Posted: 2/14/2026 2:48:05 PM EDT
[Last Edit: 2ANut][Edited]
|
For background, I am a career software engineer and occasional electronics designer/tinkerer. I have a custom home "dashboard" system that uses a 32" touchscreen monitor and a Raspberry Pi 5, wall-mounted. Software is all C#/.NET stuff. My app has a "sleep" mode that blanks out the display. But the monitor's backlight remains on so the display still visibly glows if the room is dark. I want to add software control over the backlight power circuit using a 3.3V GPIO from the Pi and a MOSFET circuit, that way my app can shut down the backlight without actually turning the monitor off. I've opened up the monitor and identified the power leads to the LED backlight. It appears to be 45 volts DC. Current is unknown but probably low--the entire monitor runs on about 25 watts at full brightness. I have not yet measured current because I'm not quite ready to cut wires. I know how to design and build basic electronics circuits, but I have almost no experience with MOSFETs and I don't really understand MOSFET datasheets. Digikey's MOSFET selection is overwhelming and I really need some help narrowing down the options. So, I'm looking for suggestions for specific models of MOSFET that will help me meet these goals. Once I have some parts, I'm confident I can design and build a circuit myself.
So, any suggestions? |
| I might be overthinking things. I have 12 VDC input on the monitor...I think if I use a BJT as a driver to switch the 3.3V GPIO output up to 12V, I can use that with just about any MOSFET with a drain-source voltage of > 45V and be fine. Plus the BJT will give me the inverting logic I need so the backlight remains on if the Pi isn't present. I can tuck the whole thing inside the monitor and run a control/ground pair out to the GPIOs, and the monitor ought to act normal when those are disconnected. |
|
Maybe something like this? https://www.digikey.com/en/products/detail/taiwan-semiconductor-corporation/TSM900N06CH-X0G/7360597 60V rated for 11A or 25W, exceeds my requirements. Drive voltage is listed as 4.5V, 10V (which I'm not sure how to decipher) but I figure if I drive it with 12V it should work fine? I've already got a bunch of BJTs in stock, probably just pick anything to invert the 3.3V GPIO and use it to control the MOSFET. Looks like there are two separate leads to the backlight so I guess I need two MOSFETs in order to keep those separated. |
|
Originally Posted By targetworks: I'm not sure which AIs have demonstrated skill at answering circuit design questions, but have you considered presenting your requirements to an AI to see what it may spit out? Maybe Claude or Perplexity?  I have no AI accounts. My offline Ollama install was no good at providing definitions for MOSFET datasheet values, and online AI-powered search results were even worse--those gave incorrect or incomplete definitions sourced from misleading Reddit posts from 10 years ago. I came here because I need human help and not AI help. In any case, I ordered a batch of those parts linked above. I can play with a few on a breadboard and a benchtop power supply to see if they'll do the job before I cut up the wires in my monitor. |
|
Originally Posted By 2ANut: I might be overthinking things. I have 12 VDC input on the monitor...I think if I use a BJT as a driver to switch the 3.3V GPIO output up to 12V, I can use that with just about any MOSFET with a drain-source voltage of > 45V and be fine. Plus the BJT will give me the inverting logic I need so the backlight remains on if the Pi isn't present. I can tuck the whole thing inside the monitor and run a control/ground pair out to the GPIOs, and the monitor ought to act normal when those are disconnected. This is exactly how it's done if you don't have a secondary mosfet controller IC. There's two ways to do the switching, high side and low side. Low side is easiest and you just a a N-FET and that switches when the gate is pulled up, but low-side is difficult if you have any other connection to the device -- you're basically ungrounding the power connection. With high side, you use a P-FET and a pull-up on the gate, and use the transistor (NPN) to pull down the gate to turn the P-FET on. This is what I would do. |
|
I ordered some generic MOSFETs from Amazon that would ship quickly. Started with N-channel MOSFETs and tried low-side switching. Didn't work right because there's apparently a PWM controller behind the backlight "ground" line I tapped into, and after a few revisions there was nothing that was reliable. So I just ordered the complementary P-channel MOSFETs from Amazon, along with some 2N2222 transistors which are 60V rated. My next plan is to send the 3.3V signal to a 2N2222 to switch the MOSFET gate between 45V and ground, and do high-side switching. I have no idea what these parts are. Definitely Chinese origin with poorly-translated specs. But I measured the backlights. There are two backlights and they draw 250 mA each at full brightness, so I'm definitely not dealing with much power here. The P-channel ones on the way claim to be good for 100VDC and 23 amps, at about 80 cents a piece. The annoying part was my original N-channel build worked perfectly when it was on a breadboard and using a benchtop DC power supply for part of it. But when I soldered up a protoboard and tied it all together, I ended up with a PWM-modulated ground connecting to other grounds, which caused problems, and attempts to isolate them only caused more problems. The new P-channel version won't touch that PWM "ground" at all and will hopefully work as expected. New schematic. Resistor values are arbitrary and intended to keep BJT and GPIO currents as low as possible. I can always reduce resistor values and raise currents if I notice any flakiness when breadboarding. 
