Series Connected Voltage Boost Circuit for a Battery Operated LED Lantern

A circuit that goes in series with a white LED and a 1.5 volt cell to make > 3 Volts to drive a white LED.
This circuit and a similar article was first published as an idea for design in the March 15, 2004 issue of Electronics Design Magazine.


Photo. This is the test circuit -the basic driver is only two transistors, two resistors, the circuit was evaluated using a white LED, but when it was time to button it up and archive it, I replaced the expensive white LED with a cheap green one.

Othere LED Driver Circuits on this site:
White LED drive from 1.5V
White LED inductorless drive from 1.5V and 3V.
Fast precision LED driver
White LED Stroboscope With Constant Duty Cycle and Constant Current Drive

When Dave and I were kicking around ideas for the best arrangement of components for white led flashlight, Dave suggested putting the voltage converter in place of the second cell in a two cell flashlight, and substitute an LED assembly for the incandescent lamp. "Yeah, I that would be ideal..." was my hesitant response, though I didn't see how this would be possible. It took someone like Dave, who didn't' know it was impossible, to suggest doing it. Eventually  a topology emerged that can actual run a 3.5 volt LED when it and the driver in series with a 1.5 volt flashlight cell. Well kind of, anyway. The trick was to put one additional component, a diode that is actually part of the driver, in parallel with the LED. Read on.

 

Figure 1. Arrangement of components inside a flashlight.

This was identified as the ideal arrangement for a retrofit kit.

I was already using energy stored in an inductor that is part of a blocking oscillator to boost the voltage available for an LED (see the schematic below).

Figure 2. This entire circuit was built entirely inside
a light bulb base, but building it was a challenge.

While this provided the necessary voltage boost, it required that the blocking oscillator be powered from both terminals of the battery. It seemed that some re-arrangement of the parts might result in a practical topology that would allow current built up in the inductor to be dumped through the LED. The circuit below emerged.

Figure 3. The series connected voltage boost circuit is composed of three sections: The 1.5 volt  flashlight cell, the inductive switching circuit, and an LED assembly with its shunt diode. Both the 4.7k and 100k resistor are optional.

This circuit above is that of a series connected voltage booster. The schematic represents an LED assembly (D3 and D4 with an optional 4.7k resistor), a 1.5 flashlight cell, and the series boost circuit. If you compare the circuit in Figure 3 with the basic blocking oscillator circuit in Figure 2, it should be apparent that the series connected voltage booster is designed around two blocking oscillators in parallel with a single inductor core in common. The oscillators are locked together by virtue of their using the same core. The PNP blocking oscillator connects and disconnect the positive power supply to the inductor and the NPN blocking oscillator connects and disconnects the negative supply to the inductor.

In order for the oscillators to run, a diode, D3, needs to be placed across the LED (D4) to power the circuit while current builds up in the inductor. When the transistors turn off, the inductor current flows through steering diodes D1 and D2, and  LED D4. Diode D3 that conducted during the build up of current in the inductor is reverse biased while the inductor discharges. The 4.7k resistor across D3 and D4 is optional; it lowers startup voltage to below 1.2 volts, about 100 millivolts lower than the normal 1.3 volt startup voltage.

Figure 4. Current builds up in the inductor by passing through the two conducting transistors and diode D3, which shunts the (not shown) LED.

Somehow the transistors turn on (the startup of a blocking oscillator is covered in the other LED flashlight page on this site and also covered in many text books, so I'm not going to do it again here). I will point out that there is a 100k resistor to make sure something conducts when power is first applied. Without it, the oscillator relies on leakage currents to get started, and that really makes me uncomfortable. Yours will probably work fine without either the 100k or 4.7k resistors. Once the transistors are conducting, current builds up in the inductor until the transistors come out of saturation. The time it takes for the transistors to come out of saturation is determined by the largely characteristics of the inductor (initial inductance, core saturation characteristics, and in some cases, winding resistance). The current path is shown in Figure 4, and is through D3, the inductor, the two transistors, and the 1.5 volt flashlight cell.

Figure 5. When the inductor discharges, current flows through D1 and D2 instead of the transistors, and also through LED, D4.

Once the transistors turn off, inductor current continues to flow, but since the transistors are off, the current flows through previously revers biased diodes D1, D2, and the LED D4. As shown in Figure 5, the discharge current also flows through the 1.6 volt flashlight cell and of course, the inductor, which is the source of the current during this phase of operation.

It should be noted that all the inductor ripple current passes through the 1.5 volt flashlight cell. If this were to be a large current, I would be concerned about heating in the cell from I²R losses, where R is the internal resistance of the cell.

The inductor was made by a continuous winding of 45 turns of #32 single Beldsol magnet wire on various cores. The taps were made at 15 and 30 turns (each tap was 15 turns from its nearest end).

Here are some cores that I tried, and the results obtained

Ferroxcube TC8.2/3.7/4-3E7,  This core gave the best performace of those tested. With this core, the supply started with less than 1.1 volts applied (with the 4.7k start up resistor). The oscilltion frequency was about 20 kHz with a 1.5 volt input. The core is an 8 mm diameter high permeability (ui = 15,000). This gave an inductance of about 1.7 mH in the 15 turns between the taps..  

Very large high permeability toroid core, about 4 cm in diameter. Even better than the 9 mm core, but it was really too big for this kind of use.

Ferrite RF tuning slug 10 mm long: Very poor performance - needed about 1.8 volts before the LED would visibly start to light.

RF torrid abut 10 mm in diameter: Even worse than the slug for some reason. Needed about 2.0 volts to get a faint glow from the LED. It has a large enough winding window for a lot more wire, and it probably would have worked at a lower voltage if I had put more turns on it.

What these quick experiments with various cores confirmed is that this circuit worked best when operating with moderate inductance's (2 millihenry per winding) and operating in the tens of kilohertz. Perhpas faster transistors and maybe litz wire would make this work better at higher frequencies.
   
As my objective was to prove the topology would actually run an LED connected in this configuration, I did not try to optimize the inductor, and only did some experiments to demonstrate the behavior of a few cores I had on hand. If you optimize this circuit and would be so kind, please send the details to me so I can post them here. My email address is below.  (Note that the address is a .GIF image to evade Spambots.)

 

The circuit shown on this page doesn't use a decoupling capacitors across the power source. That is because adding a decoupling capacitor in parallel with the batter would often be impractical.  In some cases, where such a capacitor is practical, there might be an improvement in performance if a capacitor were to be placed across the battery.

Danger of Eye Damage From Visible Light Emitting Diodes

 

Dick Cappels' Project Pages          http://projects.cappels.org    

Contents ©2003, 2004, 2006 Richard Cappels All Rights Reserved  This page first published in March, 2004.
HOME