The Truth About 'Instant-On' LEDs: Why Some 120V AC Drivers Still Have 0.3-Second Delay (and How to Test It)
Calling an LED “instant-on” is like calling a toaster “instant-bread.” Technically true — it doesn’t need to warm up like a halogen filament — but wildly misleading if you’ve ever flipped a switch in a dark hallway at 2 a.m. and waited, blinking, while your bathroom light *decided* whether to commit.
I learned this the hard way after installing six Feit Electric 9W A19 bulbs in my stairwell — marketed as “instant-on,” “dimmable,” “energy-efficient,” and “just like incandescent.” They were none of those things when it came to response time. One night, I fumbled down the stairs in near-total darkness for 300 milliseconds — long enough to question my life choices, short enough that no one believed me when I said “the bulb *hesitated*.”
So I grabbed my oscilloscope. And yes — I own one. No, I’m not an EE. Yes, I bought it on eBay for $85 after watching three YouTube videos and swearing I’d “only use it for lighting stuff.” Turns out, “lighting stuff” includes proving that “instant” is just marketing’s version of “we didn’t measure it.”
How ‘Instant-On’ Got Its Name (and Lost Its Meaning)
In the early 2000s, LED bulbs were clunky, expensive, and took half a second to glow — if they glowed at all. Early drivers used basic rectified AC with minimal filtering. The LED would flicker violently or turn on in stuttery bursts. Then came soft-start circuits: simple RC networks or microcontroller-based ramps designed to limit inrush current and extend capacitor life. These worked — too well. A typical soft-start would ramp current over 200–400 ms. That’s not delay. That’s a *fade-in*. Like a stagehand slowly lifting a curtain.
By 2012, manufacturers started touting “instant-on” as a competitive differentiator — even though most bulbs still used the same slow-ramp drivers. The claim wasn’t false per se; LEDs *do* emit photons faster than any human can perceive — but only once the driver says “go.” And many drivers? They say “go… in a moment.”
I pulled apart five budget bulbs from 2015–2018. Every single one had a soft-start circuit built into its constant-current buck driver — usually a 10 µF electrolytic cap charging through a 100 kΩ resistor. Do the math: τ = RC = 1 second. That’s the *time constant*, not the full ramp — but even hitting 63% of final current takes ~1 second. Most manufacturers throttle it earlier, landing in that suspiciously consistent 250–350 ms window.
Why 0.3 Seconds Feels Like Forever (Especially in Context)
Human visual persistence is about 1/16th of a second (~60 ms). Anything longer than that registers as lag — especially when contrast is high (dark room → bright light) and expectation is immediate (you flipped the switch *now*).
Let’s ground this in real space:
- A standard stairwell is ~2.4 m tall. At average descent speed (~0.5 m/s), you cover ~15 cm in 300 ms — enough to misstep on step #2.
- In a closet or pantry, where light = safety, 300 ms is the difference between spotting the shelf edge and bonking your temple.
- In a bathroom at night? That’s the gap between “I think I’m alone” and “oh god why is there a shadow behind me.”
This isn’t theoretical. I timed it — not with gear, but with my dumbest tool: my phone’s slow-mo camera. 240 fps means each frame is ~4.17 ms. At that rate, 300 ms = 72 frames. I filmed six different bulbs switching on. The Feit Electric 9W A19 lit at frame 68. The Philips SlimStyle? Frame 12. The Bridgelux EB Gen 4 integrated module (yes, I wired it bare into a junction box — don’t try this at home)? Frame 4.
That’s not semantics. That’s physiology meeting engineering.
What’s Really Happening Inside That Bulb
Here’s the brutal truth: “instant-on” depends entirely on the driver topology — not the LED chip. You could mount a $0.02 2835 diode on a premium driver and get sub-20 ms latency. Mount a $2 Bridgelux chip on a $0.35 Chinese driver IC with soft-start enabled? You’ll wait.
Most budget bulbs use a single-stage buck converter with integrated PWM controller — often variants of the BP5132 or SY5823. These chips include configurable soft-start by default. Why? Because electrolytic capacitors hate surge current. A cold start can draw >2 A peak on 120V AC. Without soft-start, the input cap fails in ~5,000 cycles. With soft-start? 25,000+. So yes — it’s a reliability feature. But it’s also a UX tax.
Premium drivers skip soft-start entirely — or replace it with active inrush limiting (e.g., NTC thermistors + MOSFET bypass). The Bridgelux EB Gen 4 driver uses a dual-stage architecture: a fast AC-DC front-end followed by a tightly regulated DC-DC stage with <5 µs gate drive latency. Its turn-on is limited only by propagation delay in silicon — not RC timing.
I measured it: 17.3 ms from zero-crossing of AC input to stable 350 mA output (±2%). No ramp. No hesitation. Just *on*. Like flipping a mechanical switch — because at the semiconductor level, that’s what’s happening.
How to Test Turn-On Latency Yourself (No Oscilloscope Required)
You don’t need lab gear. You need:
- A smartphone with slow-motion video (iPhone 8+, Samsung Galaxy S9+, Pixel 4+)
- A completely dark room (blackout curtains help)
- A mechanical toggle switch (no dimmers, no smart switches — they add their own latency)
- A tripod or steady surface (your palm won’t cut it)
Step-by-step:
- Set phone to slow-mo mode (240 fps minimum — avoid 120 fps; too coarse).
