wellllllllll mercury vapor lamps can maintain 80% of lumens at 60,000 hours as it has already been documented with such lamps made in the 60s-80s.

Maybe 80% of the visibility, but the light output is really around 20..30%. The drop is noticeable, but the visibility is worse only slightly, so it does not appear as bad. The output would have to drop to 1/10 or even less, in order to really impact the visibility. Eyes are very adaptive and unless there are glare bombs near by, they suffice with really low illumination level. What is more important is how even is the illumination. And that is not influenced by the lamp aging, but how the fixture optics degrade.

They do light somehow after 175+khours, but that is way far below any useful output.

But they can not be beaten on the reliability: Even after 24khour rating only few percent are really dead. And even from these, most could be tracked down to faulty ballast or installation being the primary cause, destroying the lamp only as the consequence.
So you can really replace all the lamps every four years and then forgot about them for the next 4 years. And if you happen to be delayed, nobody will notice it anyway, as the lamps degrade slowly, so the slight loss of illumination won't be suspicious, as even 50% of the lumens look like only small drop...

by the way an electrician friend of mine said his city is having problems with the induction retrofits in the city parks, the driver units keep failing. he said he is considering changing them back to hid as they fail and their warranties expire.

Well, I would expect that...

In the bottom line, visibility is what matters. Good quality mercs can maintain good light output for 10 years, and relamping once in 10 years is not bad at all

LEDs can outlast mercury lamps (and they are about the only lamp that can do that), but it depends on quality (engineering and manufacturing)



If any lamps are really good for outdoor lighting, i'd say its mercury for the higher poles, fluorescent for wall recessed lights etc, and LED for architectural lighing

Mercury is te least efficient of those, bu its light color is perfect for the night - you can suffice with very low power mercury to get visibility, so no much energy savings in sodium and MH

Now on the power density (watts to fixture size), only the mercury out of all 3 is good for the bigger stuff

So by banning mercury, what is left ? Inappropriate sodium and MH, large size fluorescent, and inappropriate power LED

In the bottom line, visibility is what matters. Good quality mercs can maintain good light output for 10 years, and relamping once in 10 years is not bad at all

Well, the visibility should be with reasonable cost. If you allow the lamp to loose 70% of it's light output and still be OK with the visibility, it would become way cheaper to use a lamp with half of the wattage rating and replace it every 4 years. The same visibility, but half the power consumption... So keeping the lamps for 10 years and claiming they are giving still enough light mean you did the wrong design at the start by specifying too high wattage...

LEDs can outlast mercury lamps (and they are about the only lamp that can do that), but it depends on quality (engineering and manufacturing)

Outlast in median, maybe. But I doubt they would reach the same reliability. Way too many interconnections to eventually fail, even with the magnetic ballast. With the magnetic you are stuck with putting all LED's in series (in order to reach reasonable ballast efficiency it have to drop as little voltage as necessary to still reasonably regulate the current) and that mean any open circuit make the chain to fail, unless you insert backup bypass devices (thyristor + 15V Zener across each triplet), but that make the system more complex again.
So there is still an area, where MV's perform the best and where there is no sound replacement.

If any lamps are really good for outdoor lighting, i'd say its mercury for the higher poles, fluorescent for wall recessed lights etc, and LED for architectural lighing

Well, the LED's excellent optical control ability, together with quite good color properties (for 80..90lm/W you reach Ra of 60..70) would make them the prime choice for regular road lighting, as it already have the capability to become the cheapest light for smaller roads with up to 4m poles.
For larger roads the LED's are still too expensive to reach higher output packages, so the HPS become the cheapest light there. Their higher raw efficacy (140lm/W, compare to ~80lm/W for modern LED's) keep the margin for the worse fixture efficiency (~50% compare to 80% with LED's for pole spacing of 4x their heights) But it's color is indeed not as good.
For in-house soft lighting, the fluorescents beat all other on thge efficacy combined with rather high color rendering and the ability to make any CCT you want. But their sensitivity to temperature make them very poor performers outdoors. Moreover their large surface make them nearly unusable for any road lighting, what want to bear the label "efficient"...

