What a switching power supply consists of, part 3

What exactly is an inverter?
This circuit converts DC into AC. In this case, we have 310 V DC at the input, which needs to be fed to a transformer. But since transformers don't work on DC, we need an inverter.


An inverter consists of two main sections.
The PWM controller.


And the high-voltage output transistors. As it happens, the transformer that drives these transistors is also in the shot.


An inverter can also be much simpler, as in this well-known power supply.


An IC and a handful of components—that's the whole PWM controller.


The power supply topology and power rating are quite different from the previous example, so there's only one transistor.


Here's another version: the input filter capacitors are on the left, the transformer is on the right, and the inverter is between them.
Since the power transistor dissipates a significant amount of power, it's usually mounted on a heatsink.


But let's take a little detour into history and look at how it all began. Of course, it may not have started with this, so more accurately, let's look at where I started.
As you can imagine, PWM controllers didn't exist back then, and sometimes even a basic linear regulator was hard to find. But progress didn't stand still, and hobbyists tried replacing large transformers with switch-mode power supplies.
The diagram shows a typical self-oscillating circuit, though there were also circuits that used simple logic to generate pulses.


Back then, circuits for power supplies like these often appeared in the Radio magazine in articles about power amplifiers. But I first came across one in a power supply for a Sinclair. By the way, the photo shows one of them, which I kept as a memento :)
The circuit above did require selecting transistors, though, and in my case it overheated badly.


A self-oscillating circuit is considered the simplest option. In this example, it doesn't even regulate the output voltage.


Despite the huge variety of modern ICs, the circuit shown above has also found a place in the modern world as an "electronic transformer" for halogen lamps.


These lamps are gradually being replaced by LEDs, but electronic transformers are still quite popular, mainly because they're simple and inexpensive.


Much later, circuits like these got a second wind. International Rectifier, a well-known company, released a very simple IC for electronic ballasts for fluorescent lamps. It turned out that this IC also worked very well as the oscillator for a switch-mode power supply. These include the IR2151, IR2153, and similar ICs.
Some hobbyists even built regulated power supplies around this IC, but they don't always work properly.


This IC needs only a few small components and a pair of MOSFETs—that's the whole inverter circuit. I used this very IC to build the primary-side power supply for my bench supply.
By the way, I recommend this IC for powering audio power amplifiers: it's undemanding and quite reliable. I should also mention that unregulated power supplies tend to produce less noise.


This is what a 300 W, three-channel power supply with PWM fan control looks like. You can find more details in my review of the bench supply.


Single-ended power supplies based on self-oscillating circuits are also quite common. They were especially common in ATX power supplies as standby supplies.


You can also find them in very cheap phone chargers. A self-oscillating circuit is the simplest type of inverter.


There are exceptions, though. For example, the power supply in a fairly expensive branded air conditioner also used a self-oscillating circuit, but it was built quite well and included voltage regulation.


The next time I came across switch-mode power supplies was in the new televisions of the time. After the large, heavy transformers, this was progress.


The circuit design was pretty awful, it was difficult to repair, and I wouldn't call it compact either. The power supply in the photo is rated at 80 W.
At first, these were also built around a self-oscillating circuit, but later they started using an IC. That didn't really change much, though.


And now we come to more modern inverters. By this point, power supplies had reached the circuit-design level we see in modern units today.
Yes, the frequency went up, the operating range widened, and power increased, but the basic idea remained the same as it was 30 years ago. However, since integrated PWM controllers were still not very advanced back then, they were made as modules.


Even modern power supplies aren't above using these kinds of standardized modules. In some ways, they're actually quite convenient.


A typical block diagram for common inverter models consists of five sections.
1. Supply voltage monitoring, with protection against undervoltage and overvoltage.
2. Auxiliary power supply or startup circuit.
3. Power switch and current sensor. This section can vary significantly depending on the power supply topology.
4. The PWM controller itself—the brains of the power supply.
5. The PWM controller's main power supply section.


Let's look at how most power supplies start up. This information can help when troubleshooting.
Once the high voltage is applied, it passes through a resistor and into the PWM controller's power supply circuit.


As soon as the voltage reaches the startup threshold, the PWM controller starts up, drawing power from the capacitor in its supply circuit.
If your power supply shows no signs of life, check whether the PWM controller is getting power at its input. Sometimes these resistors go open-circuit.


Next, the PWM controller checks whether the supply voltage is within range. Not all inverters have this circuit, so if yours doesn't, you can go straight to the next step.


If the supply voltage is fine, the controller starts sending drive pulses to the power transistor. At the same time, it monitors the current through the transistor, and if it exceeds the limit, the PWM controller switches to protection mode.


If everything is working properly, after just a few cycles the main supply circuit produces its operating voltage, which then powers the controller. This is another point of failure: if that voltage is missing, the power supply will keep cycling on and off.


Once startup has completed successfully, PWM regulation takes over. I always compare it to a barrel that we fill in batches and drain through another tap at varying flow rates. The controller's job is to keep the water level in the barrel constant, even though the inlet tap has only two states: open and closed.
By the way, many of you have probably seen a resistor connected across the output of a power supply. It's there to provide a minimum load, since power supplies have trouble operating with very narrow pulses.

For example, here's the pulse width under a light load.


If you increase the load, the PWM controller will send more energy to the transformer and, through it, to the load.


Even if, for example, you load the power supply to its full capacity, the pulses won't reach their maximum width.


That headroom is needed to compensate for a drop in the input voltage.


