A voltage multiplier is applied as a high source impedance, 36V, 300mA, 12W LED driver. The high source impedance is ideal for driving power LEDs because it accommodates variations in LED junction voltage so that there is minimal change in LED current and brightness. The cascade multiplier is essentially a charge pump that is similar to the popular off-line capacitor limited LED drivers, but offers a substantially higher load current capacity as well as transformer isolation. Note that this circuit topology and application is new to the world.
Voltage Multiplier LED Driver Schematic
The problem with LEDs
LEDs are essentially constant voltage devices. Unfortunately, LEDs vary in voltage due to manufacturing variations and temperature. Thus the “constant voltage” is not constant at all—this leads to variations in current when devices are paralleled and also presents a problem when operated from constant voltage power sources—and a serious problem when powered via automotive systems that range from 11 to 14.5V. As a result, various types of ballasts are used in an attempt to regulate current:
This power supply circuit is designed around the “garden variety” 12VAC landscape lighting transformer. Transformer availability and selection has long been a problem for the experimenter, and has often driven them to experiment with dangerous off-line capacitor limited power supplies. However, the long neglected 12V landscape lighting transformer is a good solution due to numerous advantages such as low cost, availability, and power variations (VA ratings).
Directly rectifying the 12V transformer secondary is not a good choice because the rectified voltage is generally way high (typically 18VDC) due to peak detection of the sine wave—in the past, this required a large, inefficient ballast resistor to drop the extra 6V. However, a voltage multiplier (quadrupler) has some very interesting properties — not only does it increase the voltage substantially, but it offers low voltage regulation that is ideal for powering constant voltage devices such as LEDs.
The combination of the 12V lighting transformer and the voltage multiplier is like a match made in heaven. The multiplier delivers 36V (loaded). To accommodate the higher output voltage, (3) 12V LED devices are connected in series to match this voltage. The voltage may be adjusted simply by varying capacitor size. Capacitors are garden variety aluminum electrolytics. The common 3 to 4W LED is bright, but a little low to be useful in general purpose lighting. Connecting (3) in series provides 9 to 12W, a much more useful level of illumination. The 12V LED unit is commonly available due to automotive and 12V battery applications.
The initial circuit considered was the classic Cockcroft-Walton multiplier. While it functions, the output current tends to be too limited for this application. Note that all circuits tested here are quadruplers — there are references to triplers on the web, but this is in error because multipliers multiply via cascading doublers so that there can be no odd factors.
Cockcroft-Walton Voltage Multiplier Schematic
To increase the current rating, I repositioned the input and output capacitors so that they are common to either the transformer secondary high side or ground. This variation seems simple enough, but I have been unable to locate “prior art,” so I am suggesting that this is new to the world. I have coined the name: “Stretch Voltage Multiplier.” All capacitors are the same value.
Voltage Multiplier Test Circuit Schematic
Voltage Regulation Graph
In the process of analyzing the graph, I plotted the effect of a fixed resistor — this is the pink line. It blew my mind that the voltage multiplier source impedance behaves exactly as the fixed resistor —I fully expected the plot to be non-linear. The discontinuity below 20V is likely caused by reverse voltage on the polarized electrolytics — under this condition, electrolytic capacitors tend to act as rectifiers — operation in this region is not recommended. Unloaded, the voltage swings up to 74V—fortunately, my capacitors did not smoke — 63V capacitors are robust enough to tolerate a surge of this magnitude. For this reason, the power must be switched at the transformer primary rather than the DC output.
I think I have drawn the load line correctly. The lower the slope of the load line, the better the current regulation — it indicates what happens to LED forward voltage over the temperature range, as well as the limits of manufacturing variations. The graph source is Bridgelux, an LED manufacturer:
The input current indicates the presence of harmonics, but the waveform is not all that bad — peak current is well under control. I did not expect the high degree of symmetry. Power factor is obviously leading as expected. Capacitor ripple current is well within the current rating of garden variety capacitors. Check out this link for high current, small size, low cost Nichicon PW series capacitors — note that the capacitor ripple current ratings are provided:
Another curiosity is the impedance transformation. I should not be surprised because it is a law of physics, but I had in the past limited it to the electromagnetic transformer turns ratio in which the impedance ratio = the turns ratio squared. Well here, in this multiplier the same phenomenon exists — the input impedance * voltage multiplier ratio (4) squared = the load resistance.
Practical LED driver circuit
With the feasibility demonstrated, I designed the previous voltage multiplier LED driver.
The output current may be scaled simply via adjusting the capacitor values. All capacitors are of equal value. The 3 output capacitors are required to reduce the ripple voltage to 3Vp-p. Dividing this value by 3 yields 1Vp-p for each LED — this is about as high a ripple as recommended—of course, the output capacitor may be oversized for even lower ripple — (3) output capacitors are suggested to merely keep all capacitors identical. Note that being half-wave, the ripple frequency is 60hZ and the ripple voltage calculation falls in line with C = 0.7 * I /(delta E * F). See:
Note that capacitor size is frequency sensitive. All measurements were performed at 60hZ. For 50hZ operation, increase capacitor size by 20%.
If you scale up the current rating significantly, be sure to also increase the diode current rating.
The bleeder resistor (R1) is necessary to discharge the capacitors because the LEDs stop conducting below their Vf threshold voltage. This leaves a substantial residual charge in the output capacitors and could pose a hazard if accidentally shorted. The energy is in the order of 1 joule — healthy arc!
Trimming the current
DO NOT attempt to trim via measuring the LED voltage — LED voltages are nominal — let the voltage fall where it may. Current is what lights the LED and this is what may be either too high or too low, so measure and adjust the CURRENT. Adjusting C1, C2 or C3 adjusts the output current. A 10% change in only one of these capacitors changes the output voltage by approximately 1V (resistive load). It is easy to increase the current by padding the capacitor(s) with 100uF capacitors. To keep the input current symmetrical, I suggest padding either C1 & C2 equally, or C3 alone. While padding increases the output current, decreasing the current requires a new capacitor selection.
How about using a 24V transformer to power a 72V LED string? why not? This gets into some serious power.
You can see that I was lacking one capacitor so I had to “punt” for the final capacitor…
For the future — this opens new possibilities…
- Voltage doubler LED driver
- Constant current LED ballast/driver
- Buck regulator LED ballast/driver
Undocumented words and idioms (for our ESL friends)
prior art – noun – idiom related to inventions and patents in which a search is made for past technology in order to determine patentability of new ideas or devices.
pad or padding – verb – increasing the value of a component by physically adding parallel capacitance or resistance — in the early days of radio, a padder was a variable capacitor consisting of spring loaded leaves and mica insulators that were compressed via a screw adjustment.
punt – verb – idiom — football term in which all plays have been exhausted and the ball must be released to the opposing team via a punt or kick — in electronics (or other technologies), it indicates an inferior solution…