Efficiencies of power supplies with PWM

[MrAl]

Hello again,

Sounds like an interesting project but more so is the lighting effects you are going for.  Should be cool to watch.  Maybe you could add some lightning and sound effects like thunder and rain.  Lightning could just be high intensity flashes of white light followed by a short delay and then the thunder.  Simple rain can be a random number generator output driving a speaker and you can play with the randomness to get the rain to sound realistic.

[ElCaron]

I always wanted to create the illusion of waking up to a day of bad weather *g*
There will also be 40W worth of controlled COB spots, so lightning brightness shouldn't be an issue.

[MrAl]

I always wanted to create the illusion of waking up to a day of bad weather *g*
There will also be 40W worth of controlled COB spots, so lightning brightness shouldn't be an issue.

Hi,

he he :-)

The sound of rain can be very soothing and blocks out other sounds like street noise.  The thunder and lightning is just for the fun of it really and anyone that comes over to your place might get a kick out of it.  I think it would be really cool, but yeah, you would not want it running all the time :-)

[Southpark]

The whole circuit will take a mean power of 12V*(mean current).

Is that 12V DC? If 12V DC input to LEDs... then the current will be DC as well.

[ElCaron]

So? Yes, of course 12V DC. But PWM controlled, so there is an on-duty and a mean current.
At sone points in this thread DC has also been used as a contrast to "switching PWM fast enough so tgat the system cannot be described with static full-on/full-off states anymore".

[Southpark]

So? Yes, of course 12V DC. But PWM controlled, so there is an on-duty and a mean current.
At sone points in this thread DC has also been used as a contrast to "switching PWM fast enough so tgat the system cannot be described with static full-on/full-off states anymore".

I see! Thanks. So that means the average power will be a time-averaged value of the product between the 'switched 12V signal' and the 'switched current'? So not just 12V times the mean current?

[ElCaron]

That is a question basically the this whole thread is about.

If I read correctly< what you said, that is the same. The mean current is
Im = I_on * T_on/T_periode
The mean power
Pm = P_on * T_on/T_periode = (U*I_on) * T_on/T_periode = U * (I_on * T_on/T_periode) = U*Im

BUT: A switching power supply has different efficiencies (GrumpyMike, I would still like to know why that is a wrong use of the word) at different loads. So a 100W supply that has a load of 70% for 10% of the time will most probably draw less overall power than a 100W supply that has a load of 7% for 100% of the time.
If the LEDs switch on and off slow enough, the first situation will be the case. If they get too fast, however, the supply will not be able to resolve the on/off states and see a constant 7% load, at a worse efficiency.
My first experiment showed that for the supply I have, PWM frequencies of a few hundred Hertz are fine. After that, I am not sure if the logic-level MOSFET couldn't follow because of the gate capacity, or the supply became less efficient.

For all consideration regarding PWM vs current-limiting by resistor, read my summary in post #40. Short version:
Efficiency is the same, current-limiting might lead to a better distribution of heat, but may make color balancing of multiple RGB strips quite difficult. That is why I will go with well-cooled, PWM-controlled strips for now.
Even better would be voltage control down to the maximum desired brightness, if that is much smaller than 100% duty cycle in PWM.

[MrAl]

Hello again,


What were you using to drive the MOSFET gate for that test?

[ElCaron]

An Arduino Nano with a 220Ohm resistor between the pin and the gate of the IRLZ44N.
If I remember correctly, the gate would be saturated after some time in the order of 10^-5 - 10^-6 seconds (though I am not sure I calculated that correctly), so at least at the higher Arduino PWM frequencies, this might matter. I do not own an oscilloscope to test that.

Should be very similar to the Adafruit PCA9685 breakout I am going to use eventually.

[MrAl]

An Arduino Nano with a 220Ohm resistor between the pin and the gate of the IRLZ44N.
If I remember correctly, the gate would be saturated after some time in the order of 10^-5 - 10^-6 seconds (though I am not sure I calculated that correctly), so at least at the higher Arduino PWM frequencies, this might matter. I do not own an oscilloscope to test that.

Should be very similar to the Adafruit PCA9685 breakout I am going to use eventually.

Hi again,

With a gate charge of 48nC and voltage of 2.5v and 220 ohms i am getting around 5us.
But the best test is to use two i/o pins instead of one.  You can try connecting two together and switching them at the same time, but better yet use two 220 ohm resistors to drive the gate, one resistor per i/o pin, and that reduces the risk of one pin being high while the other is low and causing a huge current spike.  If that doesnt make a difference then it could be something else.

[ElCaron]

That is a good idea. I'll be on the project again this weekend.