Request for feedback on LED PWM driver output circuitry

[daviegravee]

I am looking for feedback, thoughts and general input on a design I am starting out on. I am designing a 24 channel LED driver (capable of driving 6 x RGBW strips). I am designing the output stage now. I have attached a screenshot of my LTSPICE file that simulates one single output stage. There will be 24 of these in total.

The key points of the current design are:

• PWM, constant voltage driver;
• Each channel must be able to handle 4A, at 48V (24V will be more common in practice);
• Needs to operate in ambient environments of up to 55C;
• Switching MOSFET is the DMNH6021SPDQ : https://www.diodes.com/assets/Datasheets/DMNH6021SPDQ.pdf;
• MOSFETs are driven by MCU pins (no gate driver IC);
• Minimum switching frequency will be in the range 200-400Hz;
• Maximum switching frequency will be in the range of 15-20kHz;
• The only heatsinking method that will be used in this design is copper of the PCB (see page 2 of the datasheet for thermal characteristics). I’m not up to layout yet, but I will be using the 2oz copper, 1-inch square copper plate to achieve a Junction to Ambient thermal resistance of 54C/W;

Thank you in advance!

[daviegravee]

It looks like some of my attachments got missed. Here's a zip containing the simulation file and the datasheet for the switching MOSFET.

[ArdWar]

MOSFETs are driven by MCU pins (no gate driver IC);

Pick logic level MOSFET with lower V_GS, or get better gate drive. 3V max V_GS(TH) is too high for comfort. Just look at the curves with V_GS=3.0V!
Or maybe just get better gate drive anyway. Depending on your MCU's output drive strength, the switching losses may be quite noticeable with even with low Q_G.

[PCB.Wiz]

... There will be 24 of these in total...
...Each channel must be able to handle 4A, at 48V (24V will be more common in practice)

That's significant power being turned on and off, you will need to consider parasitic inductances of a practical wiring system.
Some clamping may be needed on the FET, and to manage RFI you may need to carefully slow the gate drive slew rate, and trade off some loss for lower RFI.
An increased gate resistor is sometimes enough.

As already mentioned, 3v3 is a bit feeble for MOSFET power drive, and with that many amps & channels, you need a gate driver with some ground bounce difference tolerance.


eg a newer gate driver like UCC44273 (etc) Single-Channel Low-Side Driver has added features like
• 5-V UVLO
• Ability to handle negative voltages (–5 V) at inputs
• Fast propagation delays (13 ns typical)
• Fast rise and fall times (9 ns and 7 ns typical)
• 4.5-V to 18-V single supply range
• Outputs held low during VDD UVLO (ensures glitch-free operation at power up and power down)
• TTL and CMOS compatible input-logic threshold (independent of supply voltage)
• Hysteretic-logic thresholds for high-noise immunity

You might even consider opto isolation, to keep the noisy power signals completely away from your MCU.

[daviegravee]

Thanks!

Some clamping may be needed on the FET

Voltage clamping between source and drain? Is this due to overshoot that might occur from wiring inductances?

That's significant power being turned on and off,

I should clarify: any single channel needs to be able to handle 48V of LED strip, taking in 4A (192W of strip) but the design will not be rated for 24 x 192W simultaneous drive, and certainly not 192W dissipated by the MOSFET. Your point still stands, but I think my original wording of the spec might be misleading.

[daviegravee]

I'll start hunting for a gate driver IC.

Regarding switching times and switching losses, I understand that slower switching time means more power dissipated by the MOSFET, however the flipside is greater EMI as switching speed is increased.

I'm comfortable calculating (or rather, simulating) switching losses as a function of gate drive speed but i'm pretty inexperienced with EMI. Do you have any resources or wisdom to impart here? Should I be adding filters to the output? Anything to add is appreciated, even a link to an app note.

[BennoG]

I use attached schematic.
There is a feedback to the MCU that measures the average current of the LED.
This is to detect led defects.
O the LED voltage LEDVCC can be between 12 and 48V

For the next generation I will add a snubber circuit to prevent ringing on the wires to the led

Benno

[PGPG]

Switching MOSFET is the DMNH6021SPDQ :

Vgs threshold = 1..3V so typical 2V.
Figure 1 is for typical transistor (Vgs=2V)
For worst case (Vgs=3V) line you see there for 3V will be true for 4V.
From Figure 1 it looks that this transistor (typical) is really switched on for Vgs=4V so for worst case 5V is needed.
Of course you can buy 10 times more transistors than you need and select those with small Vgs threshod.

[Zero999]

I echo the comments above, regarding logic level MOSFETs and drivers.

It'll probably work, but drop a but more voltage than it should, meaning it might overheat if operated at full power for too long.

If you're going to simulate it, then go for the highest frequency you're going to use. I also think the rise/fall times of the input are optimistic. I suggest you look up the data sheet of the micro controller you're using. And please just upload the.asc file, rather than a zip.