
145 in new condition. As the lamp ages, light output is largely constant,
and its power consumption increases slightly.

I think it also depends a bit on the lamp size and ballast. Some of the
newest HF ballasts are claiming 200lm/W but at reduced operating power
and as a result longer tube life.
One significant disadvantage of sodium lighting is low scotopic/photopic
ratio. At streetlighting illumination levels of a few to several lux,
human vision is usually mesopic, which means both photopic and scotopic
vision are significantly functioning.

Low pressure sodium has an s/p ratio of .23, high pressure sodium has
s/p ratio usually around .62-.65, and higher luminous efficacy white LEDs
have s/p ratio usually 1.7-2.1. At streetlighting illumination levels,
light sources with higher s/p ratio can effectively provide a given degree
of illumination with less illumination according to the usual photometric
units such as lumens, candela and lux.

Agreed. I suspect they will be landed with a pig inn a poke.
However, I seriously agree with doubting the figure for 80% energy
consumption reduction, even with higher s/p ratio light and improved
ability to direct where the light goes.

However, in the area where light pollution would affect the observatory
(mentioned somewhere in what I snipped out), I would keep the low pressure
sodium.

- Don Klipstein ([email protected])

Lick seem to be under the impression that they will be protected by
somewhat exotic amber LEDs with 6nm fwhm output late at night. I'd be
interested in finding sample of these super amber high power LEDs as the
best I can see are typically ~20nm fwhm centred on 590nm.

Ideal would be 585nm fwhm <10nm as a crude solid state demo sodium light
(and much less fragile to carry around).

Regards,
Martin Brown
 
M

Martin Brown

Jan 1, 1970
0
Is there still a LPS lamp production line anywhere in the world? Its
been so long since I've seen active LPS in the North Eastern US..

US always was a mercury vapour country. UK and Belgium still have about
50% installed base of LPS with HPS making inroads into city centres.
They date from the 60's when the yellow light fog penetration was a
benefit and a second wave of them installed in the mid 70's for energy
efficiency during the oil shocks and 3 day week power cuts era.

Someone must still be making them. Still a stock item at RS although
their price is not the best.

Mercury street lighting is rare here although is used in my village.

Regards,
Martin Brown
 
P

Paul Keinanen

Jan 1, 1970
0
It's not quite that bad.
The above LEDs (IIRC) (cree XPG) quote 80% lumen maintainance at 70000
hours at 70C.

What exactly does this 70 C stand for ?

Assuming 3.3 V drop at 1 A (the maximum current specified for long
life specified by Cree *) the power dissipation is 3.3 W and with
junction to "solder point" thermal resistance of 6 C/W, the junction
is 20 C hotter. If the solder point temperature is 70 C, the junction
temperature is 90 C as in the old germanium transistor days :).
For any reasonable sized streetlamp, operating at night in ambients of
under 30C, this is not hugely challenging - though it does require some
thoguht.

Getting 3.3 W out of a 70 C case to a 30 C ambient temperature is
comparatively easy, just requiring a heatsink with (70-30)/3.3 W or 12
C/W thermal resistance, which should be easy to achieve.

Unfortunately, in order to produce something resembling 3000 lumens at
the light pole, a large number of LED devices would be required,
making it impossible to produce in a confined space with
LEDs+heatsinks.

*) From the Cree documentation:
 
D

Dirk Bruere at NeoPax

Jan 1, 1970
0
It's not quite that bad.
The above LEDs (IIRC) (cree XPG) quote 80% lumen maintainance at 70000
hours at 70C.

About 3 to 4 times better than sodium
 
D

Don Klipstein

Jan 1, 1970
0
Obviously then, they should install loads of gallium phosphide. Green is
supposed to be the brightest percieved color.

Photometric units (candela, lumen, lux) already take into account the
photopic response of "standard average human vision".
Maybe they should start in Sweden.. they'd be more comfortable with the
"Borg" look, right? ;-)


Hmm, monochromatic LEDs are still pretty easy to filter. GaP linewidth
might be 20nm, more than the laserlike sodium doublet, but not as bad as
phosphor anything (white LEDs, fluorescents, etc.).

More efficient LEDs have bandwidth around 20-25 nm FWHM for blue,
around 30-35 nm for green, and maybe 20-30 nm red and orange. Significant
spectral content exists for a bandwidth at least double the FWHM
bandwidth.

If I needed monochromatic lighting for around an observatory, I would
use low pressure sodium for:

1. Luminous efficacy greater than that of any colored LED on the market
2. Bandwidth a goodly 50-60 nm less
3. The "monochromatic" LEDs with highest luminous efficacy are green
ones, and green is scattered more by the atmosphere than the longer
orangish yellow wavelength of low pressure sodium.

