Engineer sharing: the whole process of making LED driver power supply

Design and manufacturing method of LED fluorescent lamp power supply

I am engaged in switching power supply. I used to make adapters, chargers, and iron-shell switching power supplies. Later, I made LED power supplies. At first, I made some 1W and 3W high-power LED drivers, but later I made less. The reason is very simple, there is no market. I found that high-power LED constant current power supplies, as long as their power exceeds 5W, basically have no market and can only be used for proofing. Because LED is too expensive. This is also a reminder to my friends who are engaged in power supply. This is my experience. I don’t know how many people have fallen into high-power LEDs. High-power LEDs are thunderous but small, and many people have lost their original capital in this area. The low-power LED market is better. But it’s not good. Now the small-power LED drivers are mostly occupied by RC step-down power supplies. Constant current switching power supplies drive small-power LEDs, which is good, but many people cannot accept its cost. I have produced a constant current low-power LED driver, a switching power supply, with an efficiency of 0.9, stability, reliability, and constant current accuracy. The price is only five yuan, but many people still think it is too expensive because they compare it with a one-yuan RC step-down power supply. Of course, the two are not comparable. In the switching power supply I made, there is an integrated MOS switching power supply chip and a transformer. The cost of these two is there, and of course the performance is there. But I believe that in the end, the low-power LED constant current driver will eliminate the RC step-down power supply. Because consumers will gradually become rational, a lamp made by a RC step-down power supply has almost no practical value and can only be used as a decoration and toy. If LED really enters the field of general lighting, the RC step-down power supply will not be able to do it. I can expect that in the future, as the performance of LEDs improves and the price decreases, the power supply cost will also become a very important part of the cost of LED lamps. For real lamps, RC step-down is not enough. The RC step-down power supply is popular, but it is just a transition. In the end, the constant current power supply is authentic.

I am still optimistic about low-power LED lamps. The main problem with low-power LED lamps is that the light decay is too large and the price is not ideal. But they are still better than high-power lamps for general lighting. I think it will take less than five years for low-power LED lamps to enter the field of general lighting and compete with energy-saving lamps. It will definitely take more than five years for high-power LEDs to enter general lighting. So now I am focusing on the research and development and production of low-power LEDs. I noticed that the main low-power LED lamps used in general lighting are LED desk lamps, LED honeycomb lamps, and LED fluorescent lamps. Especially LED fluorescent lamps, since the second half of 2007, many people have started to develop them, and it can be said that they are very popular. Basically, eight out of ten people who come to me now are doing this, so I also started to make power supplies for LED fluorescent lamps. I have been doing it for a while, so I will talk about the general methods and principles of the research and development and production of this kind of power supply. This is my personal experience.
About the appearance

Now, LED fluorescent lamp power supply is generally required by lamp manufacturers to be placed inside the lamp tube, such as T8 lamp tube. A small part is placed outside. I don't know why it is so. In fact, internal power supply is difficult to make and the performance is not good. But I don't know why so many people still require it. Maybe it's just following the trend. External power supply should be more scientific and more convenient. But I have to follow the trend. I will do whatever the customer wants. But making internal power supply is not easy.

It is quite difficult to install a power supply. Because there is basically no requirement for the shape of an external power supply, you can make it as big as you want and in any shape you want. There are only two types of built-in power supplies. One is the most commonly used one, which is placed under the light board, with the light board on top and the power supply below. This requires the power supply to be very thin, otherwise it cannot be installed. And this can only turn the components upside down, and the lines on the power supply can only be lengthened. I think this is not a good way. But everyone generally likes to do it this way. I will do it. Another one is less used, placed at both ends, that is, placed at both ends of the lamp tube, which is easier to make and has a lower cost. I have also made it, and it is basically these two built-in shapes.

