Before ordering v1.1 of the LED Driver PCB, I wanted to measure output ripple. The light has no noticeable flicker to the naked eye or when using my phone’s camera at 30, 60, 120, or 240 fps. However, I have noticed some unsmoothness during dimming transitions and wondered whether this could be due to ripple or the responsiveness of the TPS92511 buck converter circuit.
Here is the schematic for v1.0 of the LED driver.
With this circuit I measured about 456 mA peak-to-peak ripple at a switching frequency of 300 kHz. This 300 kHz number is the buck converter switching frequency (i.e., the ripple frequency while the LED is on), not the PWM dimming frequency. 300 kHz is 15× higher than 20 kHz, so the switching fundamental itself should not be directly audible (though audible noise can still occur via magnetics/ceramics excited by lower-frequency components or mechanical coupling). The PWM dimming frequency is set by firmware and is 19.5 kHz.
There are a few knobs to turn to decrease the current ripple:
Increase the switching frequency (fsw): decreases ripple and can allow a smaller inductance, but decreases efficiency due to higher switching losses. Also increases the min on time of PWM dimming. From the TPS95211 datasheet:
In order to maintain good dimming linearity, the minimum LED current pulse width is suggested to be three switching cycles. For example, if the switching frequency is 500 kHz, the minimum DIM pulse width is 6µs and the dimming frequency is 150Hz, a contrast ratio of more than 1000:1 can be achieved.
Add capacitance across the LED: decreases LED current ripple while maintaining inductor ripple, but can increase overshoot (the capacitor must charge when the buck turns on) and slows the current decay which can reduce effective dimming resolution.
Increase inductance: decreases ripple but slows current ramp-up time, which can reduce resolution dimming resolution. Another potential downside of increased inductance is it can require a larger more expensive inductor.
Another thing to keep in mind is that while reducing LED current ripple is desirable, you still want to maintain sufficient inductor ripple for proper operation. The TPS92511 is designed to operate in continuous conduction mode (CCM) and relies on inductor ripple current for its internal control scheme.
I first increased the switching frequency from 300 kHz to 500 kHz, which reduced ripple from 456 mA p-p to 300 mA p-p. Next, I added capacitance directly across the LED to filter the switching ripple at the load. This was very effective: with 1 µF across the LED the ripple dropped to 44 mA p-p, and with 0.22 µF it dropped to 92 mA p-p. The downside was that larger capacitance also increased the effective turn-off “tail,” so at high PWM duty cycles the LED current did not fully fall toward zero during the off portion of the PWM cycle (e.g., 190 mA at 95% duty with 2×0.22 µF). To improve PWM edge fidelity while keeping ripple low, I reduced the LED capacitance and increased the inductance. Moving to 100 µH–120 µH with 0.22 µF achieved ~54–56 mA p-p ripple while reducing the amount of current “held up” at high duty, although I still observed a residual floor (~110 mA at 95% duty) in this configuration.
I ended up with a result I was happy with: 120 µH and 0.22 µF across the LED, which produced ~54 mA p-p ripple at 48 V input and 500 kHz switching. I also observed turn-on overshoot in this low-ripple configuration (typically ~612–620 mA peak). To reduce ringing/peaking, I added a small series resistor with the LED capacitor. Testing 0.1 Ω and 0.5 Ω did not materially change the measured ripple or overshoot in my setup, so I kept 0.1 Ω in the final schematic primarily as a placeholder for future damping/EMI tuning if needed.
I ended up with this final circuit.
