Just on the news tonight was an article about the city's recent street light conversion from HPS to LED. 

LED Street Lights

Though it remains to be seen how well they hold up, the extreme climate may help prolong their life, and actually improves their efficiency.  According to the power company, the city cut it's light bill by 60%.

Just on the news tonight was an article about the city's recent street light conversion from HPS to LED.  

LED Street Lights

Though it remains to be seen how well they hold up, the extreme climate may help prolong their life, and actually improves their efficiency.  According to the power company, the city cut it's light bill by 60%.

The big key is that, they didn't have to pay for it. It's like getting a more fuel efficient car without having to pay for it.

"The FMATS used federal and state money for the switch so all the savings is kept with the city"

If you were to actually pay for the change out, LED change out would not come out ahead compared to best available discharge technology.

They also don't address if the new set up provides nearly the same output.  If you cut the output by 40% and save 35% on energy, you're not making it more efficient.  If it's cutting 60%, they're probably using "fudge factor" like "scotopic lumen", just as Lights of America do on their fluorescent outdoor lighting product. 

The same fudge factor can be applied to higher color temp discharge lamp too, but that doesn't make LEDs look favorable. 

"Instead of bright orange, nighttime streets will start to look a little more blue. The orange glow we usually see comes from the vintage style high-pressure sodium bulbs, the brighter and bluer lights you can see in certain neighborhoods are Light-emitting diodes or LEDs."

I never thought a sentence could get me so irritated

I think Luminaire hit on a good point.  I don't think that they are as bright as the old HPS lights, and I think that is a source of some of the energy savings.  From what I have read and seen in my travels, the standard for residential street lighting was 70-100 watts for HPS, while here they were often 150 watts.  Larger roads were often lit with a mix of, it seems like whatever they had handy at them time, often 400 watts HPS, on roads where I think 150-250 watts would have been plenty, and the light pollution was major.  They could have saved a fare amount of energy by going through, determining the light levels needed and installing fixtures to provide those light levels, instead of throwing the largest thing up that will fit on the pole.  (OK probably not the largest.  There are no M1000s that I know of here).

I would guess the severe weather would do nothing good for the electronic.
It is indeed right, then too hot operation shorten the life, but too cold cause other effects, that may be even more destructive, if not treated well, e.g.:
- ESR of electrolytic capacitors rise, so these capacitors would allow higher voltage spikes on them
- Mechanical stresses build up inside many components containing plastic materials, as these harden or are assembled or on room temperature (film capacitors,...) or at high temperatures (semiconductor's plastic mold), these may lead to cracks or voids, what may fill in with moisture. And if that freeze in next cycle, the ice would expand and damage it.
These stresses are even responsible for electrical parameter shifts, mainly in integrated circuit (their design strongly rely on matching between equivalent internal components, what is way worsened by stresses irregularly distributed over the die area), what may cause some regulated parameter shift in the way dangerous for other ballast components  (e.g. PFC output voltage could increase closer to the output stage breakdown). As these effects are random and nonlinear, it is impossible to evaluate them on only few prototypes. By the way this phenomenon is the reason, why lot of semiconductors for the automotive and/or similar demanding applications are all pieces production tested at -40degC, even if this test is the most expensive one. And due to this industrial rated devices (where the lightning belong) do not test at so extreme cold, so devices weak there may pass...
- Breakdown voltage of high voltage semiconductor junctions (inside power components) has strong positive temperature coefficient, so it drop down at cold, narrowing the safety gap between operating voltage spikes and the breakdown, so increases the chance, then some spurious spike drive them into breakdown (the breakdown cause damage semiconductors even without overheating them; this effect is even worse at low temperatures).

Try a cryogenic system, where transistor gain falls to single figures, from the low hundreds at room temp. The only benefit is the really reduced noise, even if you are only able to use mica capacitors as the other types tend to become open circuit. Resistors funny enough do not care too much, provided you do not thermal shock them too much, then they just go very noisy to open circuit. Of course you only need around 5 stages of gain before you are at a high enough level to use room temperature to provide the rest of the gain, the noise contribution is a lot less important.

@SeanB: You are talking about bipolar transistors, but these are virtually not used in present controllers except few specific places (voltage reference, thermal sensor and in older technologies higher current power stages in external FET predrivers - but today MOSes are more effective here too), where is no reasonable replacement.
MOSes tend to increase their gain (higher transconductance, lower Ron) at low temperatures, what mostly yield higher DC gain of whole gain circuits on lower temperatures.

