can an energy saving bulb be run off DC ?

[Simon]

As above. Could i run a standard energy saving bulb on 240V DC ? I'm guessing they use switch mode voltage converters in them so must be DC/DC so they must rectify the mains at the input ?

[mikeselectricstuff]

Probably, but remember that 240VAC rectified does not make 240VDC - it's around 330V.
YMMV with different makes, but all the ones I've seen rectify the mains and so should work on DC.
Note however that any internal fuse may suffer arcing problems at DC in the case of an internal fault.

[Zero999]

Yes, a CFL will work off DC.

I'm not sure if it'll work at full brightness on 240VDC. After the rectifier, there's a smoothing capacitor with just below the peak mains voltage on so you may need 325V to get full brightness.


Plenty of information can be found on CFL circuits in the Internet:
http://www.nxp.com/documents/application_note/AN00048.pdf
http://www.pavouk.org/hw/lamp/en_index.html
http://www.google.co.uk/search?client=opera&rls=en&q=CFL+circuit&sourceid=opera&ie=utf-8&oe=utf-8&channel=suggest

[Ed.Kloonk]

 

[johnwa]

As others have said, it is possible to run CFLs from DC. However, be aware that there are a few different designs of the driver circuits for them. The older ones used a bridge rectifer straight into a filter cap on the input. These will work fine on 340VDC. However, the power factor of this design is pretty terrible, so in recent years a new design has been showing up. These isolate the filter cap from the rectifer with another diode, and the tube is run from the unfiltered DC - the filter cap only supplies the control circuit. Since the tube supply in this design is only 240VRMS, the tube will be overdriven  if run from 340VDC.

I once built an inverter for driving these that supplied 340VDC, pulse width modulated to provide an appropriate RMS value. It may be possible to find a suitable DC value between 240V and 340V that provides acceptable operation. (The driver circuit is underdriven at 240VDC, and the transistors may not saturate). It would probably be best to measure the power consumption, and get it to equal the value when the bulb is driven from 240VAC.

[Zero999]

As others have said, it is possible to run CFLs from DC. However, be aware that there are a few different designs of the driver circuits for them. The older ones used a bridge rectifer straight into a filter cap on the input. These will work fine on 340VDC. However, the power factor of this design is pretty terrible, so in recent years a new design has been showing up.

I've never seen a power factor corrected CFL before. There are no requirements for it in the EU as a CFL draws under 100VA. I don't know about other jurisdictions.

These isolate the filter cap from the rectifer with another diode, and the tube is run from the unfiltered DC - the filter cap only supplies the control circuit. Since the tube supply in this design is only 240VRMS, the tube will be overdriven  if run from 340VDC.

I don't see how powering a power factor corrected CFL from 340VDC would cause any damage. It might not like it but I doubt it would go bang.

An active power factor correction unit is typically a boost converter before the DC reservoir capacitor which outputs an equal or slightly higher DC voltage than the mains peak voltage. In the more complex designs, the PWM of the switching transistor is sometimes varied to ensure the current drawn is perfectly sinusoidal. The simplest active power correction units don't vary the PWM, it's just a boost converter with a very long time constant in the feedback network so the PWM remains roughly constant (longer than a couple of AC cycles) but the current varies in phase with the AC and is not perfectly sinusoidal but better than the classic rectifier+filter capacitor design.

My guess is that if a CFL had PFC, it would use the latter of the techniques described above so powering directly from DC shouldn't be a problem.

[johnwa]

I've never seen a power factor corrected CFL before. There are no requirements for it in the EU as a CFL draws under 100VA. I don't know about other jurisdictions.

OK, perhaps these are not as common as I thought. There are a few CFL schematics at http://www.pavouk.org/hw/lamp/en_index.html, but the only one with the PFC circuit is actually an electronic ballast for a linear tube.

I don't see how powering a power factor corrected CFL from 340VDC would cause any damage. It might not like it but I doubt it would go bang.

