Get best of the week

LED driver with a BOM cost of less than $ 1. Is it possible?

LED driver development is an interesting and complex task. The market in this direction is very saturated - sometimes it seems that the production of LED lamps is everywhere. Starting from the garage and ending with huge factories. As for drivers, giants like Philips or Meanwell on the one hand, good-looking Chinese like Moso and Billion on the other, noname Chinese on the third ... Under these conditions, the task of optimizing the product for the price is added to the engineering components (circuitry and design).



So, I’m talking about the development of LED-drivers with a significant limitation on the price of components.

In my previous article, I conducted a small analysis of the requirements for LED equipment, as well as the regulatory documentation that describes these requirements. It's time to talk about the development. As you know, "without TK - the result is unpredictable", we’ll start from this.

TK requirements


  • Supply voltage 230 ± 10%
  • Power Consumption 15 W
  • Output voltage: 110 - 120V
  • Galvanic isolation: not required
  • Ripples of a light stream: no more than 5%
  • Luminous efficiency: not less than 100lm / W
  • Power factor: at least 0.9 (also work out the 0.5 option)
  • Design restrictions: Element height 14mm, maximum SMD (if possible).
  • The cost of components of the LED driver: no more than $ 1

Analysis of LED driver circuitry options


Consider the implementation options.

Option without KKM. In this case, since the output voltage is 110-120V, you need to make a buck converter. At the input of the buck converter, a rectifier and a storage capacitor, this will give a constant (ripple) voltage of approximately 310V. In order to make it more clear what is being discussed, hereinafter for each option I will give examples of microcircuits on which it is possible to implement the considered structure. Examples of microcircuits for a step-down converter without KKM: LM3444, HV9910B, HV9961, BP2831.

One-stage combined KKM / current stabilizer.An option when one stage of the conversion provides input sinusoidal current consumption and current stabilization of the LED line. Chip examples: TPS92074, BP2366, PT6917. There are exotic options like HV9931.

Line driver. An option when current stabilization is provided by dissipating part of the power on the regulator (by analogy with a linear stabilizer). Examples of NSI45090, FAN5640, PT6913, BCR402, BP5131 chips.

Two-stage: PFC + current stabilizer.The first stage is a boost KKM, after which a constant voltage of 380-400V is obtained on the storage capacitor. The second stage buck converter with current stabilization. Since this solution is usually used for more powerful LED drivers, a chip with a foreign key is usually used for the first stage (KKM), for example, NCP1650, UCC38051, LT1249.

One-stage with passive KKM like "Valley fill". KKM of this type is a fairly well-known scheme in narrow circles , I will write about it in more detail below.

Use an active filter.This is not an independent option, but an addition to one of the options, which allows to reduce ripple current, and, consequently, ripple of the light flux. An active filter can be implemented both on a field-effect and on a bipolar transistor. Example circuit:

There are also special microcircuits for this purpose, for example, BP5609, JW1210.

There are other options for constructing LED-drivers, for example, flyback converter or segmented linear driver, they are not considered, since they obviously do not fit the requirements of TK.

For ease of analysis, the pros and cons of the options considered are summarized in the table:

Driver typeprosMinuses
Option without KKM (buck)Not many elements cheaper, more compact.

In the presence of an electrolyte at the input, a low ripple of light can be provided.
Low PF.
(APFC-buck)➔ , .

PF.		20-30%.
➔ .

.

.		➔ , .

PF.
➔ .

PF.

.

TRIAC-.
➔ , .

100%.
(boost PFC+buck)PF.

.
Two converters ➔ more elements ➔ more expensive, more dimensions.
Single Stage + Valley fillVF corrector is simpler than full-fledged APFC ➔ cheaper, more compact than two-stage.The output voltage VF is pulsating, with a minimum value of Uin / 2.

It is difficult to get PF better than 0.9.

Not suitable for powerful solutions.
Active filter applicationAllows to reduce ripple of light.Dissipates power ➔ worsening efficiency.