|
Here's some pictures so far. This dashboard has existed for years, but was always on a non-interactive display tucked in a corner. Now that it's on a touchscreen, I'm adding all kinds of things. C#/.NET using AvaloniaUI as the UI framework, running on whatever Linux OS the Raspberry Pi's currently use. The calendar is hardcoded to pull from a few iCal sources (including a davical server running in my basement). There's a detailed weather view with some graphs, using data pulled from an open source online weather API:  This view uses an embedded browser frame (Chromium Embedded Framework) to bring TornadoHQ.com in-app, since I live in a place where severe storms and tornadoes are common. TornadoHQ is the best site out there for keeping tabs on severe weather.  Art/slideshow view pulls random pictures from a folder in the filesystem and changes every 30 minutes. (All I have to do is SCP or rsync new images to a specific folder and they'll get pulled in.) I overlaid a gray square over the image widget to greatly reduce the brightness. During normal daylight, the screen ALMOST looks like a painting or something. Definitely doesn't look like an interactive screen, which I like.   Here's what's left after removing the shell/bezel and all unnecessary components, then putting what remains into the picture frame. The "brain" is a 16 GB Raspberry Pi 5 with big passive heatsinks and no fans, to the right of the display.  I prepared a spot on the wall near my front door in the living room.  The in-wall media box has a 120V outlet and a CAT6 port wired up to my core switch, so all cords will be hidden. The box has a cutout in the top so I can run cables through the wall up to the screen. The Raspberry Pi and the monitor's power supply will live in here.  Test-fitting everything. I didn't have the outlet in the media box wired up yet so I had to run cables out. Final install won't have that. I simply used four 2.5" trim screws straight through the picture frame into studs, and it holds everything nice and well.  Not pictured is a "sleep" mode that blacks out the display. In its current form, the backlight remains on, which means the screen glows in the dark. So the purpose of the MOSFET circuit is to cut power to the backlight in this mode, but without actually turning off the monitor. The touchscreen remains active and this mode has a fullscreen button so a simple touch will restore the backlight and flip back to the main dashboard views. |
|
Originally Posted By 2ANut: I ordered some generic MOSFETs from Amazon that would ship quickly. Started with N-channel MOSFETs and tried low-side switching. Didn't work right because there's apparently a PWM controller behind the backlight "ground" line I tapped into, and after a few revisions there was nothing that was reliable. So I just ordered the complementary P-channel MOSFETs from Amazon, along with some 2N2222 transistors which are 60V rated. My next plan is to send the 3.3V signal to a 2N2222 to switch the MOSFET gate between 45V and ground, and do high-side switching. I have no idea what these parts are. Definitely Chinese origin with poorly-translated specs. But I measured the backlights. There are two backlights and they draw 250 mA each at full brightness, so I'm definitely not dealing with much power here. The P-channel ones on the way claim to be good for 100VDC and 23 amps, at about 80 cents a piece. The annoying part was my original N-channel build worked perfectly when it was on a breadboard and using a benchtop DC power supply for part of it. But when I soldered up a protoboard and tied it all together, I ended up with a PWM-modulated ground connecting to other grounds, which caused problems, and attempts to isolate them only caused more problems. The new P-channel version won't touch that PWM "ground" at all and will hopefully work as expected. New schematic. Resistor values are arbitrary and intended to keep BJT and GPIO currents as low as possible. I can always reduce resistor values and raise currents if I notice any flakiness when breadboarding. https://i.imgur.com/84JJy6f.jpeg This schematic does not work. A gate-source voltage of 45V makes these MOSFETs literally explode and spray shrapnel. Not fun. I had to create a voltage divider network between the BJT collector and 45V so that it can pull the gate down to about 35V instead of all the way to 0, and that avoids letting out the magic smoke. Also, the MOSFETs are opposite of what I expect so I need an additional BJT stage to invert the GPIO logic a second time. But my next problem is this: the monitor has two backlights (one on the left, one on the right) and it doesn't like it when you disconnect both. I can confirm this by just playing with the backlight leads and none of my circuitry. As long as one backlight remains powered, I can cycle the other with no problem, using either my MOSFET circuit or connecting by hand. But if both backlights get shut down, neither will turn back on until the monitor is power-cycled. Something in the monitor's backlight control circuit goes wonky when both are shut down. 
|
|
As best as I can deduce, the 45V supply has a protection circuit that shuts it down if it goes no-load, and then the entire monitor needs a power cycle to restore it. I tried putting some resistors between +45V and ground to trick it and that didn't help, so it must measure voltage drop due to the LEDs as part of its protection. If I want to continue with a GPIO-controlled backlight, I either need to provide my own 45V supply (I was not able to power the backlight with my 30V benchtop supply so I can't undervolt it), or I need a chain of diodes and some resistors to switch the 45V supply to when the backlight is off in order to provide a load and trick it... Which is a lot of effort to avoid pressing the monitor's power button when I want to turn it off. I put it back on the wall and I'm just going to use it as-is for now. |