The choice of MIS transistors for the voltage converter of an automobile ULF. Encyclopedia of radio electronics and electrical engineering

Encyclopedia of radio electronics and electrical engineering / Voltage converters, rectifiers, inverters

 Comments on the article

1. Learn to read primary sources

"Of all the parameters of the MIS transistor, the most important for us is the resistance of the open channel." Clausmobile

That's right, but it's not the only one. Let's take the documentation for the power transistor (say, IRFP054N) and analyze it piece by piece. And along the way, we will prioritize - what is important and what is not. I must say right away that according to three main parameters - the channel resistance Rds, the maximum operating voltage of the drain-source Vbrds, and the channel current Id, conclusions can be drawn, but it is advisable to operate with a full set of data. At least because the maximum permissible parameters at + 25C are guaranteed to kill the device at 100C. And, besides, the limiting data in the interpretation of different manufacturers are not always comparable.

So let's read the document.

Absolute maximums

DC drain current at Vgs=10V: Id=81A at 25C, Id=57A at 100C. And the note says - "Calculated based on the limiting (ideal) thermal resistance of the case." Therefore, in real life it is unattainable. We will determine the limiting current ourselves based on a reasonable thermal power, pulse duty cycle and channel resistance.

Impulse drain current Id=290A (with similar reservations). Great, but just as inaccessible.

Thermal power dissipated at 25С Pmax=170W and its decreasing temperature coefficient LDF(Pmax)=-1.1W/C. These two parameters always live inseparably. Indeed, when the crystal is heated to 125C (this is normal), the maximum permissible power decreases to 170-1.1 * (125-25) = 60W. Here on these 60 W, and with a margin - 50 W, and we will navigate for the time being.

Gate-source voltage limit (Vgs) - +/-20V. Safe enough for a 12V network.

Thermal resistance

PN junction-case - Rjc=0.9 C/W. This means that at 50W of heat loss, the temperature of the working area of ​​the crystal will be 0.9 * 50 = 45 degrees higher than the temperature of the transistor case (which, in turn, is lower than the average temperature of the radiator).

Radiator housing, flat surface with silicone grease - Rcs=0.24 C/W. Those. 60W will give another 12C of heat loss. With a mica gasket it will be a little worse. Another argument in favor of fully isolated transistors. Alas, while there are few of them and dogs are expensive ...

PN transition-air (in the absence of a radiator) - Rja=40C/W. What should have been proved - without a radiator, the device is useless.

Electrical parameters (at 25С at pn-junction)

Crazy settings. In view of the above, 25C at the crossing can only be in a very cold winter. Therefore, the temperature dependences of all parameters are extremely important. Thank God, IR does not lie and honestly talks about them.

The closed channel breakdown voltage is Vbrds=55V (Vgs=0V, channel threshold current is 250µA) and its temperature reduction factor LDF(Pmax)=-0.06W/C. Those. at 125C Vbrds drops to 49V. Two good conclusions. First, the drain voltage swing is equal to two supply voltages (i.e. 30V max) plus the inevitable fluctuation during switching (add another 10V) - a total of 40V, which clearly fits into the norm. Second, if 250 µA is already large enough and is considered a “breakdown” current, then the normal leakage current of a closed transistor is an order of magnitude lower (25 µA at 25C and Vds = 55V, but 250 µA at 150C). And of course, it is not necessary to disconnect it (the converter) from the battery in the non-working position.

Open channel resistance at Id=43A and Vgs=10V: Rds=12mOhm (milliOhm). Good resistance. The best single crystal IRFP064N in this respect has 6.4 mOhm (this was the lowest resistance in 1999. Times change - 2002 ...). Less - only for multi-chip modules. And how it behaves with increasing temperature is shown in graph 4. When the temperature drops to -40C, the resistance decreases by 25%. At 100C - increases by 40%. At 175C - doubles. Therefore, further in the calculations, I always operate with a double "passport" resistance.