As it turns out, the LEDs on the market with highest luminous efficacy
so far are white ones with blue-emitting chips and greenish-yellow-glowing
phosphor. I have heard of some green LED prototypes using a phosphor
(and I assume a blue LED chip) having lumens out per watt in as high as
notable recent white laboratory prototypes.

- Don Klipstein ([email protected])
 
D

Don Klipstein

Jan 1, 1970
0
Current street light fixtures are really bad, outputting light in
placed, in which it is not needed, some even output directly into the
sky causing light pollution. Others output a lot of light close to the
horizontal plain, falling directly into the eyes of pedestrians and
drivers. This will destroys the night adaption of the eye and higher
illumination levels on the ground is required.

Unfortunately, the current street infrastructure with light poles at
every 30 m is quite suboptimal. Either the distance between light
poles should be similar to the pole height or a horizontal bar should
be suspended above the street center line and the light fixtures can
then be installed every few meters on the bar. In such a situation,
LEDs would work quite effectively.

The life expectance for various LEDs used for illumination is strongly
dependent on the thermal conditions of the LED itself. To reach such
huge numbers quoted would require connecting it to a good heatsink
which is cooled by ice cold water :). With a realistic heat sink in a
light fixture, the life expectance is much shorter.

There are now LED manufacturers, especially Lumileds, who publish
projected life expectancy of lighting class LEDs at temperatures that
are very reasonable. I will have to dig back in, but I seem to recall
more recent white Luxeon ones being stated by Lumileds to be projected
to maintain 70% of their initial light output at 50,000 operatingh hours
with (I forget which) 350 or 700 mA current and some figure of peak
junction temperature that I now forget, ??? probably somewhere in the
range of 100-120 C, definitely hotter than 85 C.

(The PDF datasheet only reads in versions of Acrobat higher than the
one on my partition with an older operating system and the program that
my PC-based e-mail program is on - so I can't get to it just now.)

I can't remember exact thermal resistance of these well now either.
But I am confident that the junction temperature is sufficiently cool
for this 50,000 hours life expectancy (based on accelerated testing of
products invented less than 50,000 hours ago) when heatsink temperature
is 60 C, probably 75 C.

On the other hand, optical characteristics of these are rated for
junction temperature of 25 C. With real-world thermal conditions,
expect a few to several percent less light output than indicated by
the "25 C junction" figures in the datasheet for white Luxeons.

- Don Klipstein ([email protected])
 
D

Don Klipstein

Jan 1, 1970
0
Robert Baer wrote said:
Martin Brown wrote
?? "Low pressure sodium (lamps) are almost monochromatic narrowband
emitters" ??
Hmmm..a number of nice yellow bands, a number of red bands, a number
of green bands, not to mention a few lonely ones here and there.
Yup...monochromatic...

A good 99.5%-plus of the visible spectrum content of SOX series low
pressure sodium lamps is in the 589.0 and 589.6 nm lines in the orangish
yellow. Both of these spectral features, like the other notable (notably
around 3 orders of magnitude weaker) spectral features of low pressure
sodium, have bandwidth less than .1 nm.

So, a good 99.5%-plus of the visible spectral content of low pressure
sodium is in a closely spaced pair of narrow lines, where the total
bandwidth of that spectral feature is somewhere around .7 nm. The visible
spectrum, according to the most common definition, is 400-700 nm.

I know of one orangish-red line pair, nothing between that and 800 nm in
the very near IR, and 3 overlapping series of very weak line pairs that
produce the orangish-red line pair and plenty from very-yellowish-green to
bluish violet. Most of this is in the orangish-red line pair, a very
yellowish green line pair, and a very bluish green line pair. This whole
set of 3 series accounts for only a few tenths of a percent of visible
spectrum content of SOX series low pressure sodium lamps.

- Don Klipstein ([email protected])
 
D

Don Klipstein

Jan 1, 1970
0
YES. Almost 95% of their visible light output is in the yellow D-lines.
The next strongest line in the visible is less than 0.5% of the total.

I would say more extreme than that by an order of magnitude, as in
more like 99.5% of visible spectrum content of low pressur sodium
being in one line pair having total bandwidtyh less than 1 nm, and the
next one down probably has less than .2% of the total.
Someone has drawn the spectra of the elements and various light sources
with a linear scale spectral graph. You can just about see the third
brightest visible line if you know where to look (569nm).

The 2nd-place visible spectrum feature of a low pressure sodium lamp
is close to there, comprised of 4 lines: 567.0, 567.6, 568.2 and 568.8
nm. The first two of those are much stronger than the last two.

I saw that...

The Na entry has only one of the line pairs shown at all, the main one
in the orangish yellow.

This source cites 2nd place being .7%, which sounds a goog half an order
of magnitude high to me if it is a visible spectrum feature. (Likely
according to a spectrometer experiencing saturation or having uncorrected
non-flat spectral response or both.)
There is a 10% line output in the near IR around 820nm but modern tubes
filter it with an InO window to improve thermal efficiency.


