From cabinet grid to Resolve timeline
A “P3.9” cabinet is 500 mm across 128 pixels, which is 3.90625 mm. Type the 128. Everything on this page is one division on the numbers you type — the canvas, what your master does when it lands on it, and where the vertical cut runs out of pixels.
The canvas
| Canvas | 3,072 × 1,536 4,718,592 px · 4.72 MP |
|---|---|
| Physical size | 12.000 m × 6.000 m aspect 2:1 (2.000000) |
| Against UHD | 56.89 % 2.2756× a 1080p frame |
| Pitch, derived | 3.906250 × 3.906250 mm 256.0000 px per metre |
| Cross-check against the spec sheet | -11.52 mm over 12.000 m a 3.91 mm spec figure implies 127.8772 px per cabinet, -0.096 % |
| 4:2:0 legality | Both dimensions even — legal and both land on H.264’s 16-pixel macroblock grid |
What your master does when it lands on it
Two scale factors, not one: 0.8000 across and 0.7111 down. Every figure below is per side, centred — the alignment is a setting, not a fact.
| Scale entire image to fit | lands 2730.67 × 1536.00 170.67 px black each side, 0 px top and bottom (not a whole pixel — the resampler blends the edge across two samples) |
|---|---|
| Scale full frame with crop | lands 3072.00 × 1728.00 trims 0.00 px each side, 96.00 px top and bottom — 11.111 % of the frame gone |
| Stretch frame to all corners | 12.5 % too wide nothing in the chain reports this one |
| Centre crop, no resizing | 384.00 px each side, 312.00 px top and bottom keeps 56.89 % of the master’s pixels |
The vertical cut
| Crop at full wall height | 864.00 × 1,536 28.125 % of the wall width · nearest even width 864 px |
|---|---|
| Scale to target | 1.2500× 1,327,104 source samples land behind 2,073,600 delivered pixels (0.6400 per pixel) |
| Where the band sits | x = 1104.0 to 1968.0 35.9375 % to 64.0625 % of frame width · crop centre can travel 432.0 to 2640.0 |
| Against the thirds | outside by 80 px a subject composed on the thirds line at 1024.0 lands on the crop edge. Slide the crop and the whole background slides with it. |
| Height ceiling | the wall is 1,536 px high, under the 1,920 px target that ceiling only binds if you cut the vertical from the file you sent to the wall — a master above wall resolution can be cut at full quality |
The second surface
| Ribbon | 1,792 × 128 7.000 m unwrapped |
|---|---|
| Folds at | x = 512 and x = 1,536 front spans 1,024 px; its centre is 1,024, the canvas centre is 896 |
| Booth-centre offset | +128 px 1.000 cabinets · 500.00 mm. Centre the artwork on the canvas and it is off the front by this much. |
| Stacked master canvas | 3,072 × 1,664 aspect 24:13 |
| Raster the processor is handed | 5,111,808 px 61.63 % of a 3,840 × 2,160 raster |
| Pixels with an emitter behind them | 4,947,968 px the dead band is 1,280 px wide — 10.00 cabinets, 5.000 m |
| Fits the processor | yes a canvas can sit under a processor’s pixel budget and still be refused for being too wide. The width is the constraint. |
The Resolve timeline
Field names are Resolve’s own. This page prints no menu path for them: nobody has opened a named Resolve build to verify the paths for this page, and a path from memory is not worth more than no path.
