150W offline LED driver cannot possibly be 95% efficient?

[treez]

Hello,
A certain major streetlighting company is purporting to have Offline, 220VAC, 150W, dimmable LED streetlights which have efficiency of 95%.
This streetlight is dimmable from 5W to 150W.
Do you agree that this is not possible with anything other than ridiculously expensive circuitry?
I mean, I once made a 150W LED driver which was 95% efficient….but this was from an input voltage of 48VDC , and it consisted of three separate 50W Buck channels feeding into the same LED load. With a 48V input, one’s switching losses are far lower than switching off a 400V post PFC bus. Also, with 48VDC input, one has available some FETs with very low Rdson and low junction capacitances, so as to keep the losses down. Also, the 95% efficient version had a severly undamped drain voltage transistion (to reduce switching losses) and was severely noisy. (when the series gate resistors were reduced from 4R7 to 2R7 , then switching this LED driver on resulted in the lab digital radio going totally quiet when it had just been playing at full volume!)
To get 95% efficiency in an offline (mains connected) situation, you are going through a PFC boost converter, and then a downstream LED driver. It would in fact be pretty miraculous to get an offline PFC boost converter to operate with efficiency of 95%, let alone the cascade of PFC Boost/LED driver.
To get 92% plus efficiency from a PFC/LED driver cacade combo, then the LED driver would have to be a resonant converter. This would likely mean an LLC converter. However, the LLC converter would definitely not be able to garner 90% plus efficiency in the light load situation of dimmed down operation. This is because the switching frequency of the resonant converter would likely go high at light load, and this would mean more switching loss, since when well above the upper resonant frequency, then the LLC converter gives significantly more switching losses as a proportion of total losses. In order to get zero voltage switching at higher frequencies than f(upper) in an LLC converter, one would have to ensure high enough magnetising current to discharge the Cds’s at the lower switching period…this means more dead time, or decreasing the magnetising inductance sufficiently………but if the dead time was set that high, then this would lead to an efficiency penalty when at the nominal maximum load of 150W (since that’s where the converter is designed to operate at the f(upper) switching frequency…also, a reduction of magnetising inductance means more conduction losses.
Not to forget the natural variation in LED Vf, with both tolerance and temperature…this makes for more difficulties for the LLC which mean mitigation and subsequent efficiency penalties for the LLC converter.
Therefore, I refute the claim of 95% efficiency, do you agree?

 

[mtwieg]

The efficiency given is usually the peak efficiency. At light load the efficiency will be much less.

95% isn't unbelievable, but they are probably paying for it in some way.

If the driver is meant to be in the same housing as the LEDs, then such high efficiency basically doesn't even matter, since overall temperature rise is going to be dominated by load dissipation, not converter dissipation.

 

[treez]

If the driver is meant to be in the same housing as the LEDs, then such high efficiency basically doesn't even matter, since overall temperature rise is going to be dominated by load dissipation, not converter dissipation.

yes i agree with that, but with respect, this was not really my point.

Customers are now into wanting absolute maximum efficiency, so as to reduce electricity bills, so companies quoting 95% are going to get favoritism.
The customers now are massively motiviated by even a 1% increase in efficiency for streetlights.

 

[mtwieg]

Sounds like a marketing problem more than an engineering problem.

A 1% efficiency difference for a 150W streetlight is maybe 1gbp in electricity per year. Over the lifespan of the device, probably much less than the capital cost of obtaining that higher efficiency. But if they're beating your prices with that efficiency spec, then you've got a problem...

 

[dick_freebird]

Customers want all kinds of things until the small
matter of the check arrives. About 1% of them could
tell you how much 1% additional efficiency will save
them over the purported life* of the bulb, or make
any similar sound value decision that doesn't involve
simply believing what the package says.

So packages say all kinds of stuff.

* and you bet your a$$ there's an asterisk in there
somewhere.

 

[betwixt]

Looks like an opportunity for your sales department to score points over the competition by explaining how efficiency and overall product costs are related. If you can demonstrate the payback time for a small efficiency increase is longer than the product life, you not only show your technical understanding but debunk claims by competitors.

Don't forget to include the 'green' aspects of it, if there are any.

Brian.

 

[treez]

Yes, its interesting though, because a cheap 150W streetlight that is 5% less efficient than whichever competitor, actually only results in an extra 22kWh of electricity being used in a year......which corresponds to less than £2.
It really does stop companies from wanting to bother to do say a nice efficient resonant converter in a streetlight.

 

[KlausST]

Hi,

I agree with the "cost" aspect.

Another point of view is the power dissipation.
If you increase the efficiency from 90% to 95%..this means 50% of dissipated power as heat...and this means about half of the temperature rise..
If your ambient temperature is 25°C and the 90% temperature is 85°C then the 95% temperature is about 55°C.
This will
* increase lifetime
* can reduce heatsink size
* increase the usability in hotter ambient temperature
* increase the usability in clean-rooms, rooms with reduced air pressure or other extreme situations.

You have to choose if you aim for low cost, high volume, or special applications..

Klaus

 

[treez]

If you increase the efficiency from 90% to 95%..this means 50% of dissipated power as heat...and this means about half of the temperature rise..
If your ambient temperature is 25°C and the 90% temperature is 85°C then the 95% temperature is about 55°C.

Thanks, i agree, from the point of view of the driver smps in isolation, it is correct, however, as you know, in a streetlight, the LEDs actually dissipate far more as heat than the driver, so in fact, the heat rise due to a driver that is 5% less efficient than some other driver, is much less significant.

In both cases, the enclosure's innards, will primarly be heated by the LEDs...and any extra heating by a "5% less efficient" driver SMPS will be insignificant.