In power terms almost everything goes into the 2 sodium D-lines at 589nm
and 589.6nm. There are some faint spectral lines but they are negligible
(and unlike mercury they don't sit on any astrophysically important
wavelengths). Low pressure sodium light is pure enough to demonstrate
interference effects in lab experiments requiring monochromatic light.

By any reasonable definition it is monochromatic light.

Regards,
Martin Brown

- Don Klipstein ([email protected])
 
D

Don Klipstein

Jan 1, 1970
0
At least Cree seems to specifies the characteristics at 25 C junction
temperature.

Lumileds does the same.
In the real world, this is a valid assumption only during
a few seconds at turn on. The output drops by 20-30 % for higher
junction temperatures.

For white LEDs with reasonably accomplishable heatsinks in reasonable
ambient temperatures (even 30 C), I would say dropping by more like 10%,
maybe 15% if pushing for smaller heatsink size.
It seems to be very hard to get any reliable MTBF figures, but
assuming that the MTFB is halved for each 10 C increase in
temperature, which is typical for semiconductors, you really have to
operate the device well below the maximum junction temperature.

Lumileds states life expectancy figures with a requirement of junction
temperature being cooler than the 150 C or whatever maximum, but still
fairly hot in (IIRC) the 100-120 C range.
For long life, you will have to operate the device below 100 C
junction temperature. From the heatsink design point of view, this
resembles the problems when building audio power amplifiers using
germanium transistors a few decades ago :).

In older days, non-white LEDs were supposed to achieve 100,000 hour
life expectancy if current did not substantially exceed "characterization
current" and junction temperature did not exceed 85 C. Lately, 100
degrees C junction temperature or even a little more is OK, although
non-phosphor LEDs of color yellow-green to red tend to have substantial
negative temperature coefficient for photometric performance.
In order to obtain the claimed efficiency and long life, you really
have to operate a "5 W" device at 1 W, thus in order to get the
expected illumination, you have to use 5 times the number of devices,
thus making it even less economically attractive.
When reading specifications for power LEDs, you must be very careful
to check the actual condition in which the claims are valid. While the
marketing claims for some ordinary semiconductor part might be a bit
excessive, the marketing claims are much worse in the power LED
industry.

By the way, San Jose and other similar places close to the equator are
nasty places for LED street lights due to the long warm nights, making
it hard to dissipate the heat into the environment. While at higher
latitudes, street lights might be needed for 18-24 hours during the
winter, transferring the heat into subzero ambient temperature is much
easier. During the summer, only a few hours (if any) is needed and
even then, the heat is easily dumped into a moderate temperature air.

I would worry about hot nights far to the east of San Jose. I seem to
think of typical July and August nighttime temperature in San Jose around
11-14 C. Philadelphia, farther north, has average nighttime *minimum*
temperature (around sunrise) for July and August being 19-20 C at the
official weather station location and about 21 C in Center City. This
figure is probably 23-24 C in Orlando, Houston and New Orleans. Central
city Phoenix in July and first half of August is also very warm at night.

- Don Klipstein ([email protected])
 
D

Don Klipstein

Jan 1, 1970
0
I think it also depends a bit on the lamp size and ballast. Some of the
newest HF ballasts are claiming 200lm/W but at reduced operating power
and as a result longer tube life.

I agree that this is likely true, same as for fluorescent and germicidal
low pressure mercury vapor lamps. The similarity among these is in
radiation resulting from atomic electron transitions between states where
the lower state is "the ground state".
Agreed. I suspect they will be landed with a pig inn a poke.

Lick seem to be under the impression that they will be protected by
somewhat exotic amber LEDs with 6nm fwhm output late at night. I'd be
interested in finding sample of these super amber high power LEDs as the
best I can see are typically ~20nm fwhm centred on 590nm.

Ideal would be 585nm fwhm <10nm as a crude solid state demo sodium light
(and much less fragile to carry around).

Regards,
Martin Brown

- Don Klipstein ([email protected])
 
M

Martin Brown

Jan 1, 1970
0
Michael said:
Really? What about the decades of incandescent street lights, and
the use of other technologies I've seen. Hell, we've had gas lights at
times. Both carbide and natural gas.

In terms of the skywards light pollution from street lighting the
dominant US component is from mercury light followed by HPS.

Notable exceptions were the Golden Gate bridge which for a long time was
lit by LPS and regions surrounding world class observatories.

In the UK you can still find a few vintage gas lights in London if you
know where to look, but they contribute almost nothing to light
pollution. I don't recall ever seeing a UK road with incandescent
streetlamps or for that matter an American one.

The worst offending private lights are wall mounted 1kW quartz halogen
insecurity lights in crude tin cans so half their light goes upwards.

Regards,
Martin Brown
 
Top