I'd like to design an LED stripe dimmer using PWM and a MOSFET. My current playground is AVR-based and later versions might build on another platform. So the whole thing must be controlled from 3.3 V min. The LED stripes have 24 V to reduce the current flowing through them (not the 12 V models). The expected current per channel is around 3-5 A max.
So far I found out that it may be challenging to find a MOSFET that can be used here directly, i.e. connecting the MOSFET gate to a GPIO output more or less directly. (With some resistors.) So I might need some kind of MOSFET driver in this application to drive the MOSFETs fast enough to achieve the desired PWM frequency (2.5 kHz) and dimming resolution (~4000 levels; 100 ns shortest pulse), but also to keep the power dissipation low during switching. A gate voltage of 3.3 V and a small GPIO current of 40 mA max might be insufficient here.
The thing is, a driver needs its own power supply to, well, power the whole thing (more than the MCU). I already have an external 24 V supply for the LED stripes and the device. Inside, 5/3.3 V is generated from that to supply the MCU. The current for that isn't all that much so a small LDO should do.
But am I correct that I need yet another step-down for the driver? Most drivers seem to work up to 18 V and the MOSFET gates also don't want 24 V. So I'd probably need something around 12 V for the driver. That makes the whole thing more complicated. And what LDO could I use here? When I read that the driver can push up to 1 A into the gate (to be quick), even if only for a very short time, would the LDO need to be designed for that current? Or would something smaller do with some capacitor or so. Switching or linear regulator?
I've never used a driver and couldn't find any information online about using a driver with its own supply voltage. All other sources only show 12 V LEDs or are too abstract. Any help is appreciated here!
The driver voltage should be appropriate for the MOSFETs, either 5V or 12V depending if they are logic level or not. Drivers need good high frequency decoupling, typically 100nF and 10µF ceramic for each chip, as they
switch high currents very fast. You can add some series gate resistance to reduce the switching speed somewhat
to reduce EMI, something in the 10 to 100 ohms range might be good for 2.5kHz PWM.
Double check the drivers you choose are 3V compatible on the logic inputs, most are.
The MOSFET driver would generally be powered with 5v, assuming logic level fets. It's purpose is that it can dump an eye-popping amount of current into or out of the mosfet gate for the very brief instant that it needs to switch. Amps are common. They also - fed 5 or 10v - apply the whole voltage to the MOSFET gate, yet can be triggered by a lower voltage from the microcontroller.
It sounds like what you want is a circuit board with some nice, juicy MOSFETs - and MOSFET driver IC's connected to them. Maybe with 4 channels on a board? You know, it just so happens that I sell exactly such a board... https://www.tindie.com/products/drazzy/4-channel-logic-level-mosfet-driver-100a-up-to100v/ Honestly, I don't recall seeing anyone else selling boards with MOSFET driver IC's on them, other than myself.
There almost certainly are people doing so (judging by these forums, use cases that ought to have a driver are pretty common), so either a lot of people are giving up on their projects or drastically scaling back their loads, or they're getting driver IC's, probably mounted on a breakout board - from somewhere. They're not all getting them from me (I wish they were!)
MarkT:
Lower frequencies are very obvious during saccades.
I understand, but the default 500Hz should be fine.
The default 200Hz,what I use from a PCA9685 controlled LEDs, is not a problem to me.
Most of us use a 50/60Hz LCD TV, which indeed flickers a bit.
So again, why 2500Hz, more than ten times what you really need for humans.
Seems like creating problems for yourself, without obvious benefits.
Leo..
I've explained this. If you stare straight at something 100Hz is probably fine, if you ever move your eyes
its not, you see very obvious artifacts. For things like car tail and indicator lights this really matters.
Several kHz makes the artifacts much less visible. The more point-like the light source the more this
matters too, so screens are the least demanding applications, individual LEDs the most.
