Discrete MOSFET gate driver problems

Thread Starter

sumeryamaner

Joined May 29, 2017
109
As a hobbyist, I am working for a long time on some circuits which require a power MOSFET driven with PWM to control various loads.

I have tried low side switching with N channel MOSFETs and high side switching with P channel MOSFETs. I know about the threshold gate source voltages, about RDS ON values etc. Static switching (without PWM) is OK. But as soon as I start using PWM (from a microcontroller) problems begin to arise.

I exactly know about the parasitic capacitances of a MOSFET and that the best way for PWM switching is using dedicated gate drivers but I want to learn something and try the gate drive using discrete components.

I know that I must charge and discharge the famous gate capacitor as quickly as possible to be able to switch very fast and to achieve the shortest time period where the MOSFET is partially conducting (producing waste heat).

I decided to use a totem pole driver for the logic level MOSFET (IRLZ44). You can see the circuit in the first figure. The load is a simple low voltage light bulb. I don’t think it has a significant inductance and for this reason I didn’t use an inductor and flyback diode.

In the totem pole I used a pair of BD137 / BD 138. They have a continuous Ic of about 1.5A (pulsed 3.0 A) and hfe of about 160.

The totem pole has been driven by two consecutive 2N2222 transistors in common emitter configuration.

The circuit is being fed by an Arduino with a PWM frequency of 490 Hz.

The logic part of the circuit is supplied with 5V and the light bulb is powered by a separate 1S (and 2S) LiPo battery (4.2V – 8.4 V respectively).

With this configuration I have a very clean PWM signal at the output of the totem pole but almost a sawtooth waveform (with a short rise time and long falling fime) at the MOSFET output (measured at the terminals of the light bulb).

The waveform suggests that I can drive the gate to ON position without any problems but I cannot discharge the gate capacitor in a short time.

Now my questions:
1- Is a circuit modification (with a series inductor and a flywheel diode) like in the Figure 2 necessary even if the load is a simple light bulb?
2- Is the collector resistor of the second 2N2222 too high? Should I lower it to maybe 470 ohms so that the totem pole can produce a higher collector current to drive the MOSFET gate?
3- Even if I set a low duty cycle for the PWM signal, the bulb gets brighter and brighter until it burns out in about 30 – 40 seconds. I cannot find any explanation for this.
4- Are there any other factors I am overlooking or unaware of?

(Please bear in mind that I am a hobbyist and not a professional).

Thank you very much...

AAC1.jpg

AAC2.jpg
 
Last edited by a moderator:

crutschow

Joined Mar 14, 2008
23,310
What is the bulb rated voltage?
Yes, a diode should not be needed for a bulb load.

The Bd137 and Bd138 are big (and thus likely slow) transistors.
it would be better to use smaller ones such as the 2N2222 and 2N2907.

Make R3 about 100Ω and R2 about 1kΩ.
 

Thread Starter

sumeryamaner

Joined May 29, 2017
109
What is the bulb rated voltage?
Yes, a diode should not be needed for a bulb load.

The Bd137 and Bd138 are big (and thus likely slow) transistors.
it would be better to use smaller ones such as the 2N2222 and 2N2907.

Make R3 about 100Ω and R2 about 1kΩ.
Thank you very much.
The bulb is rated at about 3 V.
I will try your suggestions but is the Ic of 2N2907 enough do rapidly discharge the gate capacitor?
 
Last edited:

Thread Starter

sumeryamaner

Joined May 29, 2017
109
The first picture is the gate signal to the MOSFET. It is OK.
The second signal is obtained by connecting the oscilloscope between battery plus and MOSFET drain. This is the sawtooth like waveform.
But...
This signal is without the load connected. When I connect the bulb I get the third picture... Very clean PWM output. I think this change with the connection of the load can be explained by the parasitic capacitances...pic002.jpg pic003.jpg pic001.jpg
 

crutschow

Joined Mar 14, 2008
23,310
The second signal is obtained by connecting the oscilloscope between battery plus and MOSFET drain. This is the sawtooth like waveform.
Without a load you are just charging and discharging the drain capacitance through the meter resistance.
That waveform has no particular significance.
 

Thread Starter

sumeryamaner

Joined May 29, 2017
109
So the sawtooth waveform was my mistake and now I know the reason. Let's move on to my 3rd question.

3- Even if I set a low duty cycle for the PWM signal, the bulb gets brighter and brighter until it burns out in about 30 – 40 seconds. I cannot find any explanation for this.

