Getting close to ordering stuff for the VIXToscope! Kevin has done some helpful proofs-of-concept, driving addressable LEDs via an ESP32 + WLED software from the Resolume ArtNet output, and we now need to transition to doing production builds of the actual strips and panels. One of the most vital things is to fit within the total power budget, and here’s where I’m starting to run into some hard rocks. My original power budget for the entire ‘scope was to fit within 15A@110V (preferably less). My calculations this weekend have me exceeding that.
Physical Layout
Let’s first take a look at where the LEDs go on the physical structure before I start reviewing the LED choices and power budget numbers. On the end by the doorways, there are “end cap” panels on each side, and a triangle on the top. The other side of the structure has another triangle. So, that’s two end caps, two triangles.
Each side has vertical strips going down every 4′ or so, starting with each side of the door. That’s six vertical strips on two sides, or 12 strips. Each side also has a strip running along the roofline under the raincap, and along the ground. Although its not in the rendering, I also want strips on each end of the prism, along the ground and up each edge. That’s another six strips between the two sides.
The idea on the patterns is that the strip patterns will show colored pulses of light that start at the bottom of one of the end triangles, go up the edge, hit the roofline and go to the end and then down the backside triangle, with the pulse splitting off at each vertical strip to go down that to the ground and pulsing along the ground. The triangular panels and end caps up top, next to the triangular panel, would be showing patterns similar to what’s being shown inside the kaleidoscope. The end caps below the triangle would be flashing something like “IN->” in green on one side of the structure and “<-OUT” red on the other, with some flashy sprinkles to make it look cool.
LED Selection
For LEDs, there’s a choice of driving with various voltages — 5V, 12V, 24V and even 48V. Higher voltages are somewhat more convenient because longer strips can be driven without having to inject power partway down a strip. However, 5V is more efficient than any of the other voltages, and I’m after efficiency to keep the total power budget down, so 5V it is and we’ll design in power injection. More cables, but power is king and we’ll have all the cables in nice wiring harnesses and with good connectors so assembly will still be fast.
For individually addressable LEDs, there a bunch of different choices. I found a good comparison that listed specs and some pros/cons for the most common types. After looking through the comparisons, I liked WS2812B and WS2813. I prefer the two signal lines on the WS2813, but feel that the physical construction (everything glued into aluminum channel or aluminum sheet metal panels) will protect the LED strips enough that I can get away with the single channel WS2812B, and use the “ECO” model that is very slightly less draw.
The next big decision is pixel density. The WS2812B comes in 30 LEDs/meter, 60/m and 144/m. Kevin’s experiments showed that 60/m still looked a bit pixelated even through a white plastic diffuser, and that 144 was very smooth. However, the power draw on the 144 is crazy high, more than twice as much, so we can’t even consider that as you’ll see below. The 30/m would look very pixelated in the strips (that’s about one per 1.3 inches), so thinking 60/m is the sweet spot for the strips.
The panels are a different story — there’s considerably more space between the plastic panel and the LEDs so diffusion should be higher. 30/m might be a possibility here, but that makes fine patterns impossible – just not enough pixels to spell out words or have more subtle things going on. So, let’s say 60/m for these as well. That’s a heavy driver though, because its the square of the number given that we’re covering an area — so 60/m is 4x the power draw per panel compared to 30/m.
Total LEDs, Total Power
I did some back of the napkin calculations, but soon got overwhelmed with the details required. So, off to spreadsheet land and I came up with a spreadsheet that took in measurements in inches (because that’s what I’m using in the CAD program), converts to meters, figures out number of strips required for the panels given their width, takes in the LED density, and spits out total number of LEDs.
It also uses a handful of different methodologies for calculating the total power. The comparison link above is one; another is https://wled-calculator.github.io/, and another is the common Internet guidance of 60ma per LED. That 60ma is pretty much worst case — that’s 100% White at 100% power. Various real world scenarios draw less power: If you run a single color (R, G or B) then it only uses 20ma; combos of two colors (R+B for purple) would be 40ma. Running more dimly than 100% scales pretty much linearly. Pixels not on draw next to nothing.
So, the spreadsheet has a handful of columns that calculate power using the different methodologies, and various duty cycles off the worst case to illustrate colors and dimmer than 100%. As you can see, the range is all over the map, ranging from a low of 1465 watts to a high of 4979. Given that the actual VIXToscope — the computer + 85″ TV monitor — draw ~300W, all of these are out of my power budget. I guess that’s not surprising given the total number of LEDs: over 16,000!!
Options
So what to do?
First, most obviously, is that actually measuring the real-world draw is the only way to truly know how much I’m drawing. The challenge, of course, is that to do that I have to finish the piece: buy many thousands of dollars of LEDs and run patterns through them. We could perform some smaller scale experiments, but time is getting really tight and we need to start building.
Second, obviously, I could do 30/m for the panels. Doing so drops most scenarios back into range:
I think the ace up my sleeve is that I have total control over what patterns I show. I can definitely run things dimmer, run sparser patterns, run single color patterns. I could even have a “low power” version that fits everything into 15A, and a “high power” version if I have access to 30A power (assuming I get the right PDU). In Resolume, I could literally remap the panels to be a virtual 30/m array even though I physically have 60/m — I doubt I’d do that because other solutions would likely work as well and look a lot better, but its a foolproof fallback. One obvious savings is those end caps: over half of the area could be purely sparse R, G, and B: Red = Out, Green = In, Blue = sparkles.
So, I think I’d prefer keeping the flexibility of having 60/m and simply measuring the finished beast once I have everything, and modulate how much power I am drawing by changing up the patterns or dimming the LEDs to keep within a 15A budget… and maybe having a 30A PDU as well for when I have the opportunity to really let it rip!
