1
\$\begingroup\$

(In advance, please pardon my ignorance; I'm still learning)

TL;RD. I want to power N number of 12V LED strips using a PC ATX PSU.
Is it safe & what bits do I need to add to make it function as intended?

Some context: I'm basing my work on 2 pretty decent articles on the subject:

What I hope to achieve is to control an arbitrary amount of 12V (common anode) LED strips, & from what I can tell they seem to be in the region of 14.4W per metre (x 5m = 72W), based on LT12560-RGB 12V, 60 LEDs @ 4.8 or 14.4 W/mètre x 5 mètre = 24 or 72 Watts and 300 LEDs, although in the end I'll be going with cheaper units as I need to run close to 16 of these 5 mètre strips concurrently.

I bought some Arduino-friendly MOSFET boards (duinotech XC-4488) to do some prototyping before risking frying larger devices (such as the Mega I bought for this project), then some manual bread boarding using NMOS for PoC & testing, but soldering & running 48+ MOSFET is cost-prohibitive & there's probably a better way of achieving this outcome by using an IC, thinking 3x MAX6971. Effectively I'm making use of them as glorified relays attached to PWM pins.

I've sketched it out via Fritzing & went as far as running a simulation via TinkerCAD (although due to limitations I couldn't recreate it 1:1).

Fritzing sketch for LED array

& the final build looks like: (keeping it simple, since relevant resistors are build into the MOSFET boards from what I can tell)

enter image description here

I did some basic tests before firing up & it seemed to have checked out OK, but I was VERY much mistaken, since once I plugged into the ATX PSU, it wasn't long before I got the LOUD POP, flash, magic smoke, funky smell of ozone & the lights went out in the house. So I'm unsure if it was my ATX PSU that's faulty, but it seems far more likely that I screwed up somewhere.

Needless to say I'm very reticent to pop another ATX, and worry that my invention may be a danger to myself & others. Luckily I wasn't running my Arduino from the same PSU via Vin (is that legit?)

I'd really appreciate some help here, as I suspect I'm not officially out of my depth.

\$\endgroup\$
  • 2
    \$\begingroup\$ Most atx supplies won't run the 12V unless there is a load on the 5V. Aside from that, your wiring in that picture seems good for test. You wouldn't want 72W going through a breadboard though. Don't know what killed your supply, maybe you shorted something out accidentally? \$\endgroup\$ – Passerby Jun 10 '18 at 1:28
  • \$\begingroup\$ Hi @Passerby. Thanks for much for the quick ping-back! A suspicion may be that the ATX was faulty, since it's pretty old & been sitting in a box for a while, but my own limited knowledge seemed more probable. Noted re the breadboard track, thanks - will wire to the MOSFET board directly. Re the 5V draw, in theory, once it's all wired up, my Arduino should pick up that load via Vin OK? In the meantime I'll just put an LED & 10k resistor? \$\endgroup\$ – jakes Jun 10 '18 at 1:47
  • \$\begingroup\$ (also) Should I have another 10k resistor from +12V to ground? A possible suspicion may be that LED strip in there (an off-cut I had, not my final device) may either been popped, not working for some other reason, and/or the critical threshold was not triggered for it to turn up 7 use the juice. \$\endgroup\$ – jakes Jun 10 '18 at 1:47
  • 1
    \$\begingroup\$ These supply use loads in amps, an led or 10k resistor isn't enough. \$\endgroup\$ – Passerby Jun 10 '18 at 2:01
  • 1
    \$\begingroup\$ Also it seems we can no longer assume all ATX supplies behave the same for any switched load we throw at it. SMPS prefer more stable loads unless specifically designed for that. Instabilities can occur with loose long wires. Next time do a stability test with an 8W light bulb or MOBO and add a low ESR electrolytic or plastic cap to the power input of the MOS power input screw terminals and keep wires paired close together CMOS PWM signal with ground signal and tell us what coded frequency you are using. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 10 '18 at 16:24
3
\$\begingroup\$

72W at 50% PWM on 12V is 36W at some low frequency like 1kHz of AC ripple current which is excessive for most decoupling caps in ATX PSU’s.

If output caps have a dissipation factor which determines what % of load ripple power must be absorbed by the cap. Aging increases this rapidly.

Although a typical MOBO has dynamic CPU loads the change rate in frequency of load cycles is much lower.

Conclusion;

Excessive ripple current for YOUR PSU. Overheated cap, slowly turned into short circuit and faulty circuit protection. Get a better LED driver supply.

\$\endgroup\$
  • \$\begingroup\$ Thanks for that, @tony-stewart-older-than-dirt. (Please bear with with me; I'm a novice, so taking some time to pick up what you're putting down) Pushing ripples up to the PSU makes sense, thanks. I recall seeing something akin in the TinkerCAD simulation I created & was unsure how to resolve it, opting to go with how build was indicated in the articles ref'd. For smoothing out ripples, is that not what capacitors are for? My knowledge re caps are non-existent (so something I'll look up), but in your expertise, what would be appropriate, where & how much? \$\endgroup\$ – jakes Jun 10 '18 at 1:56
  • \$\begingroup\$ All ATX PSU's have OCP with safe shutdown on short circuit. A low ESR cap was missing in the photo to handle current risetime until SMPS can supply the current in xx us. Capable of such high ripple current such as 0.1uF to 1uF Polypropylene or ceramic or any other suitable type on LED driver board. – Choose C much lower ESR than RdsOn of MOSFET but much bigger C than Coss and FET has much lower RdsOn than ESR of Load \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 10 '18 at 12:52
  • \$\begingroup\$ Those flimsy cables may not be able to triger OCP on 500W+ ATX PSU, even when shorted. Note that may ATX PSU designs also need a minimum load on 3,3 and 5V rails to work properly. \$\endgroup\$ – Turbo J Jun 10 '18 at 13:16
  • 1
    \$\begingroup\$ Agreed @Turbo but that will never cascade faults to line with DC traces to cause "the lights {to go} out in the house" . Also modern ATX designs dont need a preload to operate. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 10 '18 at 13:37
  • \$\begingroup\$ Never say never. I can totally see the regulation loop going crazy without any 5V load and blowing up the primary MOSFETs as a result. Just to be clear: That is no supposed to happen, but component failure is quite common in cheap and old ATX PSUs, unfortunately. \$\endgroup\$ – Turbo J Jun 10 '18 at 13:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.