An active power factor correction unit is typically a boost converter before the DC reservoir capacitor which outputs an equal or slightly higher DC voltage than the mains peak voltage. In the more complex designs, the PWM of the switching transistor is sometimes varied to ensure the current drawn is perfectly sinusoidal. The simplest active power correction units don't vary the PWM, it's just a boost converter with a very long time constant in the feedback network so the PWM remains roughly constant (longer than a couple of AC cycles) but the current varies in phase with the AC and is not perfectly sinusoidal but better than the classic rectifier+filter capacitor design.

My guess is that if a CFL had PFC, it would use the latter of the techniques described above so powering directly from DC shouldn't be a problem.

The circuit is not very sophisticated, it just draws some of the operating power from the unfiltered source, so that the load is spread over the full mains cycle, rather than being in one lump at the voltage peak. I think the problem is that this type of circuit responds more to the RMS value of the input than the peak, so substituting an input waveform with a different RMS to peak ratio (1 for DC, 1/sqrt(2) for AC sine) could be an issue.

But it looks like most CFLs still use the standard rectifier + filter circuit, so they should be fine on 340VDC. It might still be a good idea to check that the current consumption is not significantly above that calculated from their power rating, to ensure long life.

It is also possible to run other 240VAC input SMPSs from 340VDC - this can provide benefits in efficiency and reduced complexity if you are building an inverter. The only possible problems would be the breaking capacity of the fuses and switches.

[scrat]

Why not trying to measure the voltage on the CFL when supplied with the mains, and then starting with a low DC value, and increasing unless it reaches the same output?

[Leo Bodnar]

Move on to LED lamps.  They have better power efficiency and smoother spectrum.
They are now cheap enough to beat CFLs unless you buy them from B&Q.

[Zero999]

The circuit is not very sophisticated, it just draws some of the operating power from the unfiltered source, so that the load is spread over the full mains cycle, rather than being in one lump at the voltage peak. I think the problem is that this type of circuit responds more to the RMS value of the input than the peak, so substituting an input waveform with a different RMS to peak ratio (1 for DC, 1/sqrt(2) for AC sine) could be an issue.

So you mean they just omit the filter capacitor? That would make the power factor better at the expense of flicker so it looks like a lamp powered from a traditional iron ballast.

Move on to LED lamps.  They have better power efficiency

I doubt that, in fact most LED lamps have poorer efficiency than a CFl.
http://en.wikipedia.org/wiki/Luminous_efficiency#Examples_2

LEDs tend to be better for direct light sources though such as replacing a halogen.

[NiHaoMike]

Modern LEDs are easily more efficient than fluorescent. That's one of the reasons why newer LCD backlights are LED. In addition, LEDs are physically tougher than fluorescents, can be dimmed practically to zero, and can be switched frequently without any impact on lifetime.

[Zero999]

Modern LEDs are easily more efficient than fluorescent. That's one of the reasons why newer LCD backlights are LED.

Old LCDs used cold cathode fluorescent lamps which were worse than compact fluorescents.

A naked Cree LED is better than a CFL but it's not cheap and the efficiency doesn't account for the losses in the power supply or the luminaire.

The best LED light listed on Wikipedia is    5.4 W LED screw base lamp (100 V 50/60Hz)   14.9%

The best fluorescent listed on Wikipedia is a T5 tube powerd by an electronic ballast   15.63%

Of course I know this is Wikipedia here but it does show that generally LEDs with integrated ballasts and packaging are at best around equal to CFLs but are often worse.

[Leo Bodnar]

Have a look at CFL spectrum. It is horrible.

[Zero999]

A white LED's is hardly perfect either, a large blue peak makes blue brighter than other colours.

[Leo Bodnar]

A white LED's is hardly perfect either, a large blue peak makes blue brighter than other colours.

This must be old cold white LED.  Things have moved on since then.

Yes, LEDs do not have emission spectrum of perfect black body (sic), D65 source or even the Sun but at least they do not have a few strong emission lines with huge areas in between with no light at all like CFLs. If two coloured objects happen to have colour difference in empty areas between CFL spikes their colours will look exactly the same to the eye (or any other receiver for that matter.)
Try sorting your black socks under CFL lamp - they will most probably will be wrongly paired in the morning.
This is called metamerism.

Same problem exists for "white" light generated by set of RGB LED triplets - they have three narrow spectral emission lines and nothing in between.
It looks white when you shine it on a white or grey object but as soon as you start looking at objects with complex reflection spectra the things start to look "weird."

