Discrete inverter for LED

Projects /  LED  Original date: 07/26/2014

This is a high current boost inverter for driving a LED from 1-2 AA battery. I used a hybrid BJT/MOSFET to increase its efficiency.

Introduction

This is a tiny high power boost inverter that I built for modding a Chinese "Magnite" (TI give-away at their booth) with a $0.60 1W LED. The circuit is crammed to 2 PCB - one that house the inverter while the other is for the LED. This is an ultra low budget project with what I have on hand.

The LED is surface mounted onto a large area of copper fill on the PCB. There are 4 pieces of copper tape soldered to the PCB and is glued to the aluminum casing of the flash light for heat sinking. A gold plated header pin soldered on back side of that PCB makes electrical contact with the metal tubing of the flash light and act as an improvised on/off switch. (Don't have the beam control anymore.)

The inverter circuit is on the smaller PCB which make contact with the +ve terminal with that funny 'e' shaped gold plated header pin.
I came up with the circuit when I was playing around with LTSpice. This is an old project. It was the first doubled side PCB at the time.

LTSpice simulation

I came up with this circuit as I was playing with LTSpice, a free analog simulator from Linear Tech.
LTSpice download  (alternatively just google for LTSpice)

Note: I used PMOS (and PNP) because that was the only high current MOSFET I had in the tiny SOT-23 package.  My previous version use a larger package (D PAK)  NMOS and NPN.

Initially, electrons goes from ground to the base of Q1 via L1, D1, R4, L2, R1.  This causes Q1 to conduct which in turns coupled to L2.  This has a positive feedback loop increasing the base current to Q1.

Note: L1, L2 have 1:1 turns ratio and winded together on same core.  You can use thinner wires for L2.

Green trace is the voltage between the positive rail and the MOSFET gate, the blue trace shows the collector current of Q1, red trace shows the drain current of the MOSFET.  Notice the vertical scales!  R3 sets the amount of collector current for Q1.

Finally at some point, the voltage crosses the Vth (threshold voltage) of the P-MOSFET and it starts to carry the bulk of the current inductor current.  This is the point when the collector current of the transistor suddenly drops in the graph because its VCE drops as the MOSFET starts to conduct.  It does that every cycle.

During this time, the inductor current increase until a certain point when it starts to saturates. The voltage increase no longer keeps up.  This is sort of like a bubble burst in a stock market as it cannot sustains the infinite growth and everybody start to panic to sell their stock and prices spirals down.   The inductor current has to go somewhere else, so it passes through D1 into the capacitor C2 which runs the LED.

Green trace shows the inductor current while the blue one shows the voltage across the MOSFET/PNP.  Notice that the MOSFET from the previous graph carries the bulk of the current, that's why the circuit has a low voltage drops which peaks out at about 280mV and just before its switches off and rise rapidly.  There is some ringing when the MOSFET/Transistor turns off.  This is due to LC oscillation of the parasitic capacitance and the inductor.

Q2 is the voltage (negative) feedback control that sets a peak bias point for the MOSFET/PNP.  The voltage R5/R2 sets the peak output voltage of the inverter.  The output voltage has an AC component (sort of like a sawtooth waveform) as the capacitor C2 charges up and discharges in the cycle.

C3 serves multiple purposes.  It is a low impedance AC path for the gate current to the P-MOSFET and the PNP.  It also stores an average bias point for the circuit for the remainder of the cycle.

R4 bias point is taken from the output of the inverter.  This allows it to operates down to 0.3V.   It works great from a bench supply as a bragging right.  In real life, this is pointless/stupid as the battery internal resistance is so high at this point, its voltage is dropping like a rock.    This the part that most of the "Joule Thief" fans fail to understand.  So enjoy the extra few milliseconds of very dim light.

Note: Contrary to newbies' believe, MOSFET needs lots of gate current to charge/discharge it gate capacitance during switching!  Should start thinking about how circuits works in AC.

Note: The circuit is voltage control because that was all I could fit inside the case. I am running the "1W" LED at about 0.3W to give some headrooms. When a LED heats up which its junction voltage decreases thus increasing current/wattage and this can spiral out of control blowing itself up. Also that was done to give a long battery 10 hours + battery life. You should really use a constant current source for driving a LED properly.  I would have used a Sipex LED driver chip if I had one.

EagleCAD schematic and layout.

I cut the board along the circular outline and solder a hand winded toroid inductor on to the 4 pads.

SketchUp rendering of the PCB as I forgot to take a picture before completing the LED/inductor/driver PCB assembly (or was it a sandwich).   Thanks Jerome Lamy for  that EagleUp plug in!