A recent post covered how ac-drivers can eliminate magnetics in an LED driver. What about the other component bogey-man, the electrolytic capacitor? To decide if you can junk the electrolytic capacitors in your ac LED driver design, you’ll need to come to terms with flicker.
I recently watched the (excellent) webinar, AC LEDs: Perception vs. Reality, put on by LEDs magazine and presented by Seoul Semiconductor. Seoul Semi was one of the first LED manufactures to offer ac LEDs. Its first generation of ac LEDs, the Acriche product line, was a packaged LED consisting of two strings of LEDs in parallel with opposite polarities; The LED themselves rectified the ac line voltage. (For more on the origins of ac LEDs, read AC LEDs, no; AC drivers for LEDs, yes.) Seoul learned from these first ac LEDs, and moved on to separating the driver electronics, including the ac rectifier, from the packaged LEDs.
Seoul’s new ac driver module uses its Acrich2 architecture that relies on banks of LEDs, and is sometimes referred to as Tapped Linear: “Tapped” because the LEDs are configured in serial with tapped points available to select banks of multiple LEDs at a time. (Fig. 1.) As the rectified ac input voltage increases, each bank of LEDs turns on, one after the other, until at the peak voltage, all LEDs are turned on and emitting light. As the voltage then falls, the LEDs turn off in reverse order. Fig. 2 shows this stepped turn-on/off of the LEDs’ light output.
This approach is quite simple, using very few external driver components. Most notable: It eliminates electrolytic capacitors, which are a notorious weak spot (but not necessarily the weakest) in LED driver design.
If your application is a non-moving light source, then Seoul Semi’s tapped linear approach has a lot going for it, with its minimal drive components and elimination of e-caps.
However, there are two potential problems with this approach: The first is that LEDs are not cheap, and underutilizing them is not a good way to create cost-effective lumens/watt. On the other hand, Seoul Semi as an LED manufacturer probably doesn’t get too fussed about selling additional LEDs. On the other other hand LEDs are becoming so efficient that we can afford to waste a little LED real estate, as evidenced by this tweet quoting Pacific Northwest National Lab:
— LEDs Magazine (@ledsmagazine) September 9, 2013
The second potential problem with Seoul Semi’s tapped linear approach is that turning LEDs on and off can be visually distracting and can physically affect some people. And this brings up the subject of flicker.
Flicker and artificial lighting go way back: All forms of electrical light – including older technologies such as incandescent and fluorescent – have some degree of variation in their intensity. An LED’s light is quite stable when powered by a stable current source. However, most currently available LED bulbs use a pulse-width-modulated (PWM) dc-dc converter which can cause a periodic variation in the light. Couple it with a phase-cut dimmer switch and the light may flicker and even flash.
There are two flicker metrics. (See Fig 3.) The older, simpler number, “percent flicker” is helpful in quantifying flicker in light sources that vary periodically, like you’d find in a fluorescent light – similar to what you’ll see when the tube is near the end of its life. Percent flicker is the difference between the maximum and minimum light amplitudes dived by their totals and multiplied by 100.
In LED lights where the maximum intensity of the light may vary – such as in a light that aperiodically flashes – the flicker index is a better metric because it’s a ratio of the areas under the curve of the light amplitude over time. For the algorithms for percentage flicker and flicker index, see the Fig. 3.
How much flicker is allowable? There is a wide variation in people’s sensitivity to flicker, ranging from headaches to fatigue and even seizures, which can affect people with photo sensitive epilepsy at frequencies below 15 Hz.
There is currently no standardized number for flicker index. Traditional electric light sources, including magnetically ballasted fluorescent lights, show a maximum percent flicker of about 40% and a maximum flicker index of about .15. As a starting point, looking at a flicker index of .15 for LEDs is probably reasonable since consumers have been coping with that level in conventional lighting. (For more on the current state of flicker research and metrics, read the excellent DOE paper, Flicker Fact Sheet.)
First-generation packaged ac LEDs such as the Seoul Semi Acriche, with their two parallel strings of LEDs of opposite polarity always have half of their LEDs off. This makes for a high flicker index at about .41. The flicker is not noticeable when stationary, but very noticeable with movement. Below is a photo of an Acriche mounted on a heatsink and connected to an ac power outlet: You can almost see my hand (faintly at the top, kinda orangey) as I wave it back and forth, and the LEDs’ stroboscopic flicker.
With the Acrich2′s tapped linear design, the stepped approach increases the duty cycle of light output compared to the original ac LED approach. This lowers the flicker index from the .41 of the ac LEDs to as low as .23 based on wave shape and using the valley fill approach which requires the addition of ceramic capacitors. While .23 flicker index may be acceptable for some stationary applications, it’s above the .15 of traditional lighting.
Which brings us to the switched linear approach TI uses in its new TPS92411 driver. The LEDs are also driven directly by rectified ac line voltage but the architecture uses the LEDs more efficiently by powering them all on, all the time. It accomplishes this by using energy stored in electrolytic capacitors.
The reference design achieves a flicker index of .15, good enough to match traditional electric light sources. But note the use of electroytics, albeit very small ones. To get the best performance, capacitance is needed and electrolytic capacitors are the most economical way to store a lot of energy. They have a bad reputation for being a weak point, but the tradeoffs often make their use worthwhile. (And, as an aside, I have never done an LED light tear-down where I didn’t find electrolytics.)
To summarize, eliminating electrolytic caps comes at the cost of greater flicker or a more complex/expensive driver.
As with most electronic designs, the best circuit to use is determined by your design constraints. If your design is for a stationary, price-sensitive bulb which will probably not be Energy Star compliant, then the tapped linear driver using banks of LEDs may be a good approach.
Here’s an example of a product that fits that description: The new Walmart Great Value line of 60/40W bulbs. Yes, they have some problems with flicker and flash while dimming, but Walmart is not looking at these bulbs to get an Energy Star rating – they just want a basic LED bulb to sell at a low-price.
On the other hand, if the LED lamp application requires a performance in flicker and dimming, then the switched linear — electrolytics and all — may be a better bet.