Simplify, simplify: TI’s AC LED driver eliminates magnetics

Switched mode power supplies (SMPS)  do a very good job of controlling the input power to LED lights. So it’s not their performance that  can cause anguish for LED lamp designers – its the need for bulky magnetics, which can be tricky to design and expensive to implement. In addition, SMPSs have trouble with legacy TRIAC and phase-cut dimmer switches, because these switches expect to see a purely resistive load, such as an incandescent light, rather than the SMPS’s wonky reactive load.

And perhaps the biggest problem of all for SMPS’s is that lamp designers don’t often have the design chops needed to tweak such complex electronic circuits.

Texas Instruments believes it has come up with the perfect solution for the problem of driving dimmable LED lighting over fixed voltage ranges with the floating switch architecture used in its new TPS92411. The TPS92411 replaces flyback, buck and boost converters’ SMPS architectures with a very simple AC switched matrix technique. (For more on the evolution of AC drivers read AC LEDs, no; AC drivers for LEDs, yes.)

This architecture requires no inductors for energy storage, and its non-isolated architecture does away with an isolation transformer, making the design magnetics-free.

Another advantage of eliminating magnetics is that since magnetics generally require through-hole components, and thru-hole is difficult to implement on metal-core boards, the elimination of magnetics frees the pc board designer to use the metal-core board technology generally required by LEDs, and move towards all surface-mounted devices – a cheaper high-volume production method.

Here’s a simplified schematic of how the switches work. Each bank of LEDs has differing numbers of LEDs, depending on the forward voltage needed for the bank of LEDs.


TMS92411 simplified

The current regulator indicated at the base is a linear circuit that uses  about a dozen common (non-magnetic) components. (The overall efficiency of the driver is about 84%, in line with common non-isolated SMPS drivers.)

Note that there is a diode at the top of each bank of LEDs. These diodes allow the switches to close without discharging the capacitors, which allows continuous current flow in the LEDs. Thus the switches aren’t cutting out the LEDs, but rather switching out which bank is seeing the rectified voltage, and which bank is running off its storage capacitor. When all switches are closed (at zero volts), all LEDs will be powered by their storage capacitors. When all switches are open (which will occur at the peak 170V), the current flows through all diodes and LEDs. The three switches will open and close throughout the rectified cycle based on their set points, switching each bank’s capacitor in or out. Because the switches are floating, they require no communication. The switch trip points are set by two external resistors.

This architecture also plays nicely with legacy dimmer switches, which see the LED/switch load as purely resistive, similar to an incandescent light.

What about the size of the capacitors needed to support the LEDs  when their switch is closed? In a typical 12W design, (shown below) the three capacitors will be 68uF over the 80V stack, 135 uF over the 40V stack, and 270uF over the 20V stack. If you want to play with these values yourself, you can download the TPS92411 design calculator spreadsheet tool here: 120V version.

TPS92411 Reference Design

Texas Instruments TMS92411 AC driver swtich

Each 92411 chip has a 100-V, 2-Ohm floating MOSFET switch with 350 mA of current capability. It supports lighting designs of up to 70W, and provides >.95 power factor with less than 20 percent of total harmonic distortion. Its low-frequency, slew-rate controlled switch action produces very little EMI noise.

12W LED reference design for TMS92411 AC driver, using Cree XLamp ML-E 10V LEDs

12W LED reference design for TMS92411 AC driver, using Cree XLamp ML-E 10V LEDs

Above is a 12W reference design using Cree XLamp ML-E 10V LEDs. You can see the relatively small caps used.

Available now in volume from TI and its authorized distributors, the TPS92411 is offered in a 5-pin SOT-23 package priced at US$0.23 each in 1,000-unit quantities. An evaluation module is offered at $75.00. The TPS92411 in an 8-pin PowerPad SOIC package will be available in the first quarter of 2014.

Comments

  1. Yves Laurin says:

    This is very good news as some people are electrosensitive (they can have headache or nausea with CFL and other equipment that generate electromagnetic field).

    of course “and 270uH over the 20V stack” is a typo and should say “and 270uF over the 20V stack”, we do not want any inductor in there.

  2. Yves, thanks for catching my typo – fixed.

  3. Toby Ovod-Everett says:

    Is it just me, or do others see “MW” in that schematic? I assume that’s omega rendering as a W due to some font issue, but if those resistors really need to be able to dissipate as much as 2MW, there’s going to be a huge amount of heat coming out of the module!

  4. Toby – Yes, those “Ws” should be ohms. I took the diagram from the .xls tool to show the capacitor sizes. Here’s a link to a .pdf of the schematic with the ohm symbol: http://www.ti.com/general/docs/datasheetdiagram.tsp?genericPartNumber=TPS92411&diagramId=SLUSBQ6

  5. Toby Ovod-Everett says:

    Interesting – it also mutates the capacitor sizes – from microFarads (with a lower-case mu) to what looks like milliFarads (the m is overlaying the F slightly)!

  6. I wonder if you can get capacitors in a profile shape of a rabbit.

    If you’re going to make shadows on the walls from your LEDs, as would be the case with that reference design, it might as well be an interesting shape.

  7. What efficacy range does this design provide? Thanks!

  8. LEDguy, I think you’re asking for the lumens/Watt of the schematic pictured above. I’m checking…

  9. Thank you Margery, My understanding is that the AC technologies while improving, are still off for part of each half wave cycle and therefore lag behind somewhat in efficacy/LpW compared to DC (always on) topologies. It is very interesting to see how close they are able to get.

  10. LEDguy, I’m realizing I shouldn’t have called this an AC driver – not because it isn’t (it is, after, an architecture that takes in a rectified ac line voltage and drives LEDs in a pretty straightforward way) but because “AC driver” has so much baggage associated with it. And part of that baggage is the idea that the strings of LEDs powered by an AC driver have to be turned off for certain periods. This is not the case with the 92411– the LEDs are on all the time . Their power just comes from capacitors for certain parts of the input voltage cycle. I should have stuck with “switched linear” for describing the architecture.

    Anyway, I asked John Parry of TI for the efficacy of the design (the one at the bottom of the post) and he said:
    “We are using Cree’s ML-E Series LEDs — which are not their newest — and thus the raw lm/W is not going to be as high compared with using the same circuit with newer, higher lm/W LEDs.  In addition, we were trying for a smallish form-factor for our board and did not want to use 40+ 3V LEDs.  The ML-Es were a nice tradeoff for this reason.  Using the Cree PCT tool for the Series connected ML-E (M2 bin) parts we use, the raw LED lm/W is ~75 at the current drive we provide them ~70mA.  If our electrical efficiency is ~84%, then the LPW of the EVM is ~63lm/W.” (Here’s the link to the Cree selector tool referenced: http://pct.cree.com/dt/index.html )

  11. Did it have the 120Hz flicker noise? Thanks!

  12. James, I got the evaluation kit from TI and hooked up a photocell to look at the ac flicker. It looks virtually identical to the Cree bulb which has a conventional dc-dc power management section. I’ll write about this soon – thanks for the question.

Speak Your Mind

*