A first design goal for the light source is obtaining enough light for sufficient carrier generation. It will be shown that this design requirement for carrier generation immediately taxes the limits of LED illumination. However, it wasnt until a working circuit was developed and tested that these issues appeared. LED illumination was tested initially because an array of high-power LEDs has the distinct advantage of low-voltage operation. The voltage of a single LED is related to the energy an electron drops across the band gap of the semiconductor. Typical LEDs operate at a few volts as opposed to the hundreds required for a xenon strobe. Lower voltages suggest safer operation and maintenance as well as simplified design.
Using the brightest LEDs available at the time provided a reasonable starting point. The brightness of the LED during continuous operation is significantly lower than the maximum light it can emit. By pulsing the LED on and off quickly, the device can be pushed beyond safe design limits. The specifications of a Cree DRAGON LED limited the absolute maximum safe current to 4A pulsed. By carefully analyzing the pulsed power specifications, I chose to overshoot the current to approximately 10A. This value puts the operational current within a factor of safety of 2.5 outside design limits. Key to making the LED survive such pulses is a very short duty cycle from 10-150us at just a few tens of hertz. This duty cycle keeps the total average power below the stated maximum of 7.5W and is standard practice for a pulsed application. Well separated pulses allow the device to cool down sufficiently between to avoid thermal destruction. It is assumed that typical LED longevity is sacrificed in this off-spec design.
The design for turning on and off an LED seems straightforward; however, contributing factors quickly escalate the design difficulty. Delivering large amounts of power in short pulses is typically done using a capacitive discharge. A capacitor is charged to a specified voltage and connected to the load. The capacitor is discharged until empty or until disconnected from the load. Obtaining the high energies required for suitable excess carrier generation requires high values of capacitance. Large capacitors are good at buffering voltage drops and storing energy but poor at turning off quickly. To compensate for the large required capacitance, the board design, including components, traces, and layout have to be optimized for low resistance and inductance.
In a pulsed power application, the LED changes temperature and operational parameters such as turn-on voltage must be compensated. As the temperature and current increase, the turn-on voltage increases from 1.8 to 2.7 Volts. To create a uniform light intensity from pulse to pulse as well as allow interchangeability of LED colors, a current clamping circuit was used. The LED is connected to the capacitor bank directly and to ground by means of a fast MOSFET. The collector of a BJT is connected to the gate of the MOSFET and the emitter is grounded. A high power resistor is placed from the base of the BJT to ground and connected in series with the LED.
When the current through the power resistor creates a voltage drop of approximately 0.7V, the BJT turns on and sinks gate drive from the MOSFET. The built-in base-emitter voltage of 0.7V is used to choose the value of the power resistor. As an example, a 0.1ohm resistor requires 7A to produce a 0.7V drop. By placing a power potentiometer in its place, a variable current limit can be set. This eliminates any variation in the current due to LED temperature and associated voltage swings. Each pulse has the same shape and intensity as the previous.With the BJT-MOSFET current clamp, the output current waveform is clipped to a square shape. The turn-on and turn-off slopes are unaffected by this circuit because of the high switching speed of the clamping circuit. Switching and control of the overall circuit is accomplished with two 555 timers. One timer in astable (free-running) mode sets the repeat frequency while a second one in monostable (edge triggered) mode sets the pulse width. The output of the second timer is connected to the MOSFET gate to turn on the LED. Each of the timing outputs are low current and do not require any special board layout focus.
All the high current legs of the circuit require very careful component and design considerations to minimize inductance and stray capacitance. The first step in keeping inductance low is choosing parts with the lowest possible inductance. With high-quality parts, the board layout took shape. All the LEDs were located centrally to provide uniform illumination of the sample. The capacitors, connectors, and ancillary circuits were placed so they would not sit proud of the board thus minimizing the distance between the LEDs and the sample. All high current circuit board traces were designed as wide and as straight as possible. The return paths for current are short and low impedance.
Even applying careful design consideration to the board, some additional problems appeared during testing. The board was designed to run off a 12V power supply requiring up to 1A of current. When the LEDs were triggered, the impedance of the empty capacitors (as seen by the power supply) dipped too low. The output voltage of the power supply dropped from 12V to 3V. By adding a tenfold capacitance increase for input buffering, the output voltage dropped to 6V instead. Unfortunately, the spike of current occurred on a time scale faster than that which the power supply could recover. This current spike propagated back through the power supply and into the 120V lines. Oscilloscopes and the other photoconductance decay equipment experienced the spikes as spurious voltage and anomalous triggering. Separating the power supply onto a different physical circuit did not remove the spike entirely.
In addition to the voltage spike challenges, the LEDs did not produce enough light to make consistent measurements. When the LEDs were triggered, they could be seen flashing dimly. Increasing the pulse width increased their apparent brightness as expected. Even at their brightest level, the intensity was not enough to create a sufficient excess of carriers for meaningful results. A small blip occurred on the oscilloscope in a way that could not be distinguished from the impact of the current spike. Due to the combination of the lower energy output and the excessive current dynamics, the LEDs gave way to a brighter Xenon strobe option.
* Hat tip to Brandon Harris for pointing out the missing output resistor (shown in red).