Here are few spectra I have taken around the house: a new T8 CFL tube, cheapo Chinese GU10 LED lamps and Philips LED bulb.

[Simon]

thanks for all the replies guys, for some reason I got no notification (and not the first time - I think thee are issues, will have to have a word with dave)

Basically the source of the question is the fact that I have my nice solar panels and plans to add more, but with my recent GTI experiment I'm not sure i'm going down that road again.

So I thought, why not just charge my battery(s) and then when i get home use the power to run the lights (I have separate circuits for upstairs and downstairs so can minimize the load initially). I could build some sort of inverter that produces mains voltage DC and screw all those crap inverters on the market. Yes LED does sound attractive but that means that I'd have to use a SMPS to drop the mains voltage when I'm out of battery so not sure about efficiency.

Yes CFL spectrum's are crap, i started a thread about this a while ago as I've tried a number of bulbs ranging from 2700K (the normal shit in the shop that is just yellow) right up to 5500K which ends up looking too blue. With the narrow spectrum a good compromise despite daylight being 5500K is 3500K. What also makes 5500K look bad is having other 2700K's nearby and your used to those so you get duped. (the brain actually does auto white balance correcting, once upon a time in photography you had to use a different film for each light source when you could not visually tell the difference)

[Zero999]

Here are few spectra I have taken around the house: a new T8 CFL tube, cheapo Chinese GU10 LED lamps and Philips LED bulb.

But aren't warm white LEDs less efficient since more phosphor is applied to the die there will be more loss in converting the blue to longer wavelengths.

The spectrum of CFLs varies widely, some are better than others, same for LEDs.
http://ioannis.virtualcomposer2000.com/spectroscope/toyspectroscope.html

Buy and expensive, good quality LED light and it'll be more efficient than any CFL, get a cheap Chinese LED light and it'll be much worse, I've seen some of the crappy lights in my local store.

[Leo Bodnar]

But aren't warm white LEDs less efficient since more phosphor is applied to the die there will be more loss in converting the blue to longer wavelengths.
The spectrum of CFLs varies widely, some are better than others, same for LEDs.

There is efficiency and there is efficiency.  There is also efficacy which is efficiency projected onto human eye sensitivity.
You can have very efficient monochromatic light (like laser) but who would want to use it as home lighting?

We probably need to agree what we are discussing here. I think "efficiency" should include nice (preferably continuous) spectrum, brightness per volume, price, lifespan and dangerous materials (e.g. CFLs contain mercury.)

Sure, there are some expensive CFLs for use in colour matching in printing industry and for photo studio lighting.
I just measured what I have around an average home (stuff you buy from eBay and B&Q.)

[Zero999]

CFLs contain virtually no mercury. The greatest hazards are the glass tube and the phosphor which is often poisonous and lead solder.

LEDs have the advantage of being not being made of glass.

My point I suppose is don't just go on the press releases by Cree on efficiency, look at the real life lamps.

[Leo Bodnar]

My point I suppose is don't just go on the press releases by Cree on efficiency, look at the real life lamps.

So how one to measure efficiency of "real life lamps?"

CFLs contain virtually no mercury.

How do you know?  I hope you did not pick up this knowledge from wikipedo. What element creates the spectrum peak at 546nm then?

[Zero999]

My point I suppose is don't just go on the press releases by Cree on efficiency, look at the real life lamps.

So how one to measure efficiency of "real life lamps?"

As you probably know it's subjective.

I think we both agree it's a balance between spectrum, price and brightness vs power consumption. I just think that LEDs aren't quite there yet, unless you're willing to buy good Cree LEDs. In my experience, CFLs are always dimmer than the incandescent they're supposed to be equivalent to and LEDs seem to be even worse.

How do you know?  I hope you did not pick up this knowledge from wikipedo. What element creates the spectrum peak at 546nm then?

Multiple sources reveal the amount of mercury in a CFL is something like 5mg which is almost nothing unless you keep breaking them, the phosphor and broken glass are much more dangerous.
http://www.treehugger.com/files/2007/05/ask_treehugger_14.php

[Simon]

I think we both agree it's a balance between spectrum, price and brightness vs power consumption. I just think that LEDs aren't quite there yet, unless you're willing to buy good Cree LEDs. In my experience, CFLs are always dimmer than the incandescent they're supposed to be equivalent to and LEDs seem to be even worse.