A bit about the Valley fill corrector


Before analyzing and choosing options, you need to briefly explain what a Valley fill passive corrector is. The scheme looks like this:


Capacitors C1, C2 are charged each to half the voltage amplitude of the network. The essence of the circuit is that the capacitors C1, C2, using the diodes D1, D2, D3, are switched from serial (when charging) to parallel (when discharging to load) connection. As a result, the load is powered by the energy of capacitors only during periods when the value of the rectified mains voltage becomes less than half the amplitude value. Thus, the duration of current consumption from the network expands and the power factor increases. However, the circuit has a significant drawback - the output voltage has a significant ripple - up to half the voltage. This affects the choice of voltage of the LED line, it should be less than half the amplitude value of the input voltage plus a certain margin.

To clarify how PFC works, Valley-Fill made a spice model in LTspice:


The model is available here . You can download and experiment, see how it works.

Choice of LED driver structure


First you need to highlight the issue of galvanic isolation. The device (lamp) as a whole is a class II product for electrical safety. Why is galvanic isolation not required? If the device is in a plastic case without metal elements that can be touched by a person, then isolation is not necessary, since the protection is provided by the case. This can be seen in the example of LED lamps - drivers in LED lamps are never made galvanically isolated.

It is obvious that I had to abandon the two-stage option. Even if I can find microcircuits with built-in power keys for both stages (and for boost PFC in the first review stage (quickly) I could not find such microcircuits, except for some monster in the giant case from Power Integrations), then it's still there will be two power microcircuits and two chokes. Looking ahead, I’ll say that it is the throttle that adds a significant part to the cost of the BOM. The option turns out to be expensive, in addition, I figured out the layout and realized that it would not fit on a board of a given size.

Next, I threw back the linear drivers. Reason number one is the output voltage of 120V, which means that more than half the power will need to be dissipated on the linear gauge, this of course is not permissible. Even if we agreed on the increase in the voltage of the LED line (and I had such an opportunity), then for such capacities the linear driver is not very applicable. The large power dissipation in a compact plastic case will turn it into a heating device.

More precisely, it is possible to use a linear LED driver for such power, but only by sacrificing a power factor or a ripple coefficient of light, which I was not going to do. This is the second reason - it is impossible to achieve the specified characteristics either by PF or by KP of light.

As you remember from the requirements of TK, I need to offer two options: one without PFC, and the second with Pf at least 0.9. As a result of the analysis, the choice for the first option is obvious - this is a buck converter with output current stabilization. That is, the input filter ➔ rectifier ➔ large capacity electrolytic capacitor ➔ buck converter. This option is quite simple, and, in general, not so interesting to consider. Further I will consider only the option with KKM.

But for the second option, I came across a difficult choice: {Valley fill corrector + step-down converter} or {APFC-buck + active output filter}. Without the active filter, it would not be possible to get the given light pulsations - it was obvious to me.

There were such doubts. An active filter circuit is an additional power transistor, and, consequently, an increase in price, as well as additional losses, means a decrease in efficiency. In another option, I was confused if I could get the required Pf using the “Valley fill” scheme. On the one hand, in appnote from IR receive a power factor of up to 0.96, but there are also nuances. For example, I did not want to excessively increase the resistance of the resistor Rvf. In addition, there was a risk that there would not be enough voltage margin for normal buck regulation. Modeling showed that there is enough stock, but not the fact that it will also be in reality.

So, the option with the “Valley fill” corrector in my estimation made it possible to get a lower or the same price, with an increase in efficiency, this was decisive in my choice.

