You know the feeling. Table set, candles lit, and the room is wrong. Nothing is broken — the light is warm, the flame is there, it's even flickering. And it's still fake. You can't point at the specific defect. You just know. That knowledge is the most reliable lie detector you have, because your eye is better at catching artificial flicker than you are at explaining it.

Here's the mechanism, and it's unglamorous. The flame on a $4 candle is not a flame. It's an LED and a microcontroller running a brightness loop. The same sequence of bright and dim, repeated, at a fixed rate. The factory doesn't generate chaos; it generates a metronome, and it labels the metronome "realistic flicker."

Why your eye catches a flicker it can't name

A real candle flame is a turbulent flow. The flicker you perceive is a slow envelope — a drift in the 0.5 to 3 Hz range — sitting on top of micro-turbulence in the 50 to 100 Hz band that your eye can't resolve. That high-frequency jitter simply blurs into the slow envelope, and the result reads as something alive and unrepeatable. No two seconds of a real flame are ever identical, and the flame has a bias: it sways first, at the top, and the brightness follows a moment later.

An LED effect board has none of that. It outputs a loop — typically a handful of brightness steps cycling at a single rate, often right around 2 Hz. Single rate, fixed steps, exact repetition. Your visual system is terrible at naming that, but it's extremely good at pattern-matching it, because the pattern is, literally, a pattern. That's the "something is off" feeling. It's not a defect you can articulate. It's a statistical signature your brain rejects.

Two more tells the loop leaves behind. First, sync: candles on the same chip share the same loop, so switched on within a second they lock into phase, and the whole table pulses as one unit. A real dining table is a dozen independent flames. A dozen LED flames in unison is the fastest way to give it away. Second, amplitude: cheap loops swing from a dim minimum to a bright maximum on every cycle, so the flame breathes too hard, the same way, every cycle. Real flames have quiet moments and loud moments. A metronome doesn't.

The slow-motion test that exposes it in 30 seconds

Don't argue about it — film it. Point a phone camera at the LED candle, record at 120 fps if your phone has it, play it back at half speed next to a real flame shot the same way. At speed, the metronome is visible in two ways: the brightness steps are quantized (you see distinct plateaus, not a smooth drift), and the envelope repeats with a fixed period. A real flame, slowed down, has no period at all. If your LED's slow frame has a rhythm you can tap along to, that's the lie.

This is also how I'd buy: ask for or check the modes. The honest boards have one that isn't "steady" and isn't "flicker" — a randomized or "real flame" mode that changes its loop on power-up. If every mode on the board is a fixed cycle, you're buying a metronome in a candle sleeve.

If you're specifying the effect — the two-layer rule

The realistic profile is two layers, not one. A slow base drift, low amplitude, somewhere in the 0.5 to 3 Hz band, with a fast micro-variation on top. Roughly 70% of the perceived motion should come from the slow layer and 30% from the fast one. One-layer boards — a single modulation rate doing all the work — read as machinery at any rate. One rate that's too slow is a sine wave. One rate that's too fast is a strobe. The split is the whole trick, and it's cheap: it's firmware, not hardware.

Randomize per unit. Different loop lengths or a random phase offset at power-up is what kills the synchronized-table effect. It costs nothing in BOM and it's the single biggest realism win on a multi-candle install. And get the sequencing right: motion before brightness. A flame that brightens before it moves reads as a lamp. A flame that sways, then the light shifts, reads as fire. The order of the signals is part of the realism.

One number to sanity-check any spec sheet: if a board lists a flicker rate above 4 Hz as its "realistic" mode, that's not candlelight, that's a pulse. Candle-like motion is slow. If the spec says 20 Hz or 100 Hz, it's describing the carrier, not the effect — fine, but ask for the envelope rate. A board that can't tell you its envelope rate can't tell you anything about its flame.

Where to go from here

If you want the effect done properly, our Jumping Flame LED Candle with Wax Pillar is the board where the motion and the brightness are sequenced instead of stacked, and the flame layer moves before the light shifts — which is the exact order a real fire follows. For a table of many, the flameless flickering candles, 3 pack is the multi-unit install where per-candle variation matters most, and the steady-glow LED tea light pack of 12 is the honest answer when the room would rather have quiet light than a performance.

None of this is about the light level. A fake flame at the right brightness is still fake. Realism is a motion problem, and the motion is a loop, and loops are detectable. Your eye already knows. Now you can say why.

FAQ

Q: Why does my LED candle flame look fake even though it flickers?
A: Because the flicker is a fixed loop. A microcontroller runs the same brightness pattern over and over at a steady rate, and your eye reads repetition as machinery. A real flame is stochastic chaos: no two seconds of it are ever identical, and the slow drift on top of the fast micro-flicker is the signature your brain checks for.

Q: What flicker rate does a realistic LED candle flame actually need?
A: A slow base drift in the roughly 0.5 to 3 Hz band, layered with faster micro-variation. Below 1 Hz the flame moves like a slow sine wave and reads as a machine; above 4 Hz it starts strobing. Real candlelight reaches the eye as a slow envelope because the 50 to 100 Hz micro-turbulence of a physical flame is too fast to resolve and simply blurs into the envelope.

Q: Why do all my LED candles flicker in sync?
A: They share the same effect chip and the same loop. If you switch them on within a second of each other they lock into phase and the whole table pulses as one. The fix is per-unit randomisation: a different loop length or a random phase offset per candle, which is why quality effect boards randomise on power-up.

Q: Is LED flicker dangerous to photosensitive people?
A: Slow 1 to 3 Hz brightness modulation is generally far gentler than the mid-band strobe that triggers problems, but it is still a visible flicker. If photosensitivity is a concern in the space, choose steady mode, which every decent candle board has as one of its modes.