To prevent this, I am beginning with a somehow higher PWM value and decrease it in 200 mS steps to prevent the bulb burning out but I would like to know the real reason and prevention.
 

michael8

Joined Jan 11, 2015
38
The load is a simple low voltage light bulb. I don’t think it has a significant inductance
and for this reason I didn’t use an inductor and flyback diode.

A light bulb isn't a simple load as it's resistance varies with
temperature of the filament. So when cold the bulb will draw a much
higher current then hot. Also the lifetime is related to (1/V)**16
(16th power) so a bit too much voltage is quickly fatal.

https://en.wikipedia.org/wiki/Incandescent_light_bulb
 

Thread Starter

sumeryamaner

Joined May 29, 2017
109
The load is a simple low voltage light bulb. I don’t think it has a significant inductance
and for this reason I didn’t use an inductor and flyback diode.

A light bulb isn't a simple load as it's resistance varies with
temperature of the filament. So when cold the bulb will draw a much
higher current then hot. Also the lifetime is related to (1/V)**16
(16th power) so a bit too much voltage is quickly fatal.

https://en.wikipedia.org/wiki/Incandescent_light_bulb
You are right, it has a positive temperature coefficient but as the resistance of a light bulb increases as it is getting hot, with a fixed PWM rate it shouldn't get brighter in a few seconds.
 

ci139

Joined Jul 11, 2016
702
if you drive the L1 at "sufficient" frequency for the current to keep running through it then your "fly-back diode" would close the loop for the (induced) current flow = use lower frequency or lower duty

(also) the "1:10" rule may apply to driving misc. BJT MOSFET as provide current through 10k and shunt the base or gate by 100k (it applies at low frequencies however up to 100kHz) . . . e.g if you charge the MOSFET gate through 100Ω-s you usually shunt it with 1kΩ and reverse flyback diode from source to gate such as low reverse recovery and low junction capacity 1N4148 /// however for decent MOS-Fet-s the values of 240Ω : 2k4Ω might do up to 2MHz

at higher frequencies the resistor vales need to be less to be able to provide sufficient current for the BJT storage time and MOSFET's gate charge
 
Last edited:

Thread Starter

sumeryamaner

Joined May 29, 2017
109
if you drive the L1 at "sufficient" frequency for the current to keep running through it then your "fly-back diode" would close the loop for the (induced) current flow = use lower frequency or lower duty

also the 1:10 rule may apply to driving misc. BJT MOSFET as provide current through 10k and shunt the base or gate by 100k
at higher frequencies the resistor vales need to be less to be able to provide sufficient current for the BJT storage time and MOSFET's gate charge
Thank you very much. There is no "L1". In the original question I asked "should I use an inductor for a light bulb?"
All oscilloscope traces are without an inductor.
The PWM frequency is less than 1 kHz.
The duty cycle is between 4 - 10 %.
 

michael8

Joined Jan 11, 2015
38
> The PWM frequency is less than 1 kHz. The duty cycle is between 4 - 10 %.

So the pulse is less than 1000uS and only 40uS to 100uS long.

What is the time constant of the filament in the bulb?

google: bulb filament time constant

all 120V bulbs
https://www.researchgate.net/publication/320835572_Heating-times_of_tungsten_filament_incandescent_lamps

lower voltage
https://link.springer.com/article/10.1007/s10762-014-0130-8

It looks like 1000uS isn't long enough for the filament to reach full
power, temperature and resistance. I would think this would result in
lower brightness at "normal" rated current. If the current was pushed
higher to get to "normal" brightness the bulb lifetime would of course
decrease...
 

ci139

Joined Jul 11, 2016
702
? just a guess that the 4k7 does not let the BD138 to open . . . ? try 4k7 down from the base of your Q1 (PWM input) to GND
. . . ? increase R2 or put a forward diode 1N4148 between Q2b & Q1c or use the TTL type drive . . .
(( don't follow me )) ↓↓ a quick spice test ( = 1 pass analysis) -- not a real life one!
Mos-Fet~drive-TEST_xaA.png + v.2 Blank.png + v.3 Mos-Fet~drive-TEST_xaB.png + v.4 Mos-Fet~drive-TEST_xaCM.png
--
- - the last uses (! loosely !) approximated model of http://www.ti.com/lit/ds/symlink/sn74lvc1gu04.pdf
 
Last edited:

crutschow

Joined Mar 14, 2008
23,310
It looks like 1000uS isn't long enough for the filament to reach full
power, temperature and resistance
It's not supposed to.
The high frequency PWM signal is so that the bulb's thermal time-constant is much longer than the pulse period, thus the bulb sees the average value of the PWM signal and does not respond to the individual pulses (which would cause flicker).
 
Top