The reason why CFL's are not as good as it says "on the tin" is because of what i keep telling people who mostly don't listen or understand: colour temperature. They emit yellow light, that is not the same as the same power of white light. But then if your using a meter sensitive to yellow light you can claim certain power ratings. As i keep saying 2700K CFL's are a con and should be banned !

Multiple sources reveal the amount of mercury in a CFL is something like 5mg which is almost nothing unless you keep breaking them, the phosphor and broken glass are much more dangerous.
http://www.treehugger.com/files/2007/05/ask_treehugger_14.php

5 mg of mercury is 5 mg of mercury. Mercury is very hazardeous. From memory 1g of mercury pollutes thousands (10'000 ?) of litres of water

[Leo Bodnar]

The reason why CFL's are not as good as it says "on the tin" is because of what i keep telling people who mostly don't listen or understand: colour temperature.

The term "colour temperature" cannot be applied to highly spiky spectrum like that of CFLs or e.g. an RGB LED. It just makes no sense.

[Simon]

you can define the colour temperature of a CFL. I'm sure if you were to measure the colour it would come back with a colour. Of course that might not be the colour of the main colour spike but will be what we see it as based on all the colours emitted. If it looks yellow (2700K) that's what it is.

[Zero999]

Simon, I don't know if I agree. I recall the horrible high colour temperature blue CFL being dimmer than the warm white lamp round your house, despite being the same power rating.

Come to think of it, I've seen some good LED lamps at the swimming pool at Calvert but they probably weren't cheap and I don't know what the power rating was. All the changing rooms were lighted by LEDs, it was just the poolside which had flat panel fluorescent lamps.

I think, if you want to go for LEDs then fork out for some Cree LEDs and run the whole house lighting off DC. Use a decent SMPS for when there isn't enough energy in the batteries. Li-ion batteries are more efficient than lead acid, but they're expensive, have a limited life, need careful charging and should be kept in an outbuilding a safe distance away from your house.

[Leo Bodnar]

LOL.  We are speaking in different terms.

What is the colour temperature of purple or green?  It just does not exist.
Look at the below diagram.  The curvy line in the middle is all colour temperatures from -273C to infinitely hot.
Any colour outside this line does not have a colour temperature.

[Simon]

Simon, I don't know if I agree. I recall the horrible high colour temperature blue CFL being dimmer than the warm white lamp round your house, despite being the same power rating.

Nope, have a good look at that graph above, 5500K is quite near blue. Warm white is the con-mans version of yellow - again look at the graph. the first CFL's that were produced were certainly not 2700 K or they would have never got out. Now the horrible yellowness of them is so apparant that people think the orange street lights are CFL's. Modern CFL's are not liked for a very good reason but people are stuck with them. Ask anyone over the age of 40 what they think of these new fangled bulbs....

[alm]

There is such a thing as correlated color temperature, which is something like the color temperature of a black body radiator that is the closest to the spectrum of the light source. As the spectrum gets farther from the ideal black body radiator spectrum, this figure tends to become less meaningful.

Incandescent lights for domestic use have a color temperature of something like 2700-2900K, so I don't understand Simon's issue with 2700K CFLs (unless they're actually lower than 2700K). Anything between say 2700K and 10000K can be perceived as white by the human eye, depending on the circumstances. Whether it looks yellow or blue mainly depends on the colors of the objects around you and whether other light sources are present. Early fluorescent lights used to have a fairly high color temperature, they actually lowered it (and reduced the efficiency somewhat) for home use.

[Zero999]

Nope, have a good look at that graph above, 5500K is quite near blue. Warm white is the con-mans version of yellow - again look at the graph. the first CFL's that were produced were certainly not 2700 K or they would have never got out. Now the horrible yellowness of them is so apparant that people think the orange street lights are CFL's. Modern CFL's are not liked for a very good reason but people are stuck with them. Ask anyone over the age of 40 what they think of these new fangled bulbs....

What do you mean?

