In 2007, I and several friends dreamed up PartyMode, a sound and lighting/visualization system created with top-of-the-line mid-2000s technology. About 22 speakers were strewn around a lab, a VU meter made of LED-illuminated case fans was controlled by a parallel port and mounted in the center of the room, and some simple software allowed the CRT monitors of every computer in the lab to be taken over as music visualizers. Unfortunately, the building in which this system was built was eventually torn down, and to my unending frustration, I took few photos and even fewer videos of the system. In the mid 2010s, I recreated a miniature version of this system as PartyMode 2.0 in my bedroom, retaining the LED case fan VU meter, but adding a quartet of bright RGB LED elements around the room in place of the monitors as music visualizers, and using the 5.1 speakers of a single sound system. Now, I've moved to a new place, and it's time to think about the next version of the system.

Perhaps unsurprisingly, I'm no longer interested in custom-designing constant-current sources for a few high-power LEDs. Instead, I'm using 60-LED-per-meter WS2812B LED strips, which will be run around the top perimeter of my room (as well as inside shelves and above my desk) and will allow me to replicate many of the original PartyMode 2.0 modes, including music-reactive, low-level nighttime illumination, and the always-popular Candlelight Mode. Controlling these is easy and well-documented, but powering them is a little more complicated. I want to power what in the end is likely to be at least 70 feet (21.3 meters) of LED strip, or 1280 LEDs, which at full brightness is up to 60mA per LED = 77A. I'll need a fair bit of power if I want to run everything at full brightness at the same time (which, admittedly, is unlikely).

The purported 15A 5V power supplies on Amazon come with extremely poor reviews, so of course my thoughts jump to that power supply jack-of-all-trades, reusing ATX PC power supplies. I've done my homework, and discovered two interesting facts:

• Older power supplies tend to have more 5V current available relative to their overall rating, which is what I need.
  
• Most of these switching-mode ATX power supplies have a minimum 12V load required to regulate the 5V rail properly, which is not what I need

Other than the 12V rail load requirement, a few of the good options I've discovered include the Zalman ZM400B-APS (for example, this one on eBay (5V @ 40A), assorted Logisys supplies (like this one, 5V @ 36A but with a minimum 1A draw on both the +5V and +12V rails), the ThermalTake PurePower 480AD (like this one or this one, 5V @ 40A), or the Antec True330 like this one, 5V @ 30A but an 800mA minimum 12V load.

Any thoughts, either from general electrical engineering expertise or experience with this type of project?

70A at 5V is completely unreasonable. You probably need to deliver power individually to each strip.

Looking at wire sizing guidelines, for power transmission at 70A you'd want at least 3 AWG wiring, which has about a conductor diameter of about a quarter inch- that's an enormous cable. And pushing all that power straight through 20 meters of LED strip will definitely fail- a quick calculation on just 1m of PCB with 3oz copper (~.1mm thickness) says you'd need a track 37cm wide to keep the temperature rise to 10 degrees C. In reality you'd probably lose around 100 Watts and start a fire trying to do that all with one rail.

For large coverage I suspect you'd do fine and save a lot of power with 30 LEDs per meter, but if you want maximum fanciness then you'll just have to solve the power problems.

I think you'd want local regulation with higher voltage distribution. Just in a quick look for things that might be useful I see there exist LED strip power tap kits that sit between strips and allow you to inject power with a local regulator- this seems exactly like what you'd want.

I'd probably look at 12V or 24V distribution with local regulation at each strip to get the required 5V.

I'm sure you're well aware, but when powering that many LED's, it will be advantageous to supply power at multiple locations along the strip. As you can imagine, the little traces on the strip are not made to carry 80 amps on one end, and voltage drop will eat away at the blue and green led brightness.

Because there aren't super affordable (or safe) 80A 5V power supplies (need 6 or 4 gauge wire? Yikes), it might be a better option to get a small 6-8A power supply for every 100 LED's or so. Multiple supplies makes things cheaper and keeps the color accurate, just separate the positive rails, and keep the ground referenced between the supplies.

Another thing of note is the number of LEDs in the chain. You need 3 bytes of memory for each LED, so you're looking at close to 4kB needed on your microcontroller. It also takes about 30microseconds to send the data for each LED, so with 1280 on one string, you're looking at ~39ms to update the whole string, or about 25FPS. You could split the string into multiple strings to increase performance, at the cost of extra processing.

Ah yes, I omitted two key points here:

• I had no intention of trying to deliver all of this power from a single supply, let alone at a single point. I anticipate having probably 3+ power supplies (if I use ATX PSUs), with injection points every ~100-200 LEDs.
  