The MV's have either very low efficacy and really poor color (the clear ones), or too large surface (phosphor coated), so do not allow good beam control (the fixture efficiency would be less than 20%, when the pole spacing have to be the usual 4x the pole height). Quite a disadvantage for economic road lighting.

Mercury is te least efficient of those, bu its light color is perfect for the night - you can suffice with very low power mercury to get visibility, so no much energy savings in sodium and MH

Now on the power density (watts to fixture size), only the mercury out of all 3 is good for the bigger stuff

So by banning mercury, what is left ? Inappropriate sodium and MH, large size fluorescent, and inappropriate power LED

Well, the bigger stuff usually does not need as good color, so HPS fit the job way better.

Of course, there are niches, which require special properties:
Tunnel lighting for high reliability, so the only technology meeting that are MV's (but the HPS could close by that when designed with redundancy - and that is the prefered method here)
or dense fog areas: There excel the LPS, however that task could be suitable for special made LED's. But such market is quite small, so would not pay for the special LED type (unlike other sources, the LEDs are extremely costly per manufacturing batch and the impossibility to run smaller quantities, what would make them the cheapest light source is the very high number of final products made form that single manufacturing batch, but you would have to sell them all, in order to be profitable)

Medved: The reason why Ash puts the MV as the best lightsource for roadlighting and streetlighting, beside reliability, is because the humen eyes, have two sensors: The cones for daylight vision and the rods for nightvision. The cones have the peak vision at the orange and red, rods have their peak vision at the blue and green.
So the blue-green (Clear) or the cool daylight (Colour corrected) colour of mercury lamps is best suited for streetlighting, where lot of light isn't required and the light can be at the same intensity as the full moon, so the rods gets more active, which is why the visibility gets better with MV lamps. For a similar visibility with HPS lamps, you need a very high lux level (A residence street should be illuminated with 400-1000W HPS lanterns or floodlights to get the same visibility as with a 80-125W MV lamps, and this is a lot of unuseful light and light pollution.
MV with the emitter that turns the arctube white instead of black, can have an excellent performance for many years, and their light drops very slowly. Also, MV lamps with this emitter, have longer service life.

Medved: The reason why Ash puts the MV as the best lightsource for roadlighting and streetlighting, beside reliability, is because the humen eyes, have two sensors: The cones for daylight vision and the rods for nightvision. The cones have the peak vision at the orange and red, rods have their peak vision at the blue and green.
So the blue-green (Clear) or the cool daylight (Colour corrected) colour of mercury lamps is best suited for streetlighting, where lot of light isn't required and the light can be at the same intensity as the full moon, so the rods gets more active, which is why the visibility gets better with MV lamps. For a similar visibility with HPS lamps, you need a very high lux level (A residence street should be illuminated with 400-1000W HPS lanterns or floodlights to get the same visibility as with a 80-125W MV lamps, and this is a lot of unuseful light and light pollution.

At first that is valid only for really low illumination levels, but in most cases the level is way higher. So the difference is by far not as big, it mean about an equal wattage of both lamps (both coated, to be fair on the lantern's optic efficiency), so 4000..6500lm 80..125W coated MV yield about the same visibility as 6500..15000lm 70..150W HPS and so on.
And it happen, than LED's optimum efficacy is about the 5000..6000K with CRI of about 60, but with way better beam shape, so requiring even less raw lumens...

The major thing is, than the clear MV's have way lower efficacy than is the benefit from the color and the coated yield to very low optical efficiency of the fixture, so the HPS become more efficient. But given the eye characteristic, HPS lighting is more sensitive to the stray beams, as they could easily cause glare.

MV with the emitter that turns the arctube white instead of black, can have an excellent performance for many years, and their light drops very slowly. Also, MV lamps with this emitter, have longer service life.

That mean the light output drop to 50% not after 8000 hours (the old composition), but after 25000hours (the newer, white soot composition). The economic limit of 70% (otherwise it become cheaper to use lower wattage lamps and replace them more frequently) is at about 20000 hour. That is about the useful service life for present HPS as well. Of course the reliability can not match, as after that time there would be already few percent of failures.