- Mount phone so lens points directly at bulb’s center. Fill frame with light-emitting surface.
- Turn off all ambient light. Wait 30 seconds for eyes to fully adapt — and for capacitors to discharge.
- Start recording.
- Flip switch.
- Stop recording after 1 second.
- Scroll frame-by-frame. Look for first visible pixel illumination — not “full brightness,” but first detectable gray-to-yellow transition.
Count frames from switch flip (you’ll see mechanical movement in the video) to first light. Multiply by frame interval (e.g., 240 fps = 4.166 ms/frame). That’s your latency.
I tested ten bulbs this way. Results:
| Bulb Model | Type / Driver IC | Measured Latency (ms) | Notes |
|---|---|---|---|
| Feit Electric 9W A19 | Buck, BP5132A | 312 | Smooth ramp; no flicker |
| Philips 9.5W SlimStyle | Buck-boost, FL7760 | 118 | Noticeable “pop” at onset |
| Cree 10W BR30 | Active PFC + buck, LM3445 | 89 | Faint pre-glow (LED leakage) |
| Sylvania Ultra LED 12W A19 | Single-stage buck, OB2500 | 294 | Identical timing curve to Feit |
| Bridgelux EB Gen 4 (bare module) | Dual-stage, custom ASIC | 17.3 | No visible ramp; square-wave onset |
Fun fact: Two bulbs with identical lumen output (800 lm), CCT (2700K), and CRI (>90) differed by 295 ms in response time. That’s not “performance variation.” That’s “one team optimized for cost, the other for spec sheet integrity.”
Soft-Start Isn’t Evil — But It Should Be Optional
Let’s be fair: soft-start prevents early failures. In my teardowns, every bulb with soft-start had lower ESR input caps and cleaner ripple. Those without tended toward higher heat stress on primary-side MOSFETs. Reliability trade-offs are real.
But here’s what pisses me off: no bulb advertises its soft-start duration. None list turn-on latency in specs. You have to dig into datasheets — if they exist — or reverse-engineer the driver. Meanwhile, packaging screams “INSTANT ON!” next to a lightning bolt icon.
This works because most people never test it. They install, flip, and accept the delay as “normal.” I think that’s lazy engineering masked as user tolerance.
The better path? Make soft-start *configurable*. Some industrial drivers (e.g., Mean Well HLG series) offer dip-switches or jumper options to disable soft-start. Consumer bulbs could do the same — a tiny solder bridge on the PCB, or even a firmware toggle via smart-switch pairing. Not every homeowner needs 17 ms. But every homeowner deserves to *choose*.
What Actually Matters (Beyond the Number)
Latency isn’t just about speed — it’s about predictability.
I tested the same Feit bulb across 100 cycles. Latency varied from 291 ms to 337 ms — depending on line voltage (114 V vs. 126 V), ambient temp (18°C vs. 32°C), and capacitor aging. That inconsistency is worse than raw slowness. Your brain adapts to fixed delays. It stumbles on variable ones.
Compare that to the Bridgelux module: 17.1–17.5 ms across all conditions. Same rise time. Same overshoot (<1%). Same recovery from brownouts. That’s not just fast — it’s *trustworthy*.
And trust matters. When your hallway light responds like muscle memory — not a buffering video — you stop thinking about the light. You just move.
Where to Find Truly Instant-On Lighting (Right Now)
If you’re retrofitting sockets: avoid anything under $10/pack. Not because cheap = bad — but because $1.20 per bulb leaves ~$0.18 for driver BOM. That buys you a BP5132 and prayer.
Look instead for:
- Architectural-grade modules — e.g., Soraa Radia or Acuity EnFocus. These use multi-stage drivers with active inrush control. Typical latency: 12–22 ms. Downside: you’ll need a licensed electrician to wire them.
- Smart bulbs with local control — e.g., Lutron Caséta LED + PD-6WCL. Their drivers bypass cloud latency and use dedicated RF wake-up. Measured: 42 ms from wall switch event to light output. Still not “instant,” but feels like it.
- Linear fixtures with external drivers — e.g., a 4-ft T8 LED troffer with Philips Advance Xitanium driver. Set soft-start to “off” in dip-switch bank. Result: 24 ms. Bonus: you can hear the relay click *before* the light hits your retina.
And if you’re stuck with budget bulbs? There’s one hack that actually works: use a mechanical relay *ahead* of the bulb. Not a smart switch — a real 120V coil, SPST, 10A contact relay. I installed one feeding four Feit bulbs. Latency dropped from 312 ms to 289 ms — not huge, but perceptible. Why? Because relays eliminate contact bounce delay in plastic toggle switches. Turns out, your $2 wall switch is slower than your $2 bulb driver.
I’m not saying go full mad scientist. But if you’ve ever stood in your dark kitchen muttering “just TURN ON already,” know this: it’s not you. It’s the capacitor charging. And it’s fixable.
Next time you buy LEDs, skip the “instant-on” badge. Flip to page 3 of the spec sheet. Look for “turn-on time” — or better yet, “output rise time.” If it’s missing? Assume 300 ms. And maybe keep a flashlight in the drawer.