A simple method for selecting key transistors for switching power supplies

The photo shows the “difficult mistakes” method.
Shurik, this is not our method! When designing or assembling a ready-made SMPS circuit, one of the pressing issues is the choice of keys. And if you can somehow adjust the rest of the details (wind the transformer into 2 wires instead of 1, if there is not enough cross-section, or install two capacitors in parallel instead of one, if there is not enough capacity, etc.), then with the keys it’s not the same and just. The wrong choice leads to a big BOOM (remembering the famous Luc Besson film: “Bada-boom!”) due to thermal or electrical breakdown. And here, too, not everything is simple. An electrical breakdown will happen immediately (or almost immediately), but a thermal breakdown can take a long time to happen, and it will happen at the most inopportune crucial moment.

The first time I wondered about choosing keys was about 8 years ago. Where did I go first? On the Internet, of course, yeah. In general, I can now say this: I shouldn’t have done it. The issue of choosing keys for pulse technology on the Internet has become overgrown with a bunch of unreliable facts, myths and incorrect interpretations of graphs in datasheets. My way of choosing keys is also imperfect and incomplete. However, in the vast majority of cases in amateur radio practice, it will be enough and even behind the scenes, you yourself will not be happy. Let's begin!

↑ Transistor selection process

Now, let's try to figure out the issue of selecting a transistor.
No one should have any doubts about the issue of maximum voltage. Just for insurance, we take a key 200 Volts more than the maximum effective voltage in the circuit. For example, in SMPS I recommend 600-volt keys, no lower. The question is what to do with the temperature. It still counts! For thermal calculations, you just need to find out how many watts of loss will occur when the switch is operating and how much it needs to be cooled so that thermal breakdown does not occur. If the result is less than Tj

, then you can use such a transistor. If it’s more, alas, but you have to choose further.

What does heating consist of? To start with static losses

associated with the junction resistance
Rds on
, which affects the voltage drop across the junction, depending on the current flowing through the switch.
This voltage drop causes power to be released on the chip and heats up the transistor in the on state. It is calculated as the product of the square of the average pulse current Iimp
by the transition resistance
Rds on
and the duty cycle
Kzap
. The latter shows how much of the time the transistor is open.

In most amateur radio designs of bridge and half-bridge converters and amplifiers, Kzap is not higher than 0.45, and its further increase does not lead to anything particularly good, except for severe pain in the head or well... So, okay, we’ve sorted out the static losses.

Now dynamic losses.

These losses are the main problem in hard-switched FET converters. They occur when the key is turned on and off. So to speak, losses during transient processes. And the higher the conversion frequency, the higher the dynamic losses. And I also don’t want to make the frequency lower, because then the size of the transformer increases.

There are resonant or quasi-resonant circuits that can significantly reduce dynamic losses, but this is a complex technique to which the expression “simple calculation” does not fit.

So, dynamic losses consist of turn-on losses and turn-off losses. It is calculated as the product of the current at the beginning ( Ir

) or end (
If
) of the pulse, supply voltage (
Upit
) and rise time (
Tr
) or fall time (
Tf
), divided by double the pulse period. I would like to note right away: losses are calculated separately when turning on and separately when turning off, and then summed up.

Now cooling.

The main problem with cooling is the thermal resistance between different materials. The transistor has 2 such places: between the crystal and the transistor body, as well as between the transistor body and the radiator. These values ​​are tabular and do not require calculations. The first value is taken from the datasheet for the transistor. The second one can also be taken from there, if it is there. If not, then the average value is taken.

So, the losses have been calculated, it’s time to put them into action.

First of all, we add up the dynamic and static losses, we get the total losses - this is how many watts need to be removed from the crystal.

Then we add up the thermal resistances.

Now we multiply the total losses by the thermal resistance. The resulting result is the temperature that needs to be “blowed away” from the radiator. Let us subtract the result from the expected operating temperature, and at the output we will find the expected radiator temperature. It is from this that you can evaluate whether the transistor is suitable or not.