Light levels in most places are overkill. 5 lux are more than enough for many places. and its the amount of light where mercury and its white light give it very big advantage vs sodium. That street lighting installations often give 25 lux ad more is not right, but thats what is usually achived with msot sodium systems out there

One of e advantages of white light is, that even when in lux the light uniformity is awfull, it still looks good enough to the eye. Which means that inferior optics are still acceptable for mercury, while not for HPS. This gives advantage to the coated mercury

LED need significant heatsinking at high wattages, whih eventually runs ino the same problem as fluorescent - fixture size, with which comes windage and so on

LED reliability : Dont see any problem with using a magnetic line conditioner, then on its output capacitor-less electronic driver. Well made electronis, on clean power, at low temperature, and without electrolitic capacitors, are very reliable

What I meant is that there have been many well documented cases where mercury vapor lamps reached 50000 hours with 80% lumen maintenance, especially the clear ones where there is no phosphor to degrade. With the superior optical efficiency in American cobra head fixtures they can maintain very good lighting for 10 years or more. Ask any lineman and they say mercs are your best friends on the job.

I get to have the experience of observing the led street lights in Los Angeles, they have been using them since 2009. The ones from 2009 are still working with very few total failures, but I can see they have some lumen depreciation and color shift already.

Mercury vapor lamps last alot longer than high pressure sodium and metal halide ones. It's a shame they probably won't be available in 2016 thanks to the Lighting Dictatorship Association inappropriately named the Dark Sky Association. What's next metal halide?    

Why yes, LEDs dont pollute the sky with light

Why yes, LEDs dont pollute the sky with light

Why they should not? Well, assuming they would be installed in the same lumen outputs as other lights...
But it is true the better optical control mean less spill light and that mean less light pollution.

@silverliner: And how that was documented? Somebody really measured the output? And for more than one "lucky" lamp?
If I open any older brochure containing lumen maintenance graph, all newer show the depression to ~50..60% at 25khours (the end of their life rating), the older one (before the whitening agent was added to the emission mix) drop to 50% at about 10..15khouirs and went down way more steeply when new. And the lumen depression curves are an average for large amount of lamps...

LED's are good for certain applications but not street lighting. With an HID street light, a burned out lamp can simply be unscrewed and replaced with a new one. Whereas with an LED streetlight, you have to take down the whole fixture and order hard to find expensive parts and replacement lamp components.

I recently read that the city of Los Angeles had a total of about 190 failures out of about 100,000 led streetlights installed in the past three years. I have seen a few spot replacements of whole LED heads the replacements are the newer generation lower wattage versions and the difference in light output between these and the 3 year old ones is noticeable for up to 6 months by then the new one will have depreciated.

@medved, what I mean is that in the 60s and 70s mercury lamps happened to be overly well built, lasting much longer with better lumen maintenance than expected. They reengineered the lamps by the early 80s due to much lower than expected sales of new lamps. Do you really believe LEDs will take over and eventually become pretty much the only source of artificial illumination with near 100% conversion of energy into light?

I recently read that the city of Los Angeles had a total of about 190 failures out of about 100,000 led streetlights installed in the past three years. I have seen a few spot replacements of whole LED heads the replacements are the newer generation lower wattage versions and the difference in light output between these and the 3 year old ones is noticeable for up to 6 months by then the new one will have depreciated.

There are two aspects, why the newest LED's would fail more frequently even assuming the LED technology itself would be equal or more reliable than others:
- The LED technology is rather new and their shortcomings and strongholds are not yet fully known, nor are the method to utilize/take care of them in an cost efficient manner. Technology, what is here for more than 60 years have 60years of knowledge behind. Even the MV technology was very unreliable in it's early days compare to the previous incandescents, while till today it evolved into the most reliable general technology known (at least from those known to me; reliable mean you plan a maintenance scheme, follow it blindly for next 20 years and still you have nearly no instances of faults or below-design performance, it does not mean the single lamp would endure the whole 20 years)
- Led lanterns are mostly new. It is known, than the failure rate of any device is somewhat high just after production, but then it drop down. These failures are in fact a defects, what were not captured before leaving the factory. So the relatively new LED fixtures are mostly in the early stage of this development, so many faults are still the hidden manufacturing defects, this is called "early failures". Even the first point (the maturity of the technology) contribute a lot - because the large part of the knowledge is the production testing methodology: What to look for in the production, in order to identify the defects leading to early malfunction. The older installations are way past this point, so the defective pieces were already removed, so the installation is in it's middle age, so with minimum defects.

