All posts
By Pravit Gandhi··13 min read

How to expose S-Log3 correctly

How do you expose S-Log3 correctly? Put 18% gray at 41% and 90% white at 61%, shoot at base ISO, and read the log signal instead of your monitoring LUT.

Expose S-Log3 by placing an 18% gray card at 41% on the waveform and a 90% white card at 61%, shooting at one of your camera's base ISO values, and judging those numbers on the log signal rather than through whatever monitoring LUT you are viewing. Sony's technical summary for S-Gamut3.Cine/S-Log3 publishes those anchors as 10-bit code values 420 and 598. Everything else in this guide, the zebra levels, the exposure index question, how far to the right you can safely push, is downstream of getting those two references where they belong.

The two numbers everything hangs off

Sony's technical summary gives three reference points for S-Log3. A 0% black patch records at 3.5%, code value 95. An 18% gray card records at 41%, code value 420. A 90% white card records at 61%, code value 598.

If you shoot something other than Sony, our log reference tool returns the same three anchors for any of nine formats, on both IRE scales. Middle gray at 41% is the single most useful fact about exposing this format. It is what a light meter reading translates into on a waveform, and it holds across every Sony body that shoots S-Log3, because the encoding does not change with exposure index. Sony's document says so directly: the curve covers the full EI range and there is no need to change the conversion formula for each EI setting.

Is your waveform reading 0 to 100 or 0 to 1023?

This trips up more people than it should, and it explains most of the contradictory numbers you will find in forum threads.

Sony records S-Log3 as full range. Its technical summary states the data range runs from 0 to 1023 in 10-bit, which corresponds to minus 7 IRE through 109 IRE, and that legal range by contrast is scaled and limited to code values 64 through 940, or 0 to 100 IRE. Camera SDI output is full range too, with no option to switch.

So the published percentages are legal-range readings of full-range code values. Code value 95 sits at 3.5 on that scale, 420 at roughly 41, and 598 at 61. If your scope shows raw 10-bit values instead, look for 420 and 598 directly. Middle gray lands close either way, since 420 is 41.1% of 1023, but white does not: code value 598 is 58.5% of the full 10-bit range and 61 on a legal IRE scale. That gap is enough to make you think you are underexposing when you are not, so check what your scope is set to before trusting any number, including the ones here.

What one stop looks like on the S-Log3 curve

Running Sony's published S-Log3 encoding formula against 18% gray gives the ladder below. These are computed from the equation in Sony's technical summary, not measured off a camera, so treat them as where the encoding puts a given stop rather than as a guarantee about your sensor. Six other formats have had the same ladder computed from their own published equations, which is the place to look before carrying an exposure target across from another camera.

Stops from middle gray10-bit code valueApprox. IRE
minus 321918
minus 227925
minus 134732
middle gray42041
plus 149649
plus 257358
plus 365167
plus 472976
plus 580885
plus 688694

Two things fall out of that table. The spacing above middle gray is only about eight or nine IRE per stop, which is why highlights are hard to judge by eye on a log image and why a waveform beats intuition. And from middle gray you have roughly six stops before you reach the top of a legal scale, which sets the practical ceiling on how far right you can push before speculars leave the container.

Most people expose a face rather than a gray card. A face one stop over middle gray records around 49, two stops over around 58. Those are useful targets because they come off the same curve, not off a preset someone posted.

Zebra settings that actually work

Sony's zebra system on the FX3 offers preset levels of 70, 75, 80, 85, 90, 95, 100 and 100+, plus two custom slots. The help guide describes the difference: Custom 1 checks correct exposure, where you set a standard value and a range and the pattern appears on anything inside that window, while Custom 2 checks overexposure, where you set a minimum and the pattern appears on anything at or above it.

That maps cleanly onto the numbers above. Put a narrow Custom 1 window on your chosen reference, 41 if you are working off a gray card, higher if you are exposing skin, and use Custom 2 as a top end warning somewhere in the low 90s, which corresponds to roughly six stops over middle gray.

The critical thing to internalize: zebras read the recorded log signal, not the monitoring LUT you are looking through. A zebra at 94 is not telling you the image is clipping in Rec.709 terms, it is telling you that part of the frame is about six stops above middle gray in S-Log3. If you set zebras by eye against a contrasty LUT, you will underexpose consistently, because the LUT already looks right to you before the sensor has enough light.