• I also don't think I'll be controlling these all from a single Arduino: I might end up with a non-Arduino controller, or I might end up with several secondary controllers controlled by a master controller

I'm afraid I'm going for maximum fanciness - I'm thinking I can just scale down the brightness of each one and have a cleaner result (especially with diffusers) than 30 LEDs per meter, if I want the equivalent brightness to a 30-LED-per-meter setup.

This seems pretty ideal to me. If I'm doing my math correctly, if we assume a total of 85% efficiency from wire loss and these little guys (5V/12V) * (70A/85%) = 34A @ 12V, well within the capability of a single modern 600W PSU (although I'd still probably split it among at least two supplies, which would be 9-10 AWG wire for a single run from each supply, or 12AWG if I split each of two supplies into 2 runs (up to 9.3A in each of those 4 runs).

Edit: Obviously the gauges get even friendlier with 24V distribution, but I'm not as familiar with good sources of 24V power and 24V->5V step-downs.

You don't save a ton of money going with a PC power supply over a more "industrial" power supply so I'd question whether you really want to use a PC power supply, but it seems fine. With some hacking you could pretty reasonably install a local power switch on a PC supply and only expose 12V outputs. As a bonus you could tap the 5V standby rail to power the system controller!

An alternative that might be pretty neat but more expensive would be to use power over ethernet. With 802.3bt you could get the required power in 8 ports or so- and that would give you ready-made data connections to each segment. However, I suspect any PoE switch capable of supplying that much power is going to be quite expensive.

Cost was my primary motivator: if not a PC power supply, what kind/brand of more industrial supply might you recommend?

I didn't have anything in particular in mind, but I saw that 500W 12V supplies on DigiKey go for about $100, which seems pretty inexpensive after factoring in the digikey tax- pretty comparable to a good-quality PC power supply.

But if you were planning to go for a cheaper or possibly secondhand PC supply, that's probably hard to beat.

I'm thinking heavily about using a couple of cheap or secondhand PC supplies, including to get 5V for my controllers. Running the numbers again for sanity, 4 12V runs split across 1280 LEDs is 60mA * (1280 / 4) * (5 V / 12V) / 0.85 = 9.41A, barely outside the 9.3A margin for 12AWG wire (or well within the margin of 12A for 11AWG or 15A for 10AWG). If I were to get a pile of these buck converters (any opinions on these?), and take their 2A continuous rating as a maximum (even though I doubt I'll be frequently if ever running every LED at full brightness), I'd need one to inject power every 33 LEDs, or 39 of them total.

Incidentally, I'm planning to put my LED strips into this type of channel/diffuser, which I'm hoping I can use as a heatsink for the buck converters.

That seems like an okay step-down converter; at least it's not a completely no-name brand. 40 of them seems excessive though: why not go for 10A output in a nice case?

20A might be a little more convenient yet, but at 10A you could use 5 meter LED strips and simply inject 10A from either end to reach the maximum power requirement for the entire strip.

The channel/diffuser you found would work great for diffusing the head from the LED's, and maybe the converters too. I would be wary of trying to put 5-10 amps through the strip itself, even from both ends, as the traces within the strip would be hard-pressed to be able to handle that current. Using multiple smaller buck converters may be a good way to go. Putting multiple buck converters in parallel with diodes to prevent them back-feeding eachother would provide a nice layer of redundancy, if you are concerned with these cheaper buck converters flaking out on you. Here are some 15A low-dropout (<0.5v) diodes.

On the other hand, 12AWG wire is good for 15-20 amps, depending on the length of the run. 16AWG wire (cheaper and lighter) is still goof for 10-12 amps. So you have that going for you, which is nice.

Two reasons made me lean towards more low-power vs. fewer high-power: (1) being able to fit them inline within or roughly within a diffuser strip, and (2) passing less current through the traces in the strips (see next section).

Yes, I was thinking that too, especially with Tari's previous computation that 77A would need 37cm-wide tracks (and the 18A for one strip is an appreciable fraction of that). I'm inclined to want to inject power at more points within the strips, using 10A / 12V backbones with proper-gauge wire paralleling the strips that have periodic taps for buck converters of some capacity.

It depends on the failure mode: if they fail to off, I don't care too much, but if they fail to passing the input voltage through, that could be problematic. Definitely one potential argument for

I'd love to go for thinner wire if it's safe, but the charts I could find for power transmission suggested not. Could you point me to where you're getting that information?