Questions about the requirements and circuit structure of this power supply

My opinion is that because the power supply must be built into the lamp, and heat is the biggest killer of LED light decay, the heat must be small, that is, the efficiency must be high. Of course, a high-efficiency power supply is required. For a T8 lamp that is 1.2 meters long, it is best not to use one power supply, but two, one at each end, to disperse the heat. So that the heat is not concentrated in one place. The efficiency of the power supply mainly depends on the structure of the circuit and the components used. Let's talk about the circuit structure first. Some people also say that the power supply should be isolated. I think it is absolutely unnecessary because this kind of thing is originally placed inside the lamp body and people can't touch it at all. There is no need to isolate it because the efficiency of the isolated power supply is lower than that of the non-isolated one. Secondly, it is best to output high voltage and low current. Only in this way can the power supply have high efficiency. The commonly used one now is the BUCK circuit, that is, the step-down circuit. It is best to make the output voltage above 100 volts and the current set at 100MA. For example, if you drive 120, it is best to have three strings, each with 40, and the voltage is 130 volts and the current is 60MA. This type of power supply is used a lot. I just think there is one disadvantage. If the switch tube is out of control, the LED will be finished. LED is so expensive now. I am more optimistic about the boost circuit. I have repeatedly said the advantages of this circuit. First, the efficiency is higher than the buck type. Second, if the power supply is broken, the LED lamp will not be broken. This can ensure that everything is foolproof. If a power supply is burned out, it will only lose a few dollars, but if an LED fluorescent lamp is burned out, it will lose hundreds of dollars. So I always recommend the boost power supply first. In addition, the boost circuit can easily increase the PF value, while the buck type is more troublesome. I am sure that the benefits of the boost circuit for LED fluorescent lamps are overwhelmingly stronger than the buck type. There is only one disadvantage, that is, under the condition of 220V AC input, the load range is relatively narrow. Generally, it can only be applied to 100 to 140 strings of LEDs or two strings. For less than this number or sandwiched in the middle, it is inconvenient to use. However, the current LED fluorescent lamps generally use 100 to 140 for the 60cm long ones, and 200 to 260 for the 1.2m ones, which are still OK to use. Therefore, the current LED fluorescent lamps generally use non-isolated buck circuits and non-isolated boost circuits. This kind of circuit used in LED fluorescent lamps can be regarded as my first invention.

Opinions on high PF LED fluorescent lamp power supply and high current LED fluorescent lamp power supply:

Personally, I think these practices are often just putting the cart before the horse. Now let me first ask what are the advantages of LED over traditional lamps. First, energy saving, second, long life, and third, no need to worry about switching, right? However, the high PF method currently used is to use a passive valley-fill PF circuit, changing the original driving method, i.e. 48 series and 6 parallel, to 24 series and 12 parallel. In this way, under 220V, the efficiency will drop by about five percentage points, so the LED fluorescent lamp power supply will generate more heat, and the lamp beads will also be slightly affected.

There is another problem, that is, the 24-in-12-in-parallel method will make the wiring of the LED fluorescent lamp beads very uncomfortable and difficult to wire. In my opinion, the best method is still the 48-in-1 method, mainly because it has high efficiency, low heat generation, and easy wiring without complexity.

What's more, some people are now proposing 24 parallel and 12 series. This method is only suitable for isolated power supply, not for non-isolated power supply. Some people who don't know the common sense of power supply think that it's great if their non-isolated power supply can output 600MA constant current. In fact, they haven't tried it carefully in the lamp tube. It would be strange if it didn't overheat.

Therefore, it is a waste of effort to develop low voltage and high current LED fluorescent lamp power supply

.

What I want to talk about next is the two constant current control modes of the switching power supply, which leads to two approaches. These two approaches are quite different in principle, device application, and performance.
First of all, the principle. The first is represented by the current constant current LED dedicated IC, mainly such as the 9910 series, AMC7150, and all the brands of LED constant current driver ICs are basically of this type, and they are called constant current IC types. But I think that this so-called constant current IC is not very effective for constant current. Its control principle is relatively simple, that is, in the primary circuit of the power supply, a current threshold is set. When the primary MOS is turned on, the current of the inductor rises linearly. When it rises to a certain value, it reaches this threshold and turns off the current. The next cycle is triggered by the trigger circuit to turn on. In fact, this kind of constant current should be a kind of current limiting. We know that when the inductance is different, the shape of the primary current is different. Although there is the same peak value, the average current value is different. Therefore, when such a power supply is generally mass-produced, the consistency of the constant current size is not easy to control. Another feature of this type of power supply is that the output current is generally trapezoidal, that is, a fluctuating current. The output is generally not electrolytically smoothed, which is also a problem. If the current peak is too large, it will affect the LED. If the output stage of the power supply does not use electrolysis to smooth the current, it basically belongs to this category. That is, to determine whether it is this control method, it depends on whether the output is electrolytically filtered. I used to call this constant current a false constant current, because its essence is a kind of current limiting, not a constant current value obtained by comparison with an op amp.