Circuit design


The diagram is shown in the figure:


Description of circuit elements:

FU1 - fuse, required by safety requirements;

RV1 - varistor for suppressing microsecond pulsed noise of high energy, as well as nanosecond pulsed interference;

R1, R2 - resistors for the discharge of the input capacitor when the device is disconnected from the network;

C1 - capacitor of the input interference filter (X2-class capacitor), suppresses conducted noise in the network, and also together with RV1 helps in the fight against impulse noise;

L1, L3, R3, R4 - elements of the input interference filter (conductive, pulsed);

VD1 - rectifier bridge;

C2, C3, VD2 ... VD4, R5 - elements of the “Valley fill” corrector;

C4 - input capacitor buck converter;

R6, R8 - resistors that provide power to the chip;

R7 - a resistor that sets the protection threshold for exceeding the output voltage (when the LED line breaks);

C5 - capacitor for microcircuit power;

DA1 - a step-down converter microcircuit with a built-in power MOSFET;

R9, R10 - current shunt resistors;

VD5 - power diode buck converter;

L3 - power converter buck converter;

C6 - output capacity.

Item selection


Fuse. I did not have to look long, the compact SMD fuse 25F-010H from Hollyland for only $ 0.048.

Varistor. Then I had to sweat. It seems now I know all the SMD varistor manufacturers in China and Taiwan. From what fits and delivered, I made a list and quoted such elements:


As a result, even Chinese prices were horrified, I had to abandon SMD in this case, and the choice fell on the TVR05391KSY varistor for $ 0.027.

I also considered European manufacturers, for example, Epcos has SMD varistors, but even more expensive, unfortunately.

X capacitor. SMD options for such capacitors are very expensive, so 0.1 μF 10% 300V X2 J104K300A100 from the world's leading manufacturer Chiefcon is the best choice for $ 0.036.

Electrolytic capacitors.The choice of SMD electrolytes at 200V is not so great, and those that are turned out to be of gigantic size. Considered the series "VE", "VEJ" from Lelon, "ULR", "UUG", "UUJ" from Nichicon, etc. 10 microfarads in the size of 12.5x13.5 did not suit me. As a result, I came across an interesting Chinese manufacturer Ymin, whose website says “Small expert”. Indeed, the VKM series, 12 microfarads in the size of 8x12.5, is an excellent option and for only $ 0.046 per piece. Give two.

Diodes.The MB6S diode bridge (0.028), the power diode chose the ES1J type - everything is standard here, but I wanted to choose the smaller diodes for Valley fill and I found a very interesting version of PANJIT's GS10xxFL. Diodes for voltages up to 1000V in the SOD-123 package, are you kidding? No, they exist. As a result, the GS1006FL is only $ 0.019. Find the same ultrafast and can be used as a power diode for lowered. I left this idea before conducting thermal tests. If ES1J does not warm up, then you can think about it.

Chokes.Initially, I turned to Eurobrands, but quoting showed that the most inexpensive of what suits me is SRR1208 for $ 0.28 from Bourns. Even Wurth has not been offered cheaper than about $ 0.3. This is a power choke. As a result, I turned the search vector towards Asian brands. After reviewing and quoting the products of such offices as Ferriwo, ABC Taiwan, Fuantronics, Coilmaster, I settled on the SRI1207 variant from the Taiwanese company Coremaster. For only $ 0.142.

Chip.I decided to opt for BPS, because they have a large portfolio in the field of microcircuits for LED drivers, I met their products in many devices, and there is also a distributor in Russia - Platan. I chose a chip with a built-in power switch BP2832AJ - it has pin-2-pin compatible "older brothers" (or "sisters", forgive my gender ignorance) BP2833 and BP2836 with lower channel resistance of the built-in MOSFET (if you suddenly need to increase power or get more efficiency). In addition, this chip can be bought in the Russian Federation.

Final BOM:

$ 0.81 is a good result for the first version. There is a small margin - because you know how it happens, after testing the first iteration of the board, some kind of ... mismatch usually comes out, and you have to add a few "magic" elements of three bucks each.

Conclusion


How to reduce conducted noise without adding new elements to the circuit? Why can't thin-film resistors be used in the Valley fill corrector? How to increase the luminous efficiency of the lamp without increasing the efficiency of the LED driver? You will learn the answers to these and other questions from the second part of this article. The article will be devoted to the tests that the author, along with the first iteration of the board, will have to pass, as well as the development of the second iteration of the device.

Power is cool - deal with it.

More articles:


All Articles