How? Very simple. The expected radiator temperature cannot be lower than the ambient temperature during natural cooling. That is, if you get a result of +24°, but outside it’s +32°, then that’s it, screw it! The transistors will experience a thermal breakdown, because no superfan can cool the radiator to 24 degrees if the air temperature is higher. It is very sad if the result is negative. If you do not have a freon or nitrogen cooling system, it is better to choose a different transistor.

Transistor as an amplifier

The transistor can also act as a small-signal amplifier, meaning it can be in any position between "fully on" and "fully off".

This means that a weak signal can drive a transistor and create a stronger copy of that signal at the collector-emitter (or drain-source) junction. In this way, the transistor can amplify weak signals.

Here's a simple amplifier for driving a speaker with a square wave signal:

↑ Subtleties

Of course, a case like this has its own subtleties and peculiarities.
In general, this can be characterized by the expression “don’t take it to extremes,” which very fully explains what not to do so as not to go crazy. First of all, this concerns temperatures.
Tj
is the maximum operating temperature of the transistor crystal, in fact the ceiling of its performance. It would be at the very least absurd to use this value in the calculation. Never push parameters into a corner, always leave room for maneuver.

For example, I use a temperature 5-10° lower in my calculations, and call it “Expected temperature” - Also.

.
Since Tj
indicated around 125° Celsius, I use 115-120° in the calculation.

Further, the ambient temperature for assessment should also not be taken at random. There are approved GOST standards, although you can simply accept +35° for the middle zone and +45° for the southern regions. This is so that in a room full of people in the summer the equipment does not burn out with a blue flame. Well, for cases of temperature fluctuations. For working outdoors under the sun, there are even more stringent conditions, but this is beyond the scope of amateur radio.

Next about voltages.

It is always worth making a safety margin for the permissible voltage.
Again, in the datasheet the Vdss
is the limit. And selecting a transistor strictly for the rectified mains voltage can play a cruel joke. Let's calculate: with a network voltage of 220 Volts, the output of the bridge rectifier will be 310 Volts. However, in reality, the network rarely has 220 Volts, and surges of up to 20% are, alas, a common occurrence. And what will happen if the voltage in the network increases by this 20%? The output of the rectifier will already be 378 Volts. Add in the noise from the welder and, voila, the 400-volt switch sparks and explodes.

I had the opportunity to repair a lot of amplifiers in which numerous Uncle Liaos skimped on transistors. Don't do this, there will be much more disappointment and savings.

Somehow, while wandering around the Internet, I came across an IR application note that recommended choosing keys with a margin of 200 - 250 Volts from the maximum voltage in the circuit. Alas, I did not save this appnote, and then I could not find it. Some people have doubts that it even exists, but the recommendation itself sounds quite sober, albeit relatively expensive.

Now about the transition resistance.

In the open state, an ideal switch should pass all the current without loss.
Alas, we live in an imperfect world. So imperfect that marketers are happy to take advantage of it. Opening the datasheet of any field-effect transistor, you can see a small characteristic Rds on
, written in large font. So: this is the transition resistance at a certain “room” temperature of 20-25 degrees. For the same IRFS840B, 0.8 Ohm is indicated.

This is all beautiful in words only, but in reality the crystal will heat up during operation, which will inevitably lead to an increase in the resistance of the open junction. Few people remember this, but this is exactly what you need to rely on when choosing a suitable transistor. Most often, datasheets do not indicate these sad figures, but only provide a graph of the temperature coefficient of resistance of the TCS, here it is for the transistor we have chosen:

As you can see in the graph, when heated, the resistance of the open junction quickly increases, and for the maximum operating 120° I recommended, the TCR of the open channel is already 2.1 Ohms, which means that from a pleasant 0.8 Ohms, an unpleasant 1.68 Ohms is already obtained. It’s sad, and that’s all, but we have to take it into account.

Well, the last of the subtleties. Be sure to consider the extreme characteristics of the transistor.