The second constant current method should be called a switching power supply. This control method is similar to the constant voltage control method of the switching power supply. Everyone knows that TL431 is used for constant voltage, because it has a 2.5-volt reference inside, and then uses a resistor voltage divider. When the output voltage is higher or lower, a comparison voltage is generated, which is amplified to control the PWM signal, so this control method can be very

Precise control voltage. This control method requires a reference and an op amp. If the reference is accurate enough and the op amp amplification factor is large enough, then the setting is very accurate. Similarly, to make a constant current, you need a constant current reference and an op amp. Use resistor overcurrent detection as a signal, and then use this signal to amplify and control PWM. Unfortunately, it is not easy to find a very accurate reference signal now. Commonly used transistors have large temperature drift as a reference. Another method is to use the diode's conduction value of about 1V as a reference. This is also possible, but not high. The best is to use an op amp plus TL431 as a reference, but the circuit is complicated. However, the constant current power supply made in this way is much easier to control. And the output of the constant current controlled by this mode must be electrolytically filtered, so the output power is smooth DC, not pulsating. If it pulsates, it cannot be sampled. So to determine which one it is, just look at whether its output has electrolysis.

The two constant current control modes determine the use of two different types of devices, which in turn determines that the two circuit devices are used differently, have different performances, and have different costs.

The LED power supply made by the constant current control IC represented by the 9910 series is actually current limiting and has a relatively simple control. Strictly speaking, it does not belong to the mainstream mode of switching power supply control. The mainstream mode of switching power supply control must have a reference and an op amp. However, this kind of IC can only be used for LEDs and is difficult to use for other things, just because LEDs have extremely low requirements for ripple. But because it is only used for LEDs, the price is relatively high now. Basically, it is made by using 9910 plus MOS tubes, and the output is electrolytic. Generally, I see that many people use I-shaped inductors as power conversion inductors. This kind of power supply, generally, has a picture on the chip information of the manufacturer, and it is basically a step-down type. I will not say much, there are many more people who are proficient in this than me.
The second is represented by me, that is, the constant current driver of the switching power supply control mode. This kind of IC uses ordinary switching power supply chips as the core conversion device. There are many such chips, such as PI's TNY series, TOP series, ST's VIPER12, VIPER22, Fairchild's FSD200, etc., and even RCC with only triodes or MOS tubes can be made. The advantage is low cost and good reliability. Because ordinary switching power supply chips are not only good in price, but also classic products that have been used a lot. Such ICs are generally integrated with MOS tubes, which are more convenient than 9910 with MOS, but the control method is more complicated and requires external constant current control devices, such as triodes or op amps. Magnetic components can be I-shaped inductors or high-frequency transformers with air gaps.
I like to use transformers because although the cost of inductors is very low, I think their load capacity is not good, and the adjustment of inductance is not flexible. So I think the better device choice is an ordinary integrated MOS switching power supply chip plus a high-frequency transformer. In terms of performance and cost, it is the most ideal choice. There is no need to use any constant current IC, which is not easy to use and expensive.

Finally, one of the most important ways to distinguish between these two power supplies is to see whether their outputs are filtered by electrolytic capacitors.

Regarding the power supply problem - whether it is a current-limiting constant current controlled power supply or a constant current power supply controlled by an op amp, the power supply problem must be solved. That is, when the switching power supply chip is working, a relatively stable DC voltage is needed to power the chip, and the working current of the chip ranges from one MA to several MA. There is a kind of chip like FSD200, NCP1012, and HV9910. This kind of chip is high-voltage self-fed, which is convenient to use, but the high-voltage feeding causes the heat of the IC to rise, because the IC has to withstand about 300V DC. As long as there is a little current, even one MA, there will be 0.3 watts of damage. Generally, the LED power supply is only about ten watts, and the loss of a few watts can reduce the efficiency of the power supply by several points. There is also a typical example like QX9910. Use a resistor to pull down to get power, so the loss is on the resistor, and it will lose about a few watts. Another method is magnetic coupling, which is to use a transformer to add a winding to the main power coil, just like the auxiliary winding of the flyback power supply, so as to avoid the loss of a few tenths of a watt of power. This is also one of the reasons why I do not use an isolated power supply but a transformer, just to avoid the loss of a few tenths of a watt of power and improve the efficiency by a few points.