Reason for limited bipolar use is their cost: Bipolar transistors is not possible to make so small in occupied area on the silicon, what make them quite "expensive" components of the integrated circuit, so it is way cheaper to create even more complex circuit using only MOSes to go around the worse matching of MOS devices compare to bipolars (e.g. OPAMP with dynamic offset compensation, even when using ~5x more components, occupy on the silicon about 1/3 area compare to bipolar one of the same final offset).
The only place i know they can not be reasonably (so without the need for an extensive trimming and/or >20% manufacturing spread) replaced is the voltage reference circuit, but here we are talking about only two (topology-wise; otherwise one of them is made from multiple units in parallel to reduce the sensitivity on the offset of the next gain stages)
Note, then the use of "classic" zeners is virtually forbidden, as driving any PN junction into breakdown cause it's severe degradation in those "clean" processes required for the rest of the circuit.
And the reason to not use pure bipolar process (these are very cheap per mm^2, so allow way larger die for the same cost) is the need for more and more complex logic functions, but at the same time with virtually no current consumption, what is possible to implement only in CMOS...

What most analog circuits on the common chip rely on is the matching between identical structures (e.g. two MOSes in a differential pair should have the same threshold voltage, so the offset of this gain stage is then zero - their Vgs subtract from each other; This is valid only if working condition of these two is equal and that is the point) and at cold it is the package stress, combined with the irregularity of the plastic fill material, yielding to different forces "pushing" in different way on any individual component from the group, shifting parameters in different way (so in the differential pair each transistor have different Vt, so the difference yield an offset of that opamp), what in turn shift some analog parameter (e.g. offset in an voltage reference opamp cause the generated voltage to shift, so all circuit values based on this reference would shift as well - e.g. PFC output voltage).
And as the positions of the grains in the plastic material can not be determined, result is, then affected parameter shifts would show large variations between individual pieces...

I know about the matching issue, but in most high power applications like power supplies bipolar is still the cheapest for high power and high voltage, even though it has big issues with thermal runaway. Mosfets do tend to be fragile, and if you have a simple enough power driver in a separate package bipolar is about the best, as there are less issues with voltage stress across the die causing problems, most bipolar manufacturers know enough about Planar to get a good die that will survive years at high stress without too much parameter shifting, and they are cheap from having been designed for TV flyback use.

Making a good regulator is still a place where bipolar and steel can is the best, as you can get a very good reference and then either laser trim or zener zap it to get within the desired range. Buried junctions are still the best for reference sources, and have good long term stablity. I have used CA3036 arrays to make a regulator before, and made a few using 723 cans, after 20 years it still is in regular use, and still gives a very accurate output, even though it runs at close to maximum input voltage for the 723.

I think bipolars win today only in really simple self oscillating power stages (as only these are able to deliver enough base drive current).
For switching application they have rather limited SOA (limit of the voltage vs current combination, that should not be exceeded in any moment, include switching transients; very important, but rarely documented is "reverse bias SOA" - aplicable to switch OFF transients; you have to use snubbers, oversized transistor for low power applications and/or use them only in soft switching mode; the limit is imposed by tendency of local thermal runaways on the chip), what i would consider as one huge weakness - it or limit the maximum usable voltage (going above 400V for 0.3A current is quite difficult).

MOSFET's do not have this issue at all, so they do not need any extra component (except spike clamp in case some stray inductance is present in the switched path).
The only drawback today is their larger silicon area for the same ON state drop vs operating current and blocking voltage rating; however as described above where the 500V MOSFET is way enough (e.g. 230V halfbridge based designs without PFC), you need 750..1000Vcbo rated bipolar due to SOA limitations for the same reliability...

Moreover in hard switched applications ("classical" switching regulators) the slowness of bipolars cause more switching losses then higher conductive (due to higher ON drop) losses of an MOSFET of similar cost.

In anything designed in past 10years (and now i mean even cheepee production) i have not seen bipolars in power stages except for mobile phone chargers (only ~3W flyback; but even here they disappear in favor of integrated switchers like NCP1050 or similar) and selfoscillating fluorescent ballasts (strictly soft switching; here they seems to stay for quite some time, as due to soft switching they are not SOA limited, so do not have to be as high voltage rated)

A real test would be in Barrow, Alaska in winter !


Powell