DaVinci Resolve — LED canvas 3072 x 1536 Set the timeline before you import anything: the timeline frame rate greys out once a clip is in the Media Pool. Timeline resolution ......... Custom, 3072 x 1536 Pixel aspect ratio .......... Square Timeline frame rate ......... <the show's rate> set before import Playback frame rate ......... same as the timeline Mismatched resolution files . Scale full frame with crop Output / render resolution .. follows the timeline From a 3840 x 2160 master onto this canvas: scale is 0.8000 across and 0.7111 down — two numbers, not one fit ....... 170.67 px of black down each side crop ...... 96.00 px off the top and the same off the bottom, 11.111% of the frame stretch ... 12.5% too wide. Not a choice. 1:1 ....... 384.00 px each side, 312.00 px top and bottom, keeps 56.89% Canvas checks: 3072 / 2 = 1536 and 1536 / 2 = 768 — even, so 4:2:0 is legal 3072 / 16 = 192 and 1536 / 16 = 96 — on the H.264 macroblock grid cabinet pitch 500 / 128 = 3.906250 mm, 256.0000 px per metre 3072 px x 3.906250 mm = 12,000 mm = 24 x 500 mm Vertical cut (1080 x 1920): 864.00 x 1536 at full wall height — 28.125% of the wall width 1.2500x scale; 1,327,104 source samples behind 2,073,600 pixels Field names above are Resolve's own. This page does not print a menu path for them — see https://festivalluts.com/tools/stage-calculator for why.
Type the pixel count, not the pitch
A cabinet is sold by its pitch. A 500 mm cabinet carrying 128 pixels across is marketed as P3.9, and 500 / 128 is 3.90625 mm — so the spec sheet rounds, and the number you are given is already wrong in the third decimal. Build from it and the error multiplies by the number of cabinets.
On this wall that costs 11.52 mm across 12.000 m. It is not much, and it is also not nothing: it is the difference between artwork that lands on a cabinet seam and artwork that lands a centimetre inside one. The fix is free — get the pixels per cabinet as an integer, multiply up, and derive the pitch last. The pixel count is the number that is exactly right, because a cabinet cannot carry a fractional pixel.
The odd-looking figures on spec sheets follow from this. For the 500 mm and 960 mm cabinet families, the pitch is a cabinet dimension divided by an integer: 500 / 192 is 2.604167, 500 / 128 is 3.906250, 500 / 104 is 4.807692, 960 / 144 is 6.666667. Marketed, those become P2.6, P3.9, P4.8 and P6.67. Other cabinet families exist — this is a pattern, not a law of spec sheets.
The canvas, in one line of arithmetic
Three multiplications and nothing else. Pixels across one cabinet times cabinets across is the canvas width. Cabinet width in millimetres times cabinets across is the physical width. The pitch falls out of the first two and is never an input.
When the pitch is the same horizontally and vertically, the canvas aspect is the physical aspect exactly — 2:1 here, on both. Mesh and transparent curtain screens are often specified with different horizontal and vertical pitch, and on those it does not hold; the calculator computes both and says so when they disagree.
The cross-check line above is worth reading carefully, because it is easy to mistake for a proof. Multiplying the canvas width back by the pitch this page derived returns the physical width identically, for any pitch — it is an identity and proves nothing. Run it against the pitch printed on the spec sheet instead and it becomes a rounding detector: the number it returns is how far the marketed figure has drifted from the cabinet.
What your master does when it lands on that canvas
An LED canvas is almost never a standard resolution, so a master built at UHD arrives at a size the wall does not share. The processor resolves that with one of four behaviours, and the one thing they have in common is that the scale is two numbers, not one.
Fit takes the smaller factor and leaves black on the long axis. Crop takes the larger and throws away the overhang. Stretch takes both independently, which is the only one of the four that changes the shape of a face, and the only one nothing downstream will report. A 1:1 centre crop resizes nothing and simply discards what does not fit.
The figures above are per side and assume the picture is centred, which is a setting rather than a fact — an operator who has anchored top-left gets all of the black on one side. Where a bar or a trim lands between pixels — a half pixel, or any other fraction — the resampler has to invent that edge, and on a hard graphic edge that is visible from the pit.
Even numbers are structural. Multiples of sixteen are a codec's preference.
These are not the same rule and the page keeps them apart. 4:2:0 chroma subsampling stores one colour sample for every 2 × 2 block of luma, so an odd width or height has nowhere to put the last row. That is structural, it holds for every codec that subsamples, and a canvas with an odd dimension is a real problem.