I'm also aware that if I'm tapping these transmission lines, after a few taps the current will have dropped to where I can potentially use thinner wire for the rest of the run.

Try this chart, based on the US National Electric Code
12AWG wire is used for 30amp circuits in houses, and 20-25 amp circuits in industrial/commercial buildings. 16AWG wire is used in extension cords on a 15 amp circuit, and should be completely fine for 10 amps, especially when you are using a buck-converter on the end that doesn't particularly care if there is a little voltage drop. (Though, you shouldn't see that much drop anyway.)

Also note that when searching for these things, "Power Transmission" may be referencing to transmitting power for miles, not just a few feet. The current rating for wires of that length are completely different. Also note that DC power does not suffer from the skin effect, so a wire rated for a certain current in AC could actually handle more than that current in DC.

I'm afraid I'm real skeptical of that chart: it contradicts most things I've read and my intuition.

So annoyingly, you actually need larger-gauge wire for lower voltage, because of loss per unit length. e.g. 16AWG copper wire has a nominal resistance of 6.4mΩ/ft, so over a 100 foot (round-trip) run, the total resistance is 100*6.4/1000 = 0.64Ω. At 10A, that's a 6.4V drop (!!), which is a >50% loss for 12V, but only 5.8% for a 110V system. I believe my longest run will be about 60 feet round trip, including the distance in parallel with the LED strip and leading up to the strip: for a maximum 10% loss in the wire, that would be a maximum drop of 1.2V at 10A, or at most 0.12Ω total loss. Over 60ft, that means at most 2mΩ/ft, which annoyingly corresponds with 6AWG wire (see chart). For later in this discussion, let's call that 10A continuous over 60 feet, or 600 amp-feet: (ohms/feet) * (amp-feet) gives us ohms * amps, or volts (because the feet cancel) - trying this calculation again, 600 amp-feet * 2mΩ/ft = 1200 mΩ * A or 1.2V loss, as we calculated.

However, I think we can be clever here and think about the fact that I have 60 LEDs per meter, or ~18.3 LEDs per foot, grabbing some of that current. The wire to the beginning of the LED strand is ~10x2 = 20 feet total, so the entire ~10A needs to travel that distance with minimal loss. However, we then lose an average of 1.098A at 5V per foot along the LED strip, or with 90% buck conversion efficiency, 0.51A at 12V per foot. Thus one foot along the LED strip, we have 22 feet (10x2 + 1x2) of feeder wire, but 2 feet of that only needs to carry ~9.5A, the next 2 feet need only carry ~9A, etc. The last 2 feet of what is (10+20)x2 = 60 ft need only carry half an amp.

Using my trusty jsTIfied-emulated TI-84 Plus C Silver Edition, the zero-current point is 29.6 feet out along the strip (a 59.2-foot round-trip). It tells me that the integral of y = 10-(X>10)((X-10)*0.51) from 0 to 29.6 is 198.03926 amp-feet one-way, or 396.079 amp-feet round-trip. Going back to our amp-feet * ohms/feet calculation, but working it backwards, our 1.2V max loss = 396.1 amp-feet * max allowable ohms/feet -> max allowable ohms/feet = 1.2V / 396.1 amp-feet = 3.0mΩ/ft, between 12 and 13AWG. For sanity-checking, using another integral and the same jsTIfied instance:

• Over those first 10 feet (20 round-trip feet), we lose 0.6V at 10A (11.4V left, and half of our voltage-loss quota).
  
• In the next 2 round-trip feet, we lose another 0.058V, so 11.342V left.
  
• By 15 feet along the run (5 feet into the LED strip, 30 round-trip feet), we've lost a total of 0.86175V, or 11.138V left.
  
• By 20 feet along the run (10 feet into the LED strip, 40 round-trip feet), we've lost a total of 1.047V, or 10.953V left.
  
• Because the current is tapering off, at 25 feet (50 round-trip feet), we've still only lost 1.1558V, at 10.844V, because the current carried this far is only 2.35A.
  
• Finally, at the 29-foot mark (58 round-trip feet), we've lost 1.1877V, we have 10.8133V left, and all is well.

Let me know if you spot any logic or math errors.

This is true, I didn't take into account the voltage drop per foot, and that the effect would be amplified at lower voltages.
Interestingly, in the next table down on that page, 16AWG has a ohm/ft resistance of 4.016 mOhm, and 14AWG with 2.5 mOhm/ft.

