Sony FX3 and FX5: one conversion, two bodies

This page does not tell you which body is better in the dark. That needs measured noise at matched exposures, and this site holds none — so anything written here would be a guess dressed as advice. What this site does hold, in code you can run, is the colour half: why one conversion covers both bodies, what matching them takes, and what a LUT was never going to fix.

The specification difference, as Sony publishes it

Those four rows are manufacturer-published figures, not measurements taken here — nobody at this site has put a meter on either body, and the base-ISO row in particular should be checked against your own camera's menu before you plan a shoot around it. Everything below this table is different: it is computed from the transform library this site runs on, and you can run it yourself.

One row touches the grade: the FX5's base-ISO list is published with a middle entry, 4,000, that the FX3's does not carry. What that is worth in noise at a real exposure is a measurement question, unanswered here.

One curve, one matrix, no camera model

The Sony conversion this site ships is keyed to a log format, not a body. In the transform library the site runs on — the same file the previewer loads and the validator tests — there is exactly one Sony entry: a decode curve and an S-Gamut3.Cine to Rec.709 matrix, filed under Sony S-Log3. No FX3 entry, no FX5 entry, because there would be nothing for them to hold. The decode and encode functions each take one number in and hand one number back. Neither takes a camera model or an ISO.

Run the curve and the anchor lands where the site's exposure tip says: encoding 18% reflectance returns 0.410557, code value 420.00 of 1023 — 41% IRE, middle grey on the waveform. With no ISO term in that function, the reading is the same at every base ISO either body offers.

So there is no per-body Sony LUT to hunt for, which is the premise this site's own FX3-vs-FX5 comparison already rests on. One S-Log3 → Rec.709 conversion, or one Color Space Transform with S-Log3 / S-Gamut3.Cine in and Rec.709 out, is correct for every Sony angle recording that format. The precondition is on the camera: the matrix shipped here is S-Gamut3.Cine specifically, and the FAQ lists a conversion LUT for the wrong log format among the three reasons a LUT fails.

What matching two bodies actually takes

The conversion puts both cameras in the same space. It does not make them the same picture.

1. Same log format and gamut on both bodies. A menu check before you roll, and the one that makes the rest possible.
2. White balance, manual and identical. Every row of the S-Gamut3.Cine → Rec.709 matrix sums to 1.0, which is what keeps neutral neutral through the transform. There is no white-balance term anywhere in the chain — the site's node order puts white balance in Node 1, on the native log image, ahead of the conversion. Two bodies balanced differently going in come out that way.
3. Exposure corrected per body, before the conversion. Correct, convert, then look. The Offset wheel only behaves like a real exposure adjustment on log data, and once the transform has compressed the highlights into Rec.709's smaller range, pulling exposure down afterwards drags that block down as a whole — muddy greys, not recovered detail. A hot B-cam is a correction node ahead of the transform, one per camera.
4. One look, last, identical on both. A creative LUT expects Rec.709 and is a fixed mapping; anything you correct downstream of it is fighting a transformation that already happened.

Watch for a Color Space Transform and a conversion LUT on the same clip: each step is individually correct, which is what makes a double conversion hard to catch.

What a LUT cannot fix

A LUT maps every input colour to an output colour — a saved transformation, not an effect with settings. That is also its ceiling: it knows nothing about the frame before or after. It does nothing about:

• Blown stage colour. Saturated narrow-band LED and laser sit outside the delivery space, so the failure is out-of-gamut saturation, not over-exposure — pulling exposure down does not touch it. The control is the CST's Gamut Mapping Method set to Saturation Compression, with Saturation Knee placing the roll-off. Highlight Recovery is a Camera RAW control: right for the FX5's internal X-OCN, useless on XAVC S-Log3.
• Banding on a smooth wash. The creative looks here are 17×17×17 lattices — 4,913 points — against 33×33×33, 35,937, for the technical conversions. Seventeen points can band on a smooth gradient in a 10-bit delivery, and a haze-filled stage wash is that gradient: the file's property, not the camera's.
• Anything that is not colour-to-colour. Resolution and motion are not in the mapping, so where two bodies differ there they still differ after the conversion.

Check the colour side on your own material: every LUT page carries a "Try it on your footage" panel that runs the actual .cube on a still you drop in, in the browser, nothing uploaded. Set its selector to Sony S-Log3 — a dropped image defaults to Rec.709 — then put a frame from each body through it.

What this page does not answer

Named plainly, because skipping the hard half quietly is worse than saying so:

• Which body is cleaner at a given ISO, and what the FX5's third base is worth in practice.
• Sensor readout speed, and how each body therefore renders strobes and fast pans.
• Dynamic range in stops.
• Price, weight, mounting, internal ND, handles and audio inputs.

None of that is in this site's code or on its pages, so none of it is here. Take those to a source that publishes its method; the colour half above holds either way.

Where to go next

The FX3 and FX5 hub collects both bodies, and the S-Log3 exposure guide works the 41% IRE figure through in full. If the FX5 is recording X-OCN, Node 1 becomes the Camera RAW tab instead, and grading X-OCN in Resolve covers that path. The conversion itself sits in log → Rec.709; the concert LUTs go after it.

---

Sources

Only material from this site's own codebase and pages.

• This site's FX3 vs FX5 comparison page — the four specification rows, unchanged.
• The transform library this site ships — the one Sony S-Log3 decode curve and S-Gamut3.Cine → Rec.709 matrix, keyed on log format, taking no camera or ISO argument, rows summing to 1.0. The 0.410557 / 420-of-1023 figure is that curve at 18% reflectance.
• The site's own engineering notes — the 33³ and 17³ lattice sizes, Highlight Recovery being Camera RAW only, the FX5 sitting between the FX3 and the FX6, and the try-it panel.
• The homepage FAQ, tips and concert page — what a LUT is, the node order, the Offset wheel on log data, double conversion, matching the log format not the camera name, 41% IRE, and Saturation Compression for narrow-band LED.