I'm designing a custom PCB for an LED dimmer. It works with a strong 24 V power supply and feeds 8 LED channels. In my first configuration, two channels are combined for tunable white, so there are three wires going out (+24V, -CW, -WW) and that times four. Each channel is switched by a MOSFET and has a protection diode. I'll use wide enough PCB traces and wire bridges to handle the current.
The PSU delivers up to 13 A and each of the 8 channels is designed for 2.5 A (but not all of them obviously), depending on the LED strip length I'll install. 2 channels already work reliably and safely in a prototype with ~1.3 A/ch.
Now I'm seeking to add some protection to the circuit. I think I need a fuse to protect the PCB and MOSFET from over-current. The PSU has its own protections and the wire leading to the LEDs can handle more than that.
Is it enough to put a fuse into each cathode line between the LED cable and the MOSFET? I'm looking at a resettable fuse with Ihold of 2.5 A. The model "ESKA FRU250-30F" is currently in stock. It has an Itrip of 6 A after ~10 seconds if I read the datasheet correctly. Is that a suitable protection or should I use a fast single-use fuse? I'd prefer smaller footprints if possible.
What purpose should such a protection have? Just short-circuits on the LED line? Or maybe also induced transients from a nearby lightning event? The LED cable might be a few (< 10) metres long.
Or do I need no fuse at all if the MOSFETs can handle all the current the PSU can provide, so the PSU's internal protection (or a single fuse for the PSU line) would be sufficient? In that case the entire dimmer device would go down in the event of a fault, instead of just the affected channel. Please advise.
Fuses are there to protect the insulation, mainly wire, from melting and causing fire. Fuses take way too long to open for it to protect any solid state device. If you want to protect those devices, then you need a circuit with speed fast enough to protect those devices.
So if my cables and connectors can handle the full PSU current, I need no fuses on the board? I can only find the usual calculators for voltage drop on wires. That would be something around 15% in my case. There's no information about how warm the wires would get, but probably not relevant. I'll still need to look up the maximum current of the connectors (Molex MicroFit 3.0, 3 pins).
A fuse could protect your wiring & PCB in case of a short. It may or may not protect the MOSFET or any of the active circuitry.
If it makes you "feel better", go ahead and add fuses.
A good power supply will have a built-in fuse (usually on the AC side) or some kind of built-in protection so the power supply itself usually won't catch fire if the output is shorted.
If you do get a short you'll get MORE than the 13A power supply "rating" for a moment before the fuse blows or the power supply shuts down or dies. Or sometimes a trace will burn-off of the circuit board first (basically acting as an unintentional fuse, cutting-off power).
If there's not much chance of a short you should be able to get-by without fuses,
Sure, here you go. Note that this isn't complete yet. Pull-down resistors for each GPIO are still missing, and in this version the LED connectors are still all separated. That vertical LED+ line will be done as a wire bridge instead of a PCB trace.
I don't "feel" much here because I haven't seen fuses in action a lot yet.
The power supply will be a Mean Well HLG series which is approved for this use. It should come with decent protection to not die or burn. I haven't tried it out yet though.
Well, normally there isn't. But this application involves soldering wires to an LED strip and mounting crimped connectors to the wires. There is a small risk of making mistakes here when building it that might cause problems when air humidity increases later or some soldering points change with age or temperature. Once installed, things shouldn't be touched anymore. The dimmer device is intended to be installed together with its PSU in a closed case in the wall or above the ceiling. The LED wires should go inside the ceiling or wall, too.
After reading a bit more about this, I think I might just place a simple fuse in the common anode wire to each LED strip, outside the dimmer device. That is easy to build and to replace in case of a fault. Plus, a well-dimensioned fuse can be selected at installation time when the actual installed LED strip power is known. (The dimmer device can handle 2.5 A per channel (as per my spec) but for a shorter 1 A LED strip, a smaller fuse would be appropriate.)
Is this correct and advisable?
If something goes awry and the dimmer PCB's traces see too much current in a place where it shouldn't be, but the PSU will still provide that, then the dimmer might be destroyed. But hopefully it shouldn't burn.
The best place for your fuse is in the drain lead of the MOSFET. Be sure to oversize your MOSFET so it will blow the fuse. I would use a fast blow fuse. You can look at the fuse blow curve and see if the MOSFET curves will exceed that. While looking at the MOSFETs check for avalanche rated, that will eliminate the diodes. I would not use your current MOSFET as it is borderline at best. The output of the processor is 3.3V max, not enough to fully enhance the MOSFET. Look for a logic level MOSFET that is on (enhanced) at 3V or less. I would also recommend using something in the 25-50 ohm range in the gate leads. Your 1N4148 will be ok with LEDs but if you get some inductance they will probably blow. Your flyback diode needs to adsorb the current coming back from the load. For an inductor that is about what you put in.
Thank you for your review of my schematic. A few notes:
This device is intended to only drive LED strips and nothing else. The current in one channel must not exceed 2.5 A, I will not use longer strips than that. (Maybe I'll even reduce it to 2.0 A.) The wire length to the LED should not be too long (up to maybe 5 m, I don't know that).
I thought I already selected this MOSFET by its "logic level" feature. I assembled a prototype PCB with two of exactly these channels, with an additional MOSFET driver of two transistors (still the same 3.3 V, just more current than a GPIO) to play with. At EMI measurements (the homemade method with an SDR from <1 MHz up to a few 100 MHz) I found that using the driver interferes with FM radio! So I removed the driver from my plan; also simplifies the layout and BOM. By calculations, the gate resistor must be at least 150 Ω to not overload the GPIO. Even that caused severe short-wave interferences from 5 to 15 MHz. The 1 kΩ resistor is clean, no EMI problems. It's also still fast enough so that I can get all the levels from really dark night light with fading to off (at dynamically reduced PWM frequencies).
That prototype runs with an LED strip of total 1.3 A and ~3 m wire at all dimming levels (0 to 100%) and different PWM frequencies (300 to 2500 Hz). It's been running (w/o the driver and w/ the 1k gate resistor) for several hours now and doesn't even get warm anywhere. That makes me believe that the components should be good. (Before the EMI measurements, it even ran for days with the driver.)
I made some measurements on the scope without the diode attached to the board and saw concerning voltage spikes when switching off the LED strip. After soldering the diode to it, those spikes were completely gone. I've done some (terribly bad) spice simulations before which made me aware of the problem. I was suggested to add that diode and it helps.
I'm still a bit unsure about the fuses. I thought I'd get glass fuses and an inline fuseholder but can't seem to find a suitable holder. They few available all come with ≥2 mm² wires, but I use 0.25 to 0.5 mm² and my connector can't hold more than 0.8 (AWG 18). Also, there are fuses for 4.0 or 5.0 A available. That's why I might go with the 4 A. Just need to consider how to build the lights (LED strip lengths, aluminium profile lengths, clippings etc.). The resettable fuses are even only available for 5 A.
There are complex dependencies between all parameters…
PS: I set the current limit from the description of my intended LED strip, saying it can be used at 3 m maximum length. With specified and measured 1.5 A/m that makes 4.5 A maximum current through the strip. Hence my fuse size.
