The LED driver chip (AL8861) is specified to drive up to 1.5 amps across the LEDs. I only want to be able to drive 750mA accross the LEDs. I am using two Osram LEDs spec'd at up to 1.8A, connected in series. The input voltage to the control board is either 12V or 18V, well within the spec of the driver chip.

Examining the output current from the Illuminator Power supply, the output current is not controlling properly. The LED driver chip (AL8861) looks a lot like a buck converter power supply, except this one ignores the output voltage and monitors the current only. Looking at the schematic, a 0.1 Ohm current sense resistor (R3) goes from Vin to the Anode of the LEDs (via J2). Current returning from the cathode of the LED (J2 pin 4) goes into the inductor (L1). The lower end of the inductor goes to the switch pin on the chip (U1) and the anode of a schottky diode (D2). The current through R1, LEDs, L1 is measured by the voltage drop across R3. The current setpoint (brightness) is set via a DC voltage from a potentiometer connected to J1. Zener diode D1, R1 form a crude voltage regulator (about 1.6V) for the control setpoint.

When the chip powers up, the  LED/Inductor current is below the turn off threshold, so the switch pin on the chip grounds the lower end of the inductor, and the current rises fairly linearly with time. When the LED/Inductor current rises to the turn off threshold, the switch pin on the chip turns off. The inductive kickback drives the lower end of the inductor to a positive voltage. When that voltage rises above the Vin, D2 starts conducting and the LED/Inductor current circulates through the LEDs, L1 and D2. There is hysteresis in the comparator monitoring the voltage from the sense resistor, so the LED/Inductor current must decay to a slightly lower point than the turn off point before it turns the switch in the chip back on to start replacing the energy in the inductor that was consumed while the switch was off. The system normally runs in this region, with the LED/Inductor current oscillating between the turn off point and the turn on point. The frequency is set by the Inductor value, the LED power and the input voltage. It is designed to run around 300KHz, but the chip is spec'd to operate up to 1MHz.

Adjusting the voltage to the brightness control pin moves the turn off/turn on points, and sets the average current through the LEDs.

The design that I used for the PC board was based on the Data Sheet "Typical Application" drawing.

The problem that I am seeing is that as the current in the LEDs is increased, high speed noise is appearing on the switch pin of the chip causing the current to be measured incorrectly. When the chip is operating normally the current oscillates in a roughly triangular waveform riding on a DC level and the voltage on the switch pin is a clean square wave. When the noise starts showing up, the oscillations get much larger and less regular. The voltage on the switch pin has fast spikes while the switch is on, indicating that the switch is turning off briefly.

The first thing that i tried was to put more capacitance on the power input. A 470uF cap at the power input connector was used. This changed nothing.

Next, I put a 100pF cap in parallel with the current sense resistor (R3). This improved the behavior some, increasing the current where the noise started showing up, but it still had problems below the current that I wanted to operate the LEDs at. Some being good, more might be better... so I replaced the 100pf cap with a 0.1uF cap. This changed the behavior. The current could be increased considerably higher than with the 100pF cap, but before it reached the level that I wanted, the triangular oscillation on the current waveform disappeared and the switching frequency of the inductor current went up to about 4MHz from the 300KHz range. In this mode, I have no idea how the chip is operating. The inductor current appears to be constant. It adjusts with the brightness control input voltage reasonably well, but the switch frequency is about 4X the maximum operating point from the data sheet.

Looking at the schematic of the Diodes Inc Eval board for the driver chip, they have a capacitor across the LED's. One version of their board has a 0.1uf cap and the other has a 1.0uF cap. I removed the cap from the current sense resistor and put a 0.1uF cap across the LEDs. This allowed the current to be set at a higher point than the unmodified board, before the noise appeared, but still below what I wanted. I swapped the 0.1uF cap for a 1.0uF cap, and it can be set to a higher setpoint but still less than the 750ma that I want.

The PCB is a 2 layer board with lots of copper on the ground path and reasonably tight routing of critical connections like the sense voltage and switch node.

There are still a few things that I am going to try before punting this and using a different driver chip.