Threshold gate voltage Vgsth=2.0..4.0V at Id=250µA. Graph 3 shows the temperature dependence of the transfer characteristic. It is clear from it that 8V is quite enough for a guaranteed full opening of the channel. "Everything else doesn't matter to me."

Gate leakage current IGSS=100nA - we are not interested.

The total gate charge is 130nC at Vgs=10V, Vds=43V. This parameter defines the requirements for the trigger circuit (gate driver). For an approximate calculation of such a circuit, see the TL494 IC application material on my website. Indirectly, it also determines the thermal safety of the transistor, because the bulk of the heat is released precisely in the transient process. And graph 6 shows its dependence on the gate voltage. It can be seen that, firstly, the "capacity" of the gate is non-linear, and secondly, the charges required to open and close the channel at 12V power supply will be different. And secondly, it practically does not depend on the supply voltage on the channel.

Turn-on and turn-off time delays - all have no more than 66 ns delay, which suits us.

Input and output capacities - we have already talked about the input. The output determines the resonances of the drain circuit, which are treated with an RC damper. However, in comparison with the oscillation generated by the actual load (transformer-rectifier), they are not serious.

Flyback Diode Parameters we are not particularly interested.

What is the sum?

2. Counting on fingers

I made up a simple sign (in Excel 98), in which it is possible to evaluate the thermal regime and the efficiency of the primary circuit of the converter - i.e. losses on keys and primary winding. The losses are presented as the sum of the open state losses (see paragraph above) and the transient state.

On-state losses are proportional to the square of the input current (i.e., the square of the power consumed), transient losses are linearly proportional to the input current (power). It can be seen that transient losses dominate at low power, at high powers - losses on the resistance of the open channel increase and sharply reduce the efficiency of the primary circuit. At the same time, heat losses are quite low. Those. the choice of a transistor in an expensive massive TO-247 or TO-3 package is unjustified - the smaller TO-220 package will provide thermal conditions no worse. As for the efficiency of heat dissipation and the reliability of the design, the author is in favor of a fully insulated TO-220 (for example, IRFI1010N) with both hands.

So how do we choose a transistor for an amplifier with an output power of Ru \u200d 12.5W? Let's set the limit losses - 7.5% ​​in the open state, 13% in transients, this is only in the primary circuit at maximum power. Assuming 67% efficiency of the secondary circuit, we have a total efficiency of 67%. Assuming the efficiency of the amplifier itself is also 200% at full power Pu (say 2.2W), we have Pin = 440 Py = 440W. In this case, the average input current Iin \u12d 37W / 80V \u37d 0.8A, and the current of public keys with a total duty cycle of 46% is 55A / 33 \u2d 55A. Losses should not exceed 46W in the open state and 2W in transients. Since Popen = I ^ 26 * Rds (Joule-Lenz law, let me remind you), Rds should be no more than 054W / (1010A) ^ 100 i.e. 220 mOhm - doubled "passport" value. So, IRFP131N fits in, almost without a margin. But the IRFI0.06N and BUZ5 will fit in the same way (of course, in TO-6 and not in SMD package). But BTS200 transistors with Rds = 250 Ohm will have to be installed as many as 12-XNUMX pieces per shoulder, but the cooling requirements for each will also decrease significantly. This is often used by installing a MiniDIP or SMD battery in devices without heatsinks at all. Of course, the parallelization of transistors requires special circuitry techniques and board layout, but with an output power above XNUMX-XNUMXW, there is simply no other way out yet. I refer the curious to the historical article by Shikhman in "Master XNUMX Volt" about the device of the Lanzarov amplifier

As for the power dissipated at the fronts, it practically does not depend on Rds - only on the current and the front duration. It is quite realistic to put it in 2-3 percent of the period, and close the issue for any allowable currents.

3. Regime

We choose low-voltage transistors (Vbrds = 55-100V) in the TO-220 package, or even better TO-220 Fullpak, based on the nameplate channel resistance