Colour is subjective. To me the high colour temperature lamp round your house is a horrible dim blue/greyish lunar colour and the warm white is a more pleasant warm and creamy colour. I don't know how you can claim the 2700k lamp is less efficient because it seemed much brighter and used less power than the horrible blue light but whether you preferred it or not is your opinion. A sodium street light or green LED would probably be more efficient than either CFL but would look horrible.

In my opinion halogen lights give the best light quality, much better than normal incandescents, LEDs or CFLs.

[Simon]

The truth is that both those lights in my (or rather my dads now) house are not nice. It seems that CFL's have a narrow spectrum and it centres very closely around the main colour temperature. Because you are used to the 2700K one and also because you saw them in contrast you really noticed the defects of the 5500K. If the 2700 had not been there (illuminating a smaller space I might add) you would have not seen the stark contrast and you brain would have adapted more easily to 5500K.

The 2700K bulbs make me drowsy in the winter and i need way more "wattage" than claimed to be satisfied. I have bought 3500K compact strip lights (which I assume are also CFL's) and they were much much better, they looked white and the illuminated area was white.

The human eye will adapt within reason to different colour temperature lighting, it will autmoatically make the whitest object white and balance the rest accordingly. but 2700K is still so bad that the human eye can't quite do it. Most people don't notice or think about it as they are used to it or accept it, unfortunately i do notice and don't like it.

[Zero999]

We're all different. I can't stand 5500K because it burns my retina, is not a very flattering light to be seen in,and if it's too dim it appears grey. I have no problem with colour differentiation under 2700k as long as it's bright enough, if it's too dim colour perception is harder but at least it's not gloomy like higher colour temperature light.

[Simon]

yes each person will perceive light colour a little differently. I'm no fan of the 5500K light. it almost makes a natural light in my bedroom that is painted green (so does have some use when paired with the decor). I can manage with 2700K lighting but I need 2-3 times the equivalent power I would need from an incandescent. so that kind of almost makes CFL's not such an energy efficient option.

When i moved into my house i found 3 and 5 bulb chandeliers in the downstairs rooms. These are quite aptly fitted. It requires that many bulbs to correctly illuminate the rooms, that is rooms that according to the bulb manufacturer can be lit with one bulb. So in effect I am using up to 100W of CFL where I would otherwise use 100-150W of incandescent. Some energy saving

[Leo Bodnar]

It seems that CFL's have a narrow spectrum and it centres very closely around the main colour temperature.

I officially give up!    

By the way mid-day daylight is around 5500K.

[tsmz]

Usually, CFLs (or generally, all lights) will have a CRI (Colour Rendition Index) printed somewhere. You will absolutely need to get the highest CRI available if you're interested in good colour reproduction. You probably won't find a CRI of 90 or more in CFLs, but at least T8 fluorescent tubes with good CRI are readily available and don't cost an arm and a leg (that is, they're about twice to thrice the price of standard T8 tubes, but they're still reasonably inexpensive). The situation is a bit worse with T5 tubes, but improving.
I changed my lighting to a couple of 965 (i.e. 6500 K, CRI > 90) tubes and really can't complain. Of course, you'll have to like blueish-white sunlight or else it'll drive you nuts, but the real trick with high colour temperature is decent light intensity. Dim daylight just doesn't work and feels really nasty.
I've got an odd 865 T5 tube somewhere because I couldn't find an available 965 type, and the difference in colours is stunning. Some colours really do look slightly off, but it should nonetheless work quite well in areas where you just need light and not really good colour rendition.

The problem I've found with commercial LED lamps is that they're crap. Almost all of them have horrible cooling design, which will make them overheat and dim. I've seen quite a few supposedly long-lasting LED light strips with partially dimmer LEDs, so I don't quite think we're there yet when it comes to LED lighting. Sure, you can get some really nice high-power LEDs, but cooling the stuff is important as well and that's where the problems begin.

