LED Driver Efficiency

One limit to LED driver efficiency is that current flows through rectifiers (diodes used as one-way gates), and diodes have a forward voltage drop and thus waste power. For example, if a 1A driver goes through a diode with 0.6V drop, then the wasted power is P = I² R = 0.6W. If the driver is running in a mode where the diode is ``on'' 1/3 of the time, then the average power loss is (0.6W)(1/3) = 0.2W. If the driver is driving 10W, then the loss is 0.2W/10W = 2%.

My bad-but-almost-good inspiration was to replace each diode with an LED. Then, the power is no longer wasted. Ta-da!

Well, not quite so fast. A standard diode and an LED are different in many ways, leading to loss of the supposed good properties.

Briefly, a boost regulator works by charging up an inductor by shorting it to ground, then disconnecting the ground end. As the inductor field collapses, it tries to maintain the same current, and in doing so will express as high a voltage as it needs to in order to get current to flow.

Boost (step-up) converter. It switches between ``charging'' and ``driving'' modes. The left shows the ``charging'' mode, where a current is established in the inductor. The right shows the ``driving'' mode, where the inductor current is dumped into the load portion of the circuit. Note the voltage entering diode D is VIN plus the voltage expressed by inductor L in order to keep current flowing through L. From [SR05].

See the diode? Bad diode. Bad, bad diode. That's 2% efficiency. Wasted. Wasted! We just replace it with an LED, and -- bling! -- there was more efficient light!

Except...

• The forward voltage, V_f, for LEDs is higher than for other diodes. The inductor thus has to rise to a higher voltage, and that creates higher parasitic losses.
• LEDs are poor at resisting reverse bias. So the maximum load voltage needs to be small. Many LEDs resist a reverse bias slightly higher than V_f, but that does pretty much limit the load to a single LED or some LEDs in parallel, but definitely not a string of LEDs in series.
• Increasing the rectifier voltage means that the inductor has to do more work to feed the circuit ... it discharges faster and so needs to return sooner to the charging mode. This decreases LED losses -- but increases the transistor on time, and thus increases transistor losses.
• The discharging inductor provides a high current which charges both the capacitor and drives the ``main'' LED. So while the average current through the ``rectifier'' LED is the same as the average current through the main LED, the peak current is much higher. For example, if the regulator runs in each mode 50% of the time, the peak rectifier current is at least 2x the average LED current.

Okay, that's not such a good idea after all. Maybe it works better on a buck converter? Here's a buck converter. Note that it has an evil power-wasting diode!

Buck (step-up) converter. It switches between ``building'' and ``decaying'' modes. The left shows the ``building'' mode, where a current is established in the inductor. The right shows the ``decaying'' mode, where the inductor current is dumped into the load portion of the circuit. Note the voltage exiting diode D and entering inductor L is below zero. The voltage drop is created by inductor L in order to get a voltage across diode D in order to keep current flowing through L. From [SR05].

Unfortunately, it looks like no win here, either.

• The high V_f means the inductor has to make the rectifer end go highly negative. That means higher parasitic losses.
• We probably have to stack the ``rectifier'' LEDs to resist V_in. Ooops, that means the inductor has to go even further negative for yet more parasitic losses.
• High negative voltages means the inductor collapses quickly, so the on time of the diode stack will be small and they won't put out much light. At the same time, the switch will be on more, increasing switch losses.

So, um, points for thinking ``outside the box'', but it's back to germanium diodes or using MOSFETs as the diode.

References

[SR05] ``Switching Regulators''. No date, authorship, or other identification given. From www.national.com/appinfo/power/files/f5.pdf as of 2005/07, linked from www.national.com/kbase/category/Power.html.