The little 5V/2A buck converters should only be drawing 1A@12V, or 1.33A@9V, given a (conservative) 83.5% efficiency. And to keep them from burning out, I wouldn't suggest pushing them much harder than 1.5A, which translates to 3/4A at 12V, or 1A @9V. I am including the 9V figure because these buck converters will happily make 5V out of 8 or 9V. So, given a maximum current needed per run of 1.33A, and a maximum run length of 60ft, this gives 80 amp-feet, and with an allowable voltage drop of 3V, this leaves a maximum ohm-ft resistance of 37.5 mOhm, or (according to the chart, conservatively) 22AWG wire.

But, that's just a single run. I suppose you could run 22AWG wire to each buck converter, but that would get messy.

Your calculations look great, btw. Good to see that fnInt( function be put to a good use!


---------------------------------------------

Might I also suggest using 24V, so that there is half the current needed? Here is a ~$21 24V/10A powersupply. and here are some 11mm-wide 24-5V 2-3A buck converters.

One of my concerns about that much voltage drop is where the power is going, notably, heat.

After searching and computing, I'm definitely thinking about 24V now. It seems like Meanwell has decent quality, e.g. https://www.amazon.com/MEAN-WELL-LRS-350-24_2-Industrial-LRS-350-24_2PK/dp/B085G6Q2ZV - 24V/14.6A (365W) for $28.50 (or two for $55). Assuming the efficiency's pretty good with something like those, we can just halve the current for these ~30-foot each way runs from 10A to a generous 5A, and 10% loss (i.e., 2.4V) is now acceptable (incidentally, up to 4 runs * 5A * 2.4V = 48 watts of heat getting dumped out, but oh well): 18AWG wire, 6.4mΩ/ft only drops 1.27V (22.73V left) after 29 feet each way, so we're golden.

Working it the other way, half the current of course means half the amp-feet, or 198.04 round-trip, 2.4V max allowable loss gives us a whopping 12mΩ/ft, or 20AWG, but I'm perfectly happy using 18AWG and keeping the highest drop to 1.27V. This seems like a no-brainer, and I can just use more power from buck converters for my controller.

Efficiency on the 24V supply doesn't really matter, right? The important number for wire sizing is the local power supplies' efficiency.

But 88% typical efficiency (as specified but not characterized over load) for the 24V supplies is pretty good. Those local supplies NoahK points to are very attractively priced but perhaps questionable for efficiency: the only number they quote is 97% when basically doing nothing (6.something to 5V conversion). I'd expect it to get worse when doing larger steps down.

Oh, I'm taking about efficiency of the buck converters - I was saying 90% efficiency converting 24V down to 5V. As you say, I don't care about the 24V supply's efficiency, as long as it can deliver most of the current it claims it can.

Yeah, I'm not going to advocate for the efficiency of those little buck converters. The main aspect I was looking for was size. The aluminum channel that we linked earlier supported an LED strip with a maximum width of 10.5mm, so, assuming that is also the width of the channel, determines the maximum width of the buck converters as well. And the 24V->5V converters I linked were the smallest I could find on Amazon at 11mm. There are other channels, while not being 45 degrees, that can hold wider things

18AWG wire should be good if each section of the strip, aka each buck converter, is getting it's own home-run to the powersupply. A thicker channel would also be more accommodating to cramming wires in it alongside the LED strip, unless you were thinking of running the wires in a separate cable duct or raceway.

Not necessarily regarding power supply needs, but just a general warning about bright LEDs taking lots of current is that they get hot. Especially if they've got that IP65 coating on top of the strip. I see you're going for the non-waterproof ones (no doubt to lessen this issue) but I used to have LEDs like this mounted upside-down as a desk lamp. They'd get hot and if they're hot for long enough they start to fail one RGB element at a time. Be sure that with all that current you're pumping through it that you're thinking about cooling. They might be able to sink some heat into the surface you stick them to but I'm not sure I'd be ready to rely on that.

Yeah, I was going to put the buck converters in the channel, and figure out a way to face them against the aluminum to dissipate heat (might need some thermal paste and shims, also some mounting solution like a 3D-printed bracket?).

My computations actually suggest that I can do a single 18AWG run per set of LEDs (that is, ~320 of them) and still have very acceptable losses. I certainly was planning to run the wires alongside the LED strip, possibly using some sort of custom bracket to retain the wires and LED strip inside the channel.

I did indeed intentionally choose the non-IP65 ones to reduce heat retention, and I'm hoping to get at least some portion of the heat to be dissipated by the aluminum channels. I should reiterate that I don't plan on running these at full brightness all the time, but I want to spec everything so they can do so sometimes.