An Overview of Non-Isolated Buck Power Supply Design Methodology

The non-isolated buck type is a commonly used power supply structure, which accounts for more than 90% of the fluorescent lamp power supply. Many people think that there is only one type of non-isolated power supply, buck type. When it comes to non-isolation, buck type comes to mind, and it is said that it is unsafe for the lamp - that is, after the power supply is damaged. In fact, buck type is only one type, and there are two basic structures, namely boost and buck-boost, that is, BOOST AND BUCK-BOOST. Even if the latter two power supplies are damaged, it will not affect the LED. There is such an advantage.

Buck power supply also has its advantages. The first point is that it is suitable for 220, but not for 110, because 110V originally has a low voltage, and it will be even lower after it is reduced. In this way, the output current is large, the voltage is low, and the efficiency is not very high.

The step-down 220V AC is about 300 volts after rectification and filtering. After the step-down circuit, the voltage is generally reduced to about 150V DC, so that high-voltage and low-current output can be achieved, and the efficiency can be high. Generally, MOS is used as a switch tube. My experience is that it can reach about 90% when making a power supply of this specification, and it is difficult to go higher. The reason is very simple. The chip generally has a self-loss of 0.5W to 1W, while the power supply of a fluorescent tube is only about 10W. So it is impossible to go higher. Now the power supply efficiency is very virtual, and many people are bragging, but it is actually impossible to achieve it. It is common for some people to say that the efficiency of a 3W power supply is 85%, and it is still an isolated type.

Let me tell you, even in frequency hopping mode, the minimum no-load power consumption is 0.3W. What's more, the output is 3W low voltage, which can reach 85%. In fact, 70% is very good. Anyway, many people now brag without thinking and can fool laymen. However, not many people who work on LEDs now understand power supply.

I have said that to achieve high efficiency, first of all, it must be non-isolated, and then the output specifications must be high voltage and low current, which can save the conduction loss of power components. Therefore, the main losses of LED power supplies like this are, first, the chip's own loss, which is generally about a few tenths of a watt to one watt, and second, the switch loss. Using MOS as a switch tube can significantly reduce this loss, while using a triode will cause much greater switch loss. So try not to use a triode. Also, when making a small power supply, it is best not to save too much, and do not use RCC, because the general manufacturers of RCC circuits simply cannot make good quality. In fact, chips are now cheap, and ordinary switching power supply chips with integrated MOS tubes are no more than two yuan. There is no need to save so much. RCC only saves some material costs, but in fact, the processing and repair costs are higher, and in the end it is not worth the loss.

The basic structure of a buck power supply is to connect the inductor and the load in series with the 300V high voltage. When the switch tube is switched, the load realizes a voltage lower than 300V. There are many specific circuits, and there are many on the Internet. I will not draw pictures to explain them. Now 9910 and the constant current ICs on the market are basically implemented with this circuit. But this circuit is that when the switch tube breaks down, the entire LED light board is finished, which should be the worst part. Because when the switch tube breaks down, the entire 300V voltage is added to the light board. The light board can only withstand more than 100 volts, but now it has become 300 volts. Once this happens, the LED will definitely burn out. So many people say that non-isolated is unsafe, but in fact, it means step-down. It’s just because most of the non-isolated ones are step-down, so they think that the damage of non-isolated will definitely damage the LED. In fact, the other two basic non-isolated structures will not affect the LED if the power supply is damaged.

The buck power supply must be designed with high voltage and low current to be efficient. Why? Because high voltage and low current can make the pulse width of the switch tube current larger, so that the peak current is smaller, and the loss of the inductor is also smaller. You can know from the circuit structure that the circuit is not convenient to draw, and it is difficult to describe it in detail. Just to summarize, the advantage of the buck power supply is that it is suitable for 220 high voltage input, so that the voltage stress of the power device is small, and it is suitable for large current output, such as 100MA current, which is easier than the latter two methods and has higher efficiency. The efficiency is relatively high, the loss of the inductor is small, but the loss of the switch tube is larger, because all the power passing through the load must be transmitted through the switch tube, but only a part of the output power passes through the inductor, such as 300V input, 120 output buck power supply, only 300-120, that is, 180 part, must pass through the inductor, and the 120 part is directly turned on to enter the load, so the loss of the inductor is relatively small, but the output power must all be converted through the switch tube.