Base ISO, and why it is not optional

Sony defines base ISO as "the ISO sensitivity providing maximum latitude from the image sensor," and states that "shooting at the base ISO is required in log shooting in order to obtain maximum latitude." That is a statement about where the sensor readout is designed to give you the full range, not marketing.

Sony's help guides list ISO 800 and ISO 12800 as the two base values for the FX3 and FX3A, and ISO 800 and ISO 2500 for the FX30. Use the low base whenever you have light. Switch to the high base when you are genuinely short of it, not as a convenience to avoid opening the iris.

If your camera is in Flexible ISO rather than Cine EI you can move ISO freely, and you pay for it in latitude: Sony's own comparison table rates Flexible ISO as middle freedom for color grading in post and Cine EI as high.

What exposure index actually changes

Sony puts it plainly: "ISO sensitivity for recording corresponds to the base ISO, regardless of any changes to the EI." Changing EI does not change the sensor's sensitivity. It changes the brightness of what you monitor, which changes how you light and set your iris, which changes how much light reaches the sensor. That last part is the whole mechanism.

Rate a camera with base ISO 800 at EI 400 and you will open up a stop to make the monitor look correct. The file lands a stop brighter, further from the noise floor, with more shadow latitude and less highlight room. Rate it at EI 1600 and the opposite happens.

Sony allows EI within plus or minus 2 EV of base and recommends staying at or below base: "in general, shooting at an EI equal to or lower than the base ISO value is recommended." It also warns that emphasizing highlight gradation by rating well above base "will increase noise and graininess, and color grading will become more difficult." For the cleanest possible file, Sony's exposure index page says to shoot at the same EI as the base ISO.

Expose to the right, and where it stops working

Exposing to the right means deliberately overexposing relative to the meter so the signal sits further above the noise floor, then bringing it back down in the grade. With log it works, up to a point, and that point is defined by the top of the curve rather than by taste.

Go back to the table. If you push everything two stops right, the highlight that used to record at plus 4 (code value 729) now records where plus 6 was (886), and anything that was already at plus 6 has left the container entirely. Skin that would have sat at 49 now sits at 58. You have bought shadow cleanliness and spent highlight headroom, and the trade is not reversible: noise can be reduced, clipped highlights cannot be reconstructed.

My working default, offered as opinion rather than spec: about one stop of deliberate overexposure in controlled interior situations where you know what the brightest thing in frame is, and none at all when you have a window, a practical in shot, or a sky. Sony's guidance points the same direction, since rating at or below base ISO is itself a form of exposing to the right.

One more limit worth knowing. If you overexpose and then apply a conversion LUT built for correct exposure, the LUT maps your brighter values through the top of its own curve, and skin tones skew as they climb. The fix is to correct exposure while the clip is still in log, upstream of the LUT, or to use a color space transform and do the correction in the working space after it. Either way it happens before the creative look, which is the ordering argument in S-Log3 to Rec.709: LUT vs color space transform.

False color, and calibrating it to log

False color assigns a color band to each brightness range so you can read exposure at a glance. It is faster than zebras once calibrated and misleading before then, because palettes differ between manufacturers and most are laid out around Rec.709 levels rather than log levels.

Calibrate it once, on your own kit. Frame a gray card, confirm on the waveform that it is at 41, and note which band appears. Do the same with a white card at 61. Write both down. From then on the display is a real exposure tool for S-Log3 rather than a decorative overlay. If your body does not offer false color, an external monitor will, and the procedure is identical.

Monitoring with a LUT while exposing for log

Use a LUT for your eyes and the numbers for your exposure. Sony's log shooting guide lists three monitoring options on its Cinema Line bodies: S-Log3, s709 with film-like tones and low contrast, and 709(800%) with high contrast suited to broadcast. It also notes that a LUT applied while shooting can be embedded in the recorded file's metadata for post to pick up later.

Two practical warnings. Sony states that Gamma Display Assist is not applied to movies shown on a TV or monitor connected to the camera, so an external recorder or field monitor needs its own LUT loaded. And whichever LUT you view through, the waveform and zebras still measure the log signal underneath, which is the point of the whole system. Why that flat image is not a fault is covered in why your S-Log3 footage looks washed out.

Getting exposure consistent on the day is also the cheapest possible shot matching. Every stop of variation between angles is a stop somebody has to reconcile later, which is the recurring pain in the indie film color grading workflow and doubly so on a run-and-gun day like a wedding, where you cannot relight and cannot reshoot. When the variance is already baked in, automated shot matching is the cleanup pass, and it works far better on clips that were within a stop of each other to begin with.