[Simon]

well really I suppose colour temperature is a bad definition of a light source to start with. I mean like you say the perfect source emits a flat range of many colours. basically putting a colour temperature on a light source (and so giving it a colour) is removing that definition and is calling it a coloured light, or a varied coloured source that outputs mostly the declared colour (temperature) 5500K (blue) on it's own is useless as it 2700K (yellow), what we need is an average of 5500K that comes from a span ranging from 2700K to 9000K, we see colour as reflected light, white because it reflects all colours and each colour because it reflects that colour in the light. If the colour is missing in the light you won't be seeing it so well if at all

[Simon]

Of course, you'll have to like blueish-white sunlight or else it'll drive you nuts, but the real trick with high colour temperature is decent light intensity. Dim daylight just doesn't work and feels really nasty.

exactly. Same goes for so called warm white (yellow) as I said above, i need 5X the power to get a correct lighting, so goodby energy saving

[Zero999]

exactly. Same goes for so called warm white (yellow) as I said above, i need 5X the power to get a correct lighting, so goodby energy saving

Try halogen downlighers and go for 12V which is higher efficiency and lasts longer than 230V lamps because the filaments are thicker and can burn hotter.

I changed my lighting to a couple of 965 (i.e. 6500 K, CRI > 90) tubes and really can't complain. Of course, you'll have to like blueish-white sunlight or else it'll drive you nuts, but the real trick with high colour temperature is decent light intensity. Dim daylight just doesn't work and feels really nasty.

Sunlight is 5,800K but that's only at meridian in the tropics, it'll be a slightly lower temperature than that most of the time. Daylight averages at 6500K because of the sky being blue due to the refracted light.

The problem I've found with commercial LED lamps is that they're crap. Almost all of them have horrible cooling design, which will make them overheat and dim. I've seen quite a few supposedly long-lasting LED light strips with partially dimmer LEDs, so I don't quite think we're there yet when it comes to LED lighting. Sure, you can get some really nice high-power LEDs, but cooling the stuff is important as well and that's where the problems begin.

Yes I agree. I'd be interested to know what sorts of LEDs they used at the swimming pool I was talking about previously. The problem is domestic consumers will often choose products based on price. It took a few years for CFLs to have decent efficiency and life at life at a reasonable price and it'll take awhile for the same to happen with LEDs but when it does happen, they'll be better than CFLs.

[tsmz]

well really I suppose colour temperature is a bad definition of a light source to start with. I mean like you say the perfect source emits a flat range of many colours. basically putting a colour temperature on a light source (and so giving it a colour) is removing that definition and is calling it a coloured light, or a varied coloured source that outputs mostly the declared colour (temperature) 5500K (blue) on it's own is useless as it 2700K (yellow), what we need is an average of 5500K that comes from a span ranging from 2700K to 9000K, we see colour as reflected light, white because it reflects all colours and each colour because it reflects that colour in the light. If the colour is missing in the light you won't be seeing it so well if at all

You seem to be having a slight problem in understanding colour temperature.
There is no such thing as "white" light. How do you define white light? Same amount of radiated power in every frequency? The human eye has a high sensitivity in the greenish-yellow part of the spectrum, so this "white" light would probably look quite greenish or yellowish.
That's why we're using the definition of blackbody radiation. An ideal black body will emit light in a certain distribution when it is heated to a certain temperature. The sun is approximately a black body at about 5000 to 6000 K surface temperature. If we have established a function between temperature of a blackbody and the distribution of its emitted wavelengths, we can now define a colour temperature! The colour temperature is simply defined as the temperature a blackbody would have to be heated to to create the same (or the closest matching) distribution of wavelengths (in the visible spectrum) as the light source you're measuring.

But there are hardly any ideal blackbody radiators out there. The atmosphere will filter out some of the longer wavelengths of sunlight, making it appear more blueish, so "apparent" sunlight is usually defined somewhere around 6000 to 6700 K. Normal incandescent bulbs are mostly blackbody radiators run at quite a low temperature, wasting most of the power as infrared radiation, and that's why they're so terribly inefficient.
Of course, color temperature can be defined even if you've just got a few peak wavelengths in your spectrum, but this will get increasingly silly the more you deviate from a continuous spectrum. That's why there's the CRI, which exists in countless variants and has the strangest definitions, but the idea behind it is to give a measure of the blackbody-like quality of your light source.

[Leo Bodnar]

How do you define white light? Same amount of radiated power in every frequency? The human eye has a high sensitivity in the greenish-yellow part of the spectrum, so this "white" light would probably look quite greenish or yellowish.