If you lower the input voltage even further, the PWM controller will simply set the pulse width to its maximum. By the way, power supply PWM controllers don't generate a 100% duty cycle, because some "dead time" is always needed to protect the output transistors. During this time, the output transistors are off.
For single-ended flyback power supplies—the type used for low-power supplies—the maximum duty cycle is 50%.


The very first PWM controller I came across was the legendary TL494. The chip is very old, but it turned out to be a cheap, highly versatile controller for its designers. Even many years later, despite the availability of modern alternatives, it is still widely used in power supplies.
It is made by many manufacturers and sometimes under different names. For example, Samsung's equivalent is called KA7500.


At first glance, its internal circuitry may seem rather complicated, but in fact it isn't.


Simplified a little, the diagram looks roughly like this:
1 and 2. Power regulator and reference voltage source.
3. Pulse generator, which sets the controller's operating frequency.
4. Two comparators, one usually used for current regulation and the other for voltage regulation.
5. Dead-time generator, i.e. the minimum interval between the outputs being on.
6. Signal summing stage.
7. A flip-flop that controls the output switches and sets the operating mode: push-pull or single-ended. In some equivalents, this flip-flop malfunctioned at frequencies below 100 Hz, causing no end of surprises for people building 220 V step-up inverters.


The chip comes in a 16-pin package. It is reliable in itself, but in ATX power supplies, where it is powered from the standby supply, it sometimes fails after the supply runs out of control when the 5 V output capacitor dries out. The reference voltage regulator would break down, and a dangerously high voltage could easily appear at the PSU output. So when checking the PSU, first make sure there is 5 V on pin 14.


In AT power supplies, and later in the common Chinese enclosed PSUs, it is powered by the main transformer itself. Startup is provided by resistors in the base circuits of the power switches. When power is applied, they initially operate as a self-oscillating circuit. A small voltage appears at the transformer output, the chip starts up and takes over control. So if the PSU won't start, the first thing to check is the resistors highlighted in the schematic.


The second legendary PWM controller is the single-ended UC384x family.
I think you may have come across them before in power supplies and voltage converters.


Its internal circuitry is quite similar to the TL494, but differs in a few ways. To start with, the chip has only one output, not two.
Also, the comparators are tied to specific voltages set inside the chip, rather than being universal.
And, of course, its key feature is low-current startup. Until the chip starts operating, it draws very little current, so it can be started directly from the input voltage through a resistor. The TL494 can't do that.
For reliable startup, the chip has a threshold circuit that determines its turn-on and turn-off voltages. There are two versions, with thresholds of about 9 and 15 V.
The chip is also available with 50% and 100% duty cycles: the first is used in power supplies, the second in voltage converters.
That gives us four versions of this controller.


The chip is available in different packages, but the 8-pin version is the most common.


A typical power supply circuit using this chip looks roughly like this.


There are many power supplies on the market now that use other chips, but if you look at their schematics, you'll see a lot of similarities: the same stages and components. Any differences are minimal.


Power supply inverters can use different topologies, and I'll definitely cover that separately. But most use flyback or half-bridge circuits—the two upper schematics in the drawing. In fact, all the power supplies described today work this way.


But let's get back to PWM controllers. Earlier, I described designs where the PWM controller and power stage are separate. Fully integrated controllers have also become common, though—for example, the TOP series from Power Integrations, where practically everything is built into a single package.
Not long ago, I even came across a counterfeit. Interestingly, it's on the left, with laser marking; the original is on the right.


They became popular thanks to their very simple circuitry. In a basic design, a power supply consists of just a few components.


More advanced controllers came along later, allowing you to set the turn-on and turn-off voltages, as well as the output power limit. If desired, though, they can be configured for three-output operation by connecting the pins as shown in the earlier photo.
In any case, these controllers are much smarter and have comprehensive protection against various problems. For example, they withstood input voltages above 300 V simply by shutting down.


But their popularity was also due to the handy design software provided by the manufacturer. It let you calculate everything, right down to the transformer winding layout. And if it found problems with the calculations, it offered suggestions.


The manufacturer provided application examples for its chips. There was even a computer power supply design, but it never really caught on.


They are quite common in small power supplies, such as those used in monitors.


I’ve also been using them extensively myself for probably 15 years now.



Chinese manufacturers are keeping up too, releasing their own versions of similar chips.


They are used quite successfully in small power supplies


By the way, if you want, you can use PWM controllers without feedback from the output voltage, using the controller’s own supply winding instead. This simplifies the circuit, but of course the regulation will be a little less stable than with proper feedback.



That’s about it, in broad strokes. Sometimes it seems to me that the more I explain, the more I leave out—there’s still so much I’d like to cover in greater detail, but there just isn’t time. So there will probably be more installments on individual sections and how they work.
The video turned out way too long—I didn’t expect that myself. And that’s despite barely saying anything about the main switching transistors, and even cutting some of that out. I guess I talk too much :(


A few links to useful reviews mentioned in the video.
A decent ZXY6005S DC-DC module, or a DIY bench power supply
12 V, 6–8 A power supply that pleasantly surprised me
12 V, 5 A power supply, or how it could have been made
DC-DC converter: sometimes that’s just how it goes
S-180-12 180 W 12 V / 15 A power supply in an unusual form factor
36 V, 10 A, 360 W—or more on how power supplies work, plus a small bonus
48 V, 5 A and 240 W, or a power supply that managed to surprise me
Power supplies: small and very small