Scene luminance has nothing to do with exposure.

I very much doubt that.

I'm not familiar with 'rawproc' Glenn, so I'm guessing that you are a victim of auto-scaling. In this case there is so much dark area that the headlamp's area contribution is pushed down to nothing or almost nothing.

I remain interested in "accuracy" as it relates to histograms generally. The task itself is simple enough, so all histograms should be 100% accurate ... and, therefore, no one camera should be more accurate or less accurate than another in the histogram dept.

rawproc is my hack raw processor. I wrote it specifically to expose each and every operation required to turn a raw file into a rendition, no hidden operations like "auto-scale".

I like this image for such a demonstration. Cloudy morning, not a lot of light to fill shadows. And that danged headlight. Essentially, a low-contrast image with a high-intensity in-scene illuminant. Where, oh where, do you put the exposure?

Sorry, I was talking about the histogram Y-axis itself, not operations on the data.

Exposure 101 is very clear on that issue with regard to highlights in a scene and I'm sure you know what it says but, just in case ...

... say you have the perfect spotmeter and the engine is frozen in time ... you, not the camera, should decide - engine (say metered at 13EV) or lamp (say mewtered at 18EV)?

I would go for the engine and set my exposure to 13EV, say f/5.6 and 1/250 secs, see the table row for 13EV in en.wikipedia.org/wiki/Exposure_value

The lamp would blow for sure but it would be easily selected and fixed in post.

It's a lovely picture, BTW ...

Thanks, and you're looking at the not-best rendition. Just looked at the raw histogram, all of the image except for the headlight is at or below 10%, "underexposed" by any account. It requires a fairly aggressive tone curve to put everything in a good place:

Taken in 2015, just learning to shoot raw, a lot of learning ahead to figure out what to care about. Still futzing with it; currently trying out the Z 6 highlight-weighted matrix metering with a +1EV exposure compensation. I think I'd rather shoot manual with freaking raw blinkies... 😆

my sony a74 the train light would have flashed with an overexposed camera setting with blinkies set to 109+1 which is the equivalent accuracy of a raw file histogram.

No, it was shown and explained to you by multiple people in your other thread why your blinkies set to 109+1 is not the equivalent accuracy of a raw histogram.

NONE of you own a sony a74 camera so please stay out of the conversation with the.
camera is being discussed.

Just because you own a camera it doesn't make you an expert in how it works.

You have to ask? 😉

I nominate this entire thread for being placed on permanent exhibit in the dumpster.

I own a7rV and a6600 an they are right and you are wrong.

It was demonstrated on your own examples that it is not.

Nope, don't own Sony cameras but I've just picked through the Sony manuals for your camera and found that the zebra indicator is anchored to IRE levels, a video convention for expressing the range from black to white in a composite signal - definitely not raw.

Me too!

Opened in the GIMP, masked the lamp (not v. well), dialed it down a bit, applied contrast and brightness, three stages of USM, then "oilifield" ...

... the GIMP - don't leave home without it.

Pardon the typo.

I use GIMP mainly for clone/heal, and once in a while I'll mangle an image in to abject abstract-ness with it. Definitely OT to the thread, maybe I should start a PlayRaw thread with the raw file. Oops, wrong forum... 😆

It's astounding the sheer number of options available in the GIMP, even with no add-ons.

I use it mainly for analysis but layers and masks are extremely useful when messing around ...