There's no reason not to use something in the kHz range, there's no penalty with modern switching
devices. And if you want the thing to work through a video camera there's every reason to PWM at a rate
an order of magnitude faster than the frame rate.
MarkT:
The driver voltage should be appropriate for the MOSFETs, either 5V or 12V depending if they are logic level or not. Drivers need good high frequency decoupling, typically 100nF and 10µF ceramic for each chip, as they
switch high currents very fast. You can add some series gate resistance to reduce the switching speed somewhat to reduce EMI, something in the 10 to 100 ohms range might be good for 2.5kHz PWM.
Double check the drivers you choose are 3V compatible on the logic inputs, most are.
Thanks for your answer. I'm not sure whether I'll be using logic-level MOSFETs or not. If I use a driver with a separate voltage supply anyway, that's probably not so much of an issue, is it?
I understand that I need good decoupling capacitors. Will this suffice to provide all the charge needed to turn on the MOSFET? (Isn't that what decoupling is about?) In other words, will there still be current spikes towards the voltage supply when switching on, or should this be covered completely by the capacitors? If second, I guess the current requirements for the driver should be pretty low. Otherwise, how can I determine what current pulses the regulator needs to handle? The load current may be high, but only for a very brief moment because the gate capacitance is only so much. I don't want to overload the voltage regulator for the driver 2500 times each second, it won't be too happy about that.
I also understand that I can play with the gate resistor to skew the edges and reduce EMI (as long as the switching time is still fast enough and the MOSFET doesn't burn too much energy in the transition; I've seen formulas for that somewhere). I'll probably need to get some testing equipment for that. Maybe a simple SDR receiver or something. Could be useful for other tasks, too…
@DrAzzy, yes, I think need MOSFETs and drivers, on a board. But not just that. There needs to be more that actually puts some value in it all. There'll be a microcontroller, voltage regulator(s?) and a communication interface. I'm trying to make the design as compact as possible, so a multi-board design isn't what I'm looking for. And I was thinking of an 8-channel device if possible. I can design PCBs, have them printed and populate them. I just need to know how to design this.
Wawa:
I understand, but the default 500Hz should be fine.
The default 200Hz,what I use from a PCA9685 controlled LEDs, is not a problem to me.
Maybe not a problem to you and your behaviour. I've done tests with different frequencies in a very simple setup with the LED stripe illuminating the room. When quickly moving my hand in front of an illuminated wall, I can clearly see single frames below 1 kHz. And I do move quickly sometimes. I've read industry recommendations somewhere that a frequency above 2 kHz should be used to call it flicker-free for sensitive people (and they could be sensitive even without any movement; you never know what guests you'll have) and below 10 kHz to avoid EMI trouble.
That most consumer devices run between 100 and 1000 Hz tells me that they're cheap. It requires more effort to design this at the edge of physics, and it may require more expensive parts and more careful design. I'm by no means an expert at this, but I'm willing to learn the dependencies and how to manage them. And I want a good solution, not something half-baked that I'll worry about for years. Since halogen incandescent light bulb are increasingly harder to get (in the EU, that is, for now) and CFLs were bad from the start to their end, LEDs are my only option. LED stripes are the only practical option if you want some flexibility, and PWM is the only way to dim them. And as I'm forced to rethink room lighting now, I'm trying to get the most out of it. (Dimming, adjustable-white, automation etc.)
ygoe:
LED stripes are the only practical option if you want some flexibility, and PWM is the only way to dim them.
DC dimming is still an option, with a current controlled switching DC/DC converter.
Voltage controlled (or pot controlled) mains power LED dimmers don't use PWM for dimming.
Just did a test with my PCA9685 LED drivers.
Increased the PWM frequency from the default 200Hz to their max frequency of 1526Hz.
Did not see any difference in the room, only a slight difference in the strobing effect with flicking fingers.
As said, 200Hz doesn't bother me. Nor did the 50Hz from incandescent light bulbs.
Leo..
