# Paralleling hexfets

Discussion in 'Electronic Design' started by Bart, Jun 20, 2004.

1. ### BartGuest

Dear,

I have a question about paralleling hexfets.
I already searched a lot, but I did not find yet the rigth answer to
my question.

Let me explain what I want to do:
I created a circuit to drive four fans. (I want to regulate the fans
RPM) Explained in short it is as follows:
The gate of my FET is driven by a fan driver (MAX 6651), the source is
connected to ground and the drain is connected to the first wire of
the fan, the other wire of the fan is connected to 24V.
So by regulating the voltage at the gate, the 24V power voltage is
divided in 2 parts, one part is over the fet (Vds) the other part is
over the fan.

If I connect four fans to 1 fet (which is theoretical right) there is
a lot of heat dissipated in the hexfet. I measured temperatures of
130 °C. So I thougt to place four fets in parallel. Then the current
would divide into 4 equal parts and the dissipated heat in the fets
will decrease.

By measuring temperatures after placing them in parallel, the result
is as follows: One fet is getting very hot (120°C), 2 other fets
about 70°C and the fourth has 40°C. How is this possible?

I tried already several things. I tried with 2 fets. Then the first
ons has 100°C and the second 75°C. On the internet I found that you
have to place serial gate resistors (typical value = 100 ohm) when you
place fets in parallel. Also when I try this, the temperature does
not decrease.

Question: is there anyone who can tell me what I am doing wrong. I
would like to have a circuit by which the maximum temp is about 40 or
50 °C (instead of 120 °C). The number of fets doesn't matter.

Is it useful to place a serial source resistor?
If so, what is the typical value about that?

2. ### Tim WescottGuest

If you parallel hexfets of the same type and drive them so they are full
on then they should share current fairly well. It sounds like you are
using them as linear devices (i.e. you aren't turning them full on),
which will make them much more sensitive to variations in their
threshold voltages.

Putting individual source resistances on will help. Decide how much
voltage drop you can afford when your fans are running flat out, and
figure the allowable source resistance from that and the fan current.

3. ### Tim WescottGuest

If the fan controller is pulsing, which it doesn't sound like from his
description.

If it is, then ignore my response...

4. ### John PopelishGuest

So they don't all have exactly the same resistance.
Since the gates do not pass DC the resistor does not have any voltage
drop and does not change the voltage on the gates.
Use each fet to run one fan, with all their sources and gates tied in
parallel, but each drain going to only one fan. That way, each fet
sees the current from one fan.
This will help. A resistor with a value that is about the same as the
typical drain to source resistance will cut the variations at least in
half.

5. ### Ken SmithGuest

[...]

If you are running the HEXFETs under its rated current and the linear
mode, the drain current has a positive temperature coef. Which ever fet
has the lower threshold voltage will try to take almost all of the current
and get hot. You need a source resistor or to use the fets as source
followers to drive the fans.

6. ### artieGuest

Suggestion--

Back off to a simpler design, running one fan with one FET, and get

From the 6650/1 spec sheet (page 6), the controller is running the FET
as a linear device -- that means you need some kind of heat sink for
the FET. Figure out worst-case power dissipation, and size both the
FET and its heat sink to handle that.

The spec sheet (page 14) shows controlling 3 fans using one FET. Four
fans is going to require fooling the controller, such as using the
expansion circuitry shown in Fig. 8 (page 15), or by not connecting one
fan's tach output to anything, and assuming its speed will follow the
other fans.

You also mention using +24 volts to power the fans. The spec sheet
shows an absolute max voltage on the feedback and tach pins of +13.2V.
Page 13 of the spec sheet shows how to calculate the value of
protection resistors for these pins.

7. ### mook johnsonGuest

Go PWM and solve all of these problems. one FET switching on and off with
variable duty cycle will dissipate far less power than linear mode.

8. ### Ken SmithGuest

Or, since there are several fans, only turn on as many fans as needed.

9. ### leggGuest

Mosfets will only share linearly if Vgs variations are overcome. The
source resistor on each device is one way of doing this.

The fan loads are also not characterized to run in parallel at
anywhere near the same rpm, particularly at their starting or stalling
speeds. At that time, individual fans will expect large pulsed
currents to be available to start, and the actual voltage to rpm
relationship of each device may exhibit considerable hysterisis about
the start/stop point.

RL