Datasheet tables always indicate three values: minimum, typical and maximum (or best, typical and worst). This applies to almost everything. For example, opening time and closing time. Moreover, from a marketing point of view, the emphasis is placed precisely on typical opening and closing times. So, for example, for the IRFS840B the typical rise time is 65 ns, which is what is written everywhere, although some instances go up to 140 ns, which is more than 2 times longer! Accordingly, for the calculation it is necessary to use exactly the worst value if there is no desire to select transistors for the design.

Tables of foreign analogues of transistors

If you find an inaccuracy in the tables of analogues or want to supplement them, write about it in the comments at the bottom of the page!

Table of bipolar transistor analogues

Foreign

Bipolar transistors up to 40 V

Bipolar transistors up to 60 V

Bipolar transistors up to 70 V

Bipolar transistors up to 80 V

Bipolar transistors up to 130 V

Bipolar transistors up to 160 V

Bipolar transistors up to 200 V

Bipolar transistors up to 250 V

Bipolar transistors up to 300 V

Bipolar transistors up to 400 V

Bipolar transistors up to 500 V

Bipolar transistors up to 600 V

Bipolar transistors up to 700 V

Bipolar transistors up to 800 V

Bipolar transistors up to 900 V

Bipolar transistors up to 1500 V

Bipolar transistors over 2000 V

Unijunction transistors

Powerful field effect transistors

Weak field effect transistors

↑ Let's summarize

To select a key transistor you must:

1. Always remember that environmental conditions are not ideal
2. Use the parameters of the worst instances in the calculations
3. Always leave some margin and room for maneuver
4. Keep in mind thermal changes in parameters
5. Do not let the crystal overheat
6. Avoid overvoltage due to poor network

Everything else is considered and selected.

And here I have a bonus for you.

Since I’m still lazy, I made a table in Excel that will calculate everything itself. All that remains is to draw a conclusion about the suitability or unsuitability of the transistor.

↑ Technical specifications

As always, I believe that an amateur design, as a rule, should be simple, cheap, technologically advanced, and consist of non-scarce parts. In addition, I long ago came to the conclusion that for such purposes it is better to make small simple boards without a power supply, without a digital indicator, without a complex case. It is enough to provide clamps for connecting an external laboratory regulated power supply, an indicator in the form of a simple digital tester or pointer instrument, and, if necessary, an oscilloscope, etc.

Such devices are quickly made and remade, and most importantly, they work and are useful. If you think of a multifunctional, self-sufficient device in a separate beautiful case, it will usually remain in the spotlights. In addition, if the device is made, it suddenly turns out that it is necessary to add another function, for example, a capacitance viewer, but there is no more space on the front panel and the design must be spoiled... Therefore, I believe that unsightly amateur narrowly functional products have a right to life.

So, it is planned to test silicon transistors in the mode - current 200 mA, voltage K-E = 2 V. You can quickly change the current in the range of approximately 150...300 mA, voltage K-E up to 5...7 V. You can check (by slightly changing the settings) composite transistors with two serial PN junctions.

Using a toggle switch you can change the current, for example, 10 times. This will allow you to test low-power transistors at a current of 15...30 mA (by replacing one resistor you can set any reasonable current). I think it’s important to connect any transistors conveniently. For transistors KT814-819 there are sockets on the board, for powerful transistors in cases like TO-247, TO-3R, there are clamps. They install wires with “crocodiles”, which allow you to connect transistors in the TO-3 housing, any transistors with bent solder leads, etc.

The change in the C-E voltage is carried out by an external power source, the goal is to check the identity of the modes at higher voltages and significant heating of the transistors. At 5 V and 200 mA, we get the maximum power for KT814 without a heat sink - 1 W. For larger cases without heatsinks, thermal power is typically = 2 W.

It is easy to see that the transistor gain depends to some extent on both voltage and temperature, so determining the absolute value of the transistor gain using a microprocessor with an accuracy of up to the seventh digit does not make sense. For this reason, the simplest circuit solution was chosen, which provides sufficient accuracy for practice and allows you to do without an op-amp, microcontroller and several power supplies. Any digital tester, for example M-832, is suitable for measuring the base current.

↑ Files