Frequently asked questions

Where should middle gray sit in S-Log3?

At 41%, which is 10-bit code value 420 in Sony's technical summary. A 90% white card sits at 61%, code value 598. Those values hold regardless of which exposure index you rate the camera at, because the encoding curve does not change with EI.

What zebra level should I use for S-Log3?

Set one custom zebra as a narrow window on your exposure reference, 41 for a gray card, and a second as a minimum in the low 90s to warn you when something is roughly six stops over middle gray. The presets on Sony bodies start at 70, which is already well above middle gray, so the custom slots are where the useful work happens.

Should I overexpose S-Log3?

Slightly, and only when you know what the brightest object in the frame is. A stop of deliberate overexposure buys real shadow cleanliness. Two stops moves a plus 4 highlight up to where plus 6 used to sit and starts pushing speculars out of the container, and clipped highlights do not come back.

Does changing exposure index change what the camera records?

No. Sony states that the recording ISO stays at base ISO regardless of EI. What changes is monitor brightness, which changes how you set exposure, which changes how much light hits the sensor. That indirect path is exactly why EI is useful and why it confuses people.

How many stops above middle gray does S-Log3 go before it clips?

The curve reaches code value 1023 at 7.738 stops above middle gray. The figure you will usually be handed is six, and six is not invented — it is the recording ceiling rather than the curve. S-Log3 tops out near 94 on a legal IRE scale, which is code value 887, and 887 is 6.015 stops above middle gray. The ladder above lands on the same place: plus 6 stops is code 886, printed at 94 IRE.

Both numbers are correct and they answer different questions. That is worth stating once for all three of the questions in this group, because it is the same mistake each time — a recording or range convention read as a property of the curve. The curve is not what is stopping you. The camera is.

So pick by what you are doing. If you are comparing encodings — S-Log3's 7.738 against Panasonic V-Log's 8.000, ARRI LogC4's 11.350 or Fujifilm F-Log's 5.338 — use the curve figure, because those are all computed the same way from the manufacturers' published formulas and nothing else makes them comparable. If you are deciding how far right to push on the day, use 6.015, because that is where the file stops carrying anything.

One limit applies to both. Where a particular body clips is a sensor-dependent number Sony does not publish per model, so 6.015 is the ceiling of a recording convention and 7.738 is a property of an equation. Neither is a measurement of your camera.

How many code values does one stop take at middle gray in S-Log3?

75.69, measured from middle gray up to one stop over: Sony's formula puts gray at 420.00 and plus 1 at 495.69. The stop below is slightly tighter at 72.89, because the curve is still steepening as it climbs, and a one-stop window centered on gray gives 74.51. Use 75.69 unless you have a reason not to.

The answer in circulation is about 20, and where 20 comes from is worth seeing, because its source was right. 75.69 divided by four is 18.92. Twenty is the 8-bit figure. It comes out of correct writing about why 8-bit log recording suffers, which was doing its arithmetic in a 256-value container where about 19 code values per stop is precisely the problem being described, and it was then carried across to 10-bit without multiplying up for the four times larger container.

The corroboration comes from a different manufacturer. ARRI states that Log C spends 73 to 78 code values per stop in a 10-bit encoding. Our LogC3 figure at EI 800, computed the same way from ARRI's own formula, is 73.36: a different company, a different curve, the same band. Nothing in that band is anywhere near 20.

What code value is a 90% white card in S-Log3?

  1. That is Sony's own published table value, and running Sony's published formula returns 597.90, so the printed table is the formula rounded rather than a separate measurement.

The answer in circulation is 627, and it is worth dissecting because it states a correct method and then does not follow it. The method: Sony publishes 61 for a 90% white card, that 61 is legal-range IRE, and legal range spans code values 64 to 940. So 64 + 0.61 × 876 = 598.4, within half a code value of Sony's printed 598. The published answer sets that out and reports 627 anyway.

A second trap sits immediately beside it, and that one is the one that actually catches people. Sony's 61 is a legal-range IRE reading of a full-range code value. Read it as a full-range percentage instead and you get 0.61 × 1023 = 624 — wrong by a different route, and plausible-looking because the same misreading of Sony's 41 for middle gray gives 419.4, which is within one code value of the correct 420. The gray card lets the error through. The white card catches it. Expressed as a full-range percentage the same signal is 58.45%, not 61.

Sources