And only after long time, the failure rate start to grow, as the equipment reach it's end of life. Only the "150x 5mm LED energy saver" retrofit lamps reached this state, but the purpose made LED streetlamps are way far from this age.

And most of mercury lamps the technology is so mature, there are nearly no early failures and they normally loose their output (that is a property, not fault, as it is of systematic nature and could be well predicted on each lamp) below the limit of economic use way before they reach the limit of aging faults, so this combination make from the most reliable devices (but not the longest lasting), because they have no failures or unexpected performance degradation...

@medved, what I mean is that in the 60s and 70s mercury lamps happened to be overly well built, lasting much longer with better lumen maintenance than expected. They reengineered the lamps by the early 80s due to much lower than expected sales of new lamps.

In MV's, the lamp life is a trade off to be made between lifetime and efficacy.
Initially the lumen depression was quite strong. So the power loading of the arc tube had to be limited to ensure reasonable service life, all in order to keep the Total Cost of Ownership the lowest possible.

 Then they introduced the electrode mix, that build the white, instead of black soot, so the maintenance improved. At the same the production cost of the new lamps fell, so lamp cost become less of a factor in the TCO. It improved so much and the behaviour become so predictable, the efficacy compromise made for the previous design become not optimal anymore: The lamps still consumed too much electricity. By increasing the arctube load the energy consumption of a new installations could be reduced, thus reducing the TCO. Now the 15khour target for 70% lumen maintenance (the drop still not visible) is set so, you need to inspect the fixtures every two years anyway, while the 15khour is usually reached after two such cycles, so it was desired to allow the relamping to be scheduled on every 2'nd such inspection, so to not require extra traveling of the truck.
 So the lamps were tuned to yield the best average efficacy while meeting the lumen maintenance requirement, so allowing the lowest TCO. The 25khour rated lifetime was then published for the 50% lumen drop, as it correspond to lumen drop of large group of incandescents when reaching their rated life.

Do you really believe LEDs will take over and eventually become pretty much the only source of artificial illumination

This part yes, but mainly on account of the cost: The LED's are capable to reach so low manufacturing cost in not as far future, than it would become nearly impossible for other sources to compete simply on price alone. It is partly on the fact, than the only light fixture, what could be made simpler, is the incandescent lamp. But even when LED's require more components in the ballast, that could be mass produced by automated lines, while you need no means to allow users to replace anything in the fixture - so it would become way simpler. This expect than the LED reliability would become really under control.
The low cost production is already happening today - the  maker's pressure for LED's is backed up by the really high margins they have on these products, compare to the traditional ones. Once these margins would be pressed down by the force of market, we would see that in the LED fixture sale prices as well...
The reliability is the biggest issue with LED's today, but taking into account how infant the technology is, I see no reason, why it should not improve (this mean still staying at low cost)

with near 100% conversion of energy into light?

This is not possible at all. I think the LED's already reached most of the attainable efficacy, whenoperated at low currents and low temperatures, so it won't improve as much. For high power units, maybe to 100..150lm/W, depend on CRI: Lower CRI allow higher efficacy and vice versa, generally the energy efficiency would not go above 50%, while today they are at nearly 40%... The main problem is really the reliability mainly on the fixture manufacturer's side and for time being the cost on all sides. My guess, the cost would improve first, the reliability would somewhat follow in a pace to follow the knowledge base while maintaining the low cost...

A year and a half later, and GE is making halogen bulbs in the U.S.A. Today I heard an ad on the radio (three times) for GE soft white bulbs at Walmart, and they mention "Made in the U.S.A." in the ad. I'm happy to hear it.

Glad i have stocked up with full boxes of MV Lamps from 50watt all the way through to 400W

just got to get some 700w lamps