Sixteen is different. H.264 codes in 16 × 16 macroblocks and pads up to the next multiple, so a canvas off that grid carries a strip of padding the encoder spends bits on. HEVC codes in coding tree units of 16, 32 or 64, and AV1 in superblocks — the multiple-of-16 rule is an H.264 rule and does not carry across. LED processors and media servers commonly ingest HEVC, so check which one yours wants before treating it as a constraint.
The vertical cut, and where it runs out of pixels
A vertical reel from a horizontal capture is frame geometry, not wall geometry — the pitch never enters it. A 9:16 crop that uses the full height of the frame keeps 28.125 % of the width, and that fraction depends only on the two aspect ratios.
What changes with the wall is the ceiling. To deliver 1,920 px tall with no upscale the source has to be at least that many pixels high; this canvas is 1,536 px. It does not. That ceiling only binds if you cut the vertical from the file you sent to the wall — a master above wall resolution can be cut at full quality.
The composition consequence is the one people miss. The surviving band is centred, and the thirds lines sit 80 px outside it. A subject composed on the thirds line lands on the crop edge. You can slide the crop, but the whole background composition slides with it.
More than one surface: fold columns and the dead corner
A ribbon that wraps around the returns is one continuous canvas to the processor and three faces to the audience. The fold columns are where the artwork turns a corner, and if the returns are asymmetric — as they usually are, because the booth is not in the middle of the room — the centre of the front face is not the centre of the canvas. Centre a logo on the canvas and it is off the front by the offset above.
Stacking the wall and the ribbon into one master canvas gives the processor a rectangle, and a rectangle over two surfaces of different widths has a region with no wall behind it. That region is real pixels in the file and no pixels on the stage. The raster the processor is handed and the pixels with an emitter behind them are two different numbers, and a page that compared both against one budget would be wrong about half of them.
The Resolve timeline, and the field that locks
Set the timeline resolution before anything is in the Media Pool. The timeline frame rate greys out once a clip is imported, and the usual recovery is a new project. Pixel aspect ratio stays square — a square-pitch LED grid has square pixels, and the other options apply a stretch you do not want here.
The copy block above is generated from the numbers you typed. Field names in it are Resolve's own; there is deliberately no menu path, because nobody has opened a named Resolve build to check the paths for this page. Where this site cannot verify something it says so rather than filling the gap from memory.
What this page does not answer
It will not tell you how far back the audience has to stand before the pixels stop resolving. Every rental calculator already answers that, it is a procurement question, and the rental company is accountable for the answer in a way this page is not.
It will not give you weight, power draw, hanging load, cabinet count from a width in metres, or a price. It will not recommend a pitch for your venue, because the person specifying the wall has constraints this page cannot see. And it names no vendor products: cabinet dimensions and pixel counts are inputs here, because product lines get revised and a wrong spec attributed to a real manufacturer is worse than no spec at all.
How these numbers were produced
Every figure on this page is arithmetic on the numbers in the form, run by
public/js/stage-calc.js — one pure function with no rounding in it. The server ran
that file to fill this page before any script loaded, and the fields re-run the same function as
you change them. There is no lookup table and no constant to trust: each line is one division or
one multiplication you can redo on paper.
That matters more here than elsewhere on this site. A LUT page measures a file that is checked into this repository. This page multiplies numbers a stranger typed, so the only honest guarantee it can offer is that the arithmetic is visible and the two copies of it are the same copy.
Sources
The arithmetic is this page's own, and it is the whole of the claim. What is not from this repository is the panel geometry: cabinet sizes and pixel counts are industry conventions, and the presets are examples rather than measurements taken here — check them against your own cabinet's spec sheet.
Resolve's field names are given without a menu path for the reason above. The colour half of stage work — why a saturated LED wall clips out of gamut rather than out of exposure — is a different page: stage lasers, LED walls and what actually clips. Where a LUT goes once the timeline exists is installing .cube LUTs, and the conversion that sits under it is log to Rec.709, done right. Timeline resolution is orthogonal to colour management; nothing on this page touches it.