Another point, is that with a driver SMPS, it is only the electrolytic capacitors that have their lifetime reduced by heat. Any MOSFET, ceramic capacitor, or chip resistor, ..as long as its temperature remains below 125 degC, ..will last indefinitely.

Also, we have our streetlight SMPS’s in a “side chamber” of the ‘head’ of the streetlight…but this has to be made of metal for support….and inside this side chamber, on a hot sunny day, even with the SMPS not running, it gets extremely hot in there, as it acts like a furnace, getting stoked up by the sun…this kills off the electrolytic capacitors. Even at night, the "side chamber's walls are connected to the metal of the LED heatsink, and so that "side chamber" really does get frying hot...and our SMPS sits in there and is also enclosed in its own plastic case, which means the SMPS components really fry...and the electrolytics don't last as long as we would like.

 

[Kajunbee]

The exterior color of the chamber as you know plays a large part in how hot the chamber gets heated from the sun. I got a good example of this one time when changing a metal roof. The old metal was galvalume which is light grey in color and the new metal was white. By midday you couldn't touch the grey metal without blistering yourself. The white seemed relatively cool to the touch all day. I would think black would be worse yet. I don't know how bare aluminum is effected Since it is also has a grayish color to it.

- - - Updated - - -

Thinking back on it I wonder if the metallic nature of the galvalume coating is why that roof was so extremely hot. I can't see where the grey color alone should get that hot.

 

[KlausST]

Hi,

I can't see where the grey color alone should get that hot

Indeed it mainly depends on the absorption of the surface in the IR range.

While with usual colors one can expect that darker colors absorb more IR than light color...
But it's about impossible to compare different surface materials just by their color.

A metal or metaloxide surface may have different behaviour than plasic color.

Example: pure silicon looks like a relatively dark metal, but it's about transparent for IR.

Klaus

 

[betwixt]

Slightly off topic but it proves a point: A few days ago the air temperature here was -5C and there was 10cm of snow on the ground. At 9:00 AM the overheating alarm went off on my solar water heating panel. It only took a few minutes of sunshine to raise it's temperature to +85C which is the alarm trip point. The problem was my energy saving device (OK, I get things wrong sometimes..) which uses lack of output from a PV panel to shut the water heating down overnight. The PV panel was buried in the snow and didn't know the Sun was up!

Brian.

 

[Easy peasy]

There are plenty of single stage topologies that will give high power factor ( >0.95) and high efficiency, 95% is easily do-able in a resonant flyback with synch rect on the o/p

...

 

[mtwieg]

I agree that a good single stage can be get 95%, though with LED drivers I assume it needs to be a two stage design with electrolytic capacitors absorbing the 2xfline ripple power. A single stage design would have to pass that ripple to the LEDs, and to my knowledge LEDs don't tolerate that well.

 

[Easy peasy]

LED's can tolerate any amount of ripple, most LED designs (for signs, indicators etc) are pulsed, the amount of ripple seen at the output can be reduced by suitable sized capacitors, thus giving high PF and moderate to low ripple to the load, average led current is usually the controlled variable ....

 

[treez]

Thanks, as you know, from an offline SMPS, 150W LED loads, are best supplied with a high voltage, eg, 100 to 300VDC, as it means you can put the LEDs in series.....most LEDs are 700mA max.

As such the output current of the flyback (or whatever) would be low-ish. As you know in this particular case, sync rects would not be wanted due to their cost.
...Now i have set the scene, which would be more efficient?....., a single stage "flyback LED driver", or a "boost PFC followed by LLC LED driver"?

 

[KlausST]

Hi,

.most LEDs are 700mA max.

This is DC current.
You may use higher value current when pulsed. The datasheet should tell you.

Klaus

 

[treez]

Thanks, sorry yes i meant "maximum of the average value". Indeed, the peak can be higher than 700mA

 

[mtwieg]

LED's can tolerate any amount of ripple, most LED designs (for signs, indicators etc) are pulsed, the amount of ripple seen at the output can be reduced by suitable sized capacitors, thus giving high PF and moderate to low ripple to the load, average led current is usually the controlled variable ....

I haven't worked with modern COB LEDs recently, but I recall always seeing dire warnings about peak currents exceeding the maximum IF rating. Usually peak current ratings are 150% or less of the rated average IF, and are only tolerated at low duty cycles. Perhaps these specs are ignored in practice? Or is it standard practice to derate the LEDs so that your actual average IF does not approach the rated IF?

 

[treez]

To be honest, its a good point, ive never seen a LED datasheet which made clear that peak current can be twice the average, say but ive seen loads of products that use LEDs like that , so presume it is ok.
I presume that the delicate led bond wires are pretty low resistance, so the extra rms current through them if peak is twice average currnt, is going to be minimal. So they should be OK.
We had a contractor who told us that even if a LED bond wire does not heat up, a one off shot of current through it of say 20A for 100ns would destroy that LED.....but this is baseless.

With LEDs for streetlights, having your LEDs be the brightest on the market is what its all about....so you are constantly having to update leds........and therefore you have to make up LED lights and fixtures, etc, out of 700mA LEDs, because this is where the most choice is to be found...and where one is going to find the "Next brightest led chip".

COBs are good, but they are not as common as ~3V, 700mA chip LEDs, so not likely to feature the next brightest led on the block. (I know you didnt enter into that debate , but i am just making the point since it determines the type of topology used.......and single stage flybacks dont need sync rects.)

But it is amazing that any LED driver which comprises a boost pfc , then a downstream pwm stage, could be more than 90% efficient...surely it cannot be so? It just cannot.