I built a circuit identical to the design example in the LM3445 data sheet to drive an LED light for use on-camera with my SONY video camera and with my iPhone's built-in video camera. The circuit works well and puts put a lot of light that looks splendid to the naked eye.
To my dismay, I discovered that there are horizontal bands in the video image, most noticeable with the iPhone. There are about four pairs of alternating dark and light bands on the iPhone screen. This makes sense, since the iPhone's native scan rate is 30 Hz and the LM3445 circuit is producing 120 Hz flicker.
I have measured the amplitude of the voltage variation across the LED and it amounts to just 2% of the total (mostly DC) voltage across the LED. But that 2% is definitely noticeable by any video camera. It is subtle and the contrast is quite low, but it is there. At higher scan rates, it becomes less of an issue because the entire video scan most likely occurs within one band (either dark or light).
I have a T.I. LM3445 demo board (see AN-1935) which shows exactly the same behavior.
So, while the LM3445 advertises that it has no 120 Hz flicker, it is a matter of degree.
I am now searching for a way to get rid of this remaining 2% flicker before I abandon the LM3445 for a better solution.
The problem centers around the peak of the input voltage, which forces a peak in Vbuck (see LM3445 data sheet). This peak is getting through to the LED and is modulating the LED brightness ever so slightly.
The brute-force solutions are unattractive (and marginally effective).
Has anyone tried to solve this particular problem?