Eye is only sensitive to green/yellows because there are more of them around then other wavelengths. I reckon your flat spectrum "white" light will look pink (excessive amount of red + blue.)

[Zero999]

I think for comfortable viewing, the colour temperature should correlate with the intensity. At sunrise and at sunset when the light intensity is low, the colour temperature of sunlight is also low. Light with a higher colour temperature at this intensity will look horrible. The only time you want dim high colour temperature light is when it's very dark (moonlight) when the the eyes switch to scotopic vision which is why LED torches are so good as the rods work better at shorter wavelengths.

[Leo Bodnar]

The only time you want dim high colour temperature light is when it's very dark (moonlight)

Moonlight should have spectrum identical to the direct sunlight (around 5500K.)

when the the eyes switch to scotopic vision which is why LED torches are so good as the rods work better at shorter wavelengths.

Where do you get all this from? Rods sensitivity has a peak bang in the middle of green parts of visible spectrum.  I should say that this is purely my opinion.

[Simon]

yes and it will be no where near as bright as your localized artificial light

[Zero999]

The only time you want dim high colour temperature light is when it's very dark (moonlight)

Moonlight should have spectrum identical to the direct sunlight (around 5500K.)

Yes, i agree.

when the the eyes switch to scotopic vision which is why LED torches are so good as the rods work better at shorter wavelengths.

Where do you get all this from? Rods sensitivity has a peak bang in the middle of green parts of visible spectrum.  I should say that this is purely my opinion.

From Wikipedia, other sources are available.

Human vision has two modes:
Photopoic; this occurs in bright conditions, when the cones are dominant and has a peak sensitivity of 555nm.

Scotpoic: this occurs under low light conditions, when the rods dominate and has a peak sensitivity of 498nm. The eye's sensitivity to shorter wavelengths increases slightly but the sensitivity to the red end of the spectrum diminishes.


http://en.wikipedia.org/wiki/Photopic_vision


http://en.wikipedia.org/wiki/Scotopic

[Leo Bodnar]

Scotpoic: this occurs under low light conditions, when the rods dominate and has a peak sensitivity of 498nm. The eye's sensitivity to shorter wavelengths increases slightly but the sensitivity to the red end of the spectrum diminishes.

Ah, OK, not that much difference then.  It looks like night vision is not affected by deep red light part of the spectrum beyond 600nm.
This is probably why red light is used for temporary night light sources e.g during astronomical observations.  
You don't want somebody flash a white light onto their map and blew away your night vision that took few hours to adopt to.

[johnwa]

So you mean they just omit the filter capacitor? That would make the power factor better at the expense of flicker so it looks like a lamp powered from a traditional iron ballast.

Sort of. The half bridge driver runs from filtered DC, but the other end of the lamp is coupled back to the unfiltered supply. Have a look at the circuit 'SINECAN 5 2x 26-30W' on the page I mentioned. I guess it works by running the positive half-cycle of the HF waveform from the filter caps, and the negative half-cycle direct from the rectified mains. I have attached another circuit of a CFL that I traced out a while ago - it uses the same design.

[Zero999]

It looks like a strange thing to do.

How will it improve the power factor?

I don't see how it can. The fluorescent tube is connected to the unfiltered supply via a 47nF capacitor which makes no difference at DC and has an impedance of 68k at 50Hz so only allows the higher frequency tube current through.

The tube is powered in cold cathode mode which makes starting harder and is not good for tube life.

[johnwa]

It looks like a strange thing to do.

How will it improve the power factor?

I don't see how it can. The fluorescent tube is connected to the unfiltered supply via a 47nF capacitor which makes no difference at DC and has an impedance of 68k at 50Hz so only allows the higher frequency tube current through.

I sort of thought, running it from completely unfiltered DC will improve the power factor, so running it from unfiltered DC part of the time would give some improvement too. Perhaps I have misunderstood the circuit, although I can't see why else they would connect it like that.

The tube is powered in cold cathode mode which makes starting harder and is not good for tube life.

Yes, that is how I drew it, I think there were only two wires going from the PCB to the tube, although there might have been some other components off the board. I really can't remember, it was quite a while ago I traced it out.