# N-channel mosfet calculations

I am in the process of selecting a suitable FET to switch a resistive load (LEDs), triggered by a 3.3v MCU signal. The schematic would be:

However, I am having some trouble with my (rough) calculations for the CSD16342Q5A. From the datasheet -
$R_{DS(on)}$ = 6.1mΩ
$R_{\theta JA}$ = 123 °C/W (Pg 3 - taking into account that I would be not using additional copper for a head sink )

The rise in temperature as a function of the drain current $I_D$ is plotted as:

The questions

1. How does one factoring in the effect of the PWM switching. Should one use Figure 1. of the datasheet to factor in those effects?
2. How would one determine the gate current that the MCU has to sink/source for each toggle?
3. How does one proceed if $R_{DS}$ values at the $V_{GS}$ = 3.3v aren't available (but $V_{GS(threshold)}$ is 1.8v) and only given for $V_{GS}$ = 4.5v (eg.CSD18536KT)
• FYI, you probably want to turn your LED around and - likely - want to have some way of controlling the current through the diode. The most common method for small LEDs is a series resistor. For larger LEDs, you can get into much more complicated current-control which often involves dedicating an IC to the task. – slightlynybbled Dec 15 '16 at 0:25
• @slightlynybbled: Oops! I am going to be controlling LED strips, so they have current limiting resistors inbuilt. – Ash Dec 15 '16 at 0:42
• What is your expected LED string current at 12 V? – Jack Creasey Dec 15 '16 at 2:00
• @Ash Since you have current-limiting resistors built in, what is the value of those resistors? It matters greatly in sizing your MOSFET. – slightlynybbled Dec 15 '16 at 23:49
• @JackCreasey: About 5 amps (11 W/m * 5 meters). slightlynybbled: I will look them up, but would you be able to elaborate as to why? If the total current consumption is known isn't that sufficient? – Ash Dec 16 '16 at 11:01

1. How does one factor in the effect of the PWM switching. Should one use Figure 1. of the datasheet to factor in those effects?

Approximation will be PWM% * ID in the formula you showed above. The temp goes up with the square if ID assuming RDS(on) is more or less constant.

1. How would one determine the gate current that the MCU has to sink/source for each toggle?

There is a pulse current charging and discharging the device Gate capacitance which is approximately 1000 pf. Since your MCU output pin is probably short circuit current limited to 30 mA or so, there will be a slew rate limit to the rise and fall times, and this will likely be in the 50 - 80 nS range. If you want to limit the current out of the I/O pin to 10 mA maximum you could put a series resistor to the Gate of 330 Ohm and increase the 10k to 47k. With a VGS(th) of about 1.4 V (for IDS = 1 A) this would give you a turn on/off time of about 175 nS.

1. How does one proceed if RDS values at the VGS = 3.3v aren't available (but VGS(threshold) is 1.8v) and only given for VGS = 4.5v?

There seems to be an RDS(on) shown on the datasheet for a VGS of 2.5 V

Remember that VGS(th) is usually specified at a very low ID, in this case 250 uA, so not very useful for this application. You are really interested in when the device supports your current draw requirements and from Figure 3 you can see that at about ID = 1 A you have a VGS of ~1.4 V.

• Wouldn't the current limiting resistor and the gate capacitance form an effective LC filter, further reducing the PWM frequency? – Ash Dec 16 '16 at 11:04
• @Ash. Of course, but with 330 Ohms and 1000 pf that would be -3 dB @ 500 kHz, so not of concern with PWM frequencies of a few kilohertz. The only impact in this application is to reduce the slew rate for turn on an turn off. – Jack Creasey Dec 16 '16 at 17:25
• MOSFETs Ciss is differential parameter measured using a small AC superimposed on a bias point, it's not what you need to ballpark switching behaviour. One should rely on gate charge plots instead. – carloc Sep 1 '18 at 17:12
• @carloc It's complicated I admit, but the Ciss values are a good starting point: infineon.com/dgdl/… – Jack Creasey Sep 1 '18 at 17:36
• Yes we can get close or even spot the same result sometimes but I believe it's conceptually wrong. Also, reading gate charge in datasheet next line instead of Ciss doesn't take longer. – carloc Sep 1 '18 at 19:06