# Reverse Engineering A Real Candle

[cpldcpu] just can’t leave the mysteries of candles alone. We’ve covered his explorations of candle flicker LEDs before, but this time he’s set his sensors on the real thing. [cpldcpu] hooked a photodiode to his oscilloscope, pointed it at a candle flame, and recorded the result.

The first interesting observation was the candle slowly changed brightness, whether it was interacted with or not. Next he measured the effect when the flame was disturbed by small gusts of air. This produced a bright flicker with an oscillation at 5Hz before returning to steady state, which as [stygiansonic] mentioned in a the Hacker News comment, is a known phenomenon used in flame detectors. Neat! There’s even an equation:

Under normal gravity conditions, the flames have a well defined oscillation frequency which is inversely proportional to the square root of the burner diameter, D, and to a good approximation can be written as f » 1.5/D½, with D given in meters.

[cpldcpu] then compiled his measurements into a series of graphs and ultimately an animated gif comparing the candle steady state, a real candle’s flicker, and the flicker he recorded from a candle flickr LED. It’s surprising how different the fake is from the real thing. You can look at his measurements and code at his github.

[via Hacker News]

# Reverse Engineering Candle Flicker LEDs, Again

Flickering candle LEDs are seemingly everywhere these days, and like all fads, someone has to take a very close look at the engineering behind them.

[cpldcpu] had earlier taken a look at the controller chip in these candle flicker LEDs by measuring the current used and developing a statistical model of how these LEDs flicker. That’s math, of course, and much more fun can be had by decapsulating one of these flicker LED controller chips. It’s not very advanced tech; the LED controller is using a 1 or 2um process and a pair of RC oscillators, but it appears there could be a hardware random number generator in the silicon of this chip.

Earlier, [Cpldcpu] had taken a look at the tiny controller in these flickering LEDs and determined they used a linear feedback shift register to generate pseudorandom LED intensities. The new teardown seems to confirm that a linear feedback shift register is being used to drive the flickering LED.

Custom chips are only one way to skin a cat, or flicker a LED, and PICatout used the the tiniest PIC microcontroller (French, translation) to create his own flickering LED. Seems like making a few custom flickering LED throwies shouldn’t be too hard.

# Discussing Pulse-Width Modulation

[Michael Kleinigger] posted a lengthy discussion on Pulse-Width Modulation that goes beyond the traditional beginner tutorial. He starts a bit of background info on PWM and a tip about using a camera to judge frequency and duty cycle of LEDs. From there it’s down the rabbit hole with some testing of power-loss versus frequency.

When you change from frequencies of 50 Hz to 1 MHz how does the parasitic power loss from switching affect the overall efficiency of the circuit? It turns out there’s a rather large amount of loss at the highest level, around 1.5 mW. The greatest balance of low power loss and elimination of flicker seems to be right in the 300-500 Hz range.