[Na Horuk]

Well however it works it seems to work well on the Olympus. The ability to evaluate the final tonality of a scene without chimping is a big step forward over my DSLR if you are a RAW ETTR shooter.

Its possible that Oly reads the histogram from the feed but then modifies it so it better represents the raw? Unfortunately we never know exactly what the programmers did. Good to hear that it works well on Oly. Throughout the years we have had a couple threads that complain about the histogram on Pentax cameras. Pentax could update this feature, it is time

[wildman]

You can't trust it either.

Perhaps not but don't make the perfect the enemy of the good.

Case in point the following pic - I could see with the naked eye the DR was beyond what the small sensor could handle well but with the Olympus it was no big deal to adjust the hgram slightly to the right in order to sacrifice some highlight and gain some shadow detail just by adjusting the hgram on the fly. Try doing that with my K5 and chimping maybe 3 or 4 test frames - a real pain in the a**. Anyway by that time the damn cat would have been long gone.

See: https://www.pentaxforums.com/forums/76-non-pentax-cameras-canon-nikon-etc/171...ml#post3743973

[clackers]

Perhaps not but don't make the perfect the enemy of the good.

Case in point the following pic - I could see with the naked eye the DR was beyond what the small sensor could handle well but with the Olympus it was no big deal to adjust the hgram slightly to the right in order to sacrifice some highlight and gain some shadow detail just by adjusting the hgram on the fly. Try doing that with my K5 and chimping maybe 3 or 4 test frames

Why would you chimp 3 or 4 frames with your K-5 if this was important to you, Wildman? I won't doubt that it is, we all have preferred workflows.


You'd just enable the histogram in Live View, right?

[wildman]

You'd just enable the histogram in Live View, right?

A reasonable question...

...being old school I really find holding a camera as big and heavy as a DSLR away from me awkward, unnatural and unsteady not to mention my old eyes see much better through a viewfinder. The Olympus has implemented this feature in a natural, fast and intuitive way and everything I need is right in front of me on the viewfinder or at my finger tips. In comparison live view on the K5 feels like an awkward clumsy kludge job.

As a general rule if I take my eye from the viewfinder to fuss with gear I have lost the shot.

But, granted, "proper" ergonomics is highly idiosyncratic. Hell I still think my Nikon S2 rangefinder from 1956 was one of my fastest most responsive cameras.

---

Last edited by wildman; 08-22-2016 at 05:36 AM.

[clackers]

As a general rule if I take my eye from the viewfinder to fuss with gear I have lost the shot.

Fair enough. A Hoodman or similar loupe may give you what you need.


BTW, the overexposure blinky doesn't rely on the JPEG rendering engine AFAIK, and is quite accurate.

[wildman]

when i get home and load that image into LR5, the histogram there most of the time is cramped all the way to the left.

Basically, the histogram in the camera is slightly inaccurate. It is not of the raw data itself, but of the developed jpeg. This means if you have things like vignetting correction, shadow correction enabled, if you have a certain type of jpeg mode (bright, landscape, film reversal, etc.) and even WB will have effects on the histogram. The raw data itself will have a different histogram (usually slightly more to the left, but can even have different peaks)
Secondly, once you import the raw file, it may immediately get edited - depending on your settings of the software. This edit can cause further changes in the histogram, making it look even more different from the one in-camera.

Assuming the above quotes are true and accurate it raises a question for me:

It seems to me that whatever else the sensor "knows" at the instant of exposure it does know which pixels are going to be over or under saturated for any given exposure and this sensor behavior is independent of whatever else may or may not be happening further down the line.

Question: Why does it take rocket science to report this relatively simple sensor behavior directly back to the photographer out in the field in real time?

As a further refinement, when shooting RAW, the Tog should have the option of, in addition to the usual jpg a small thumbnail of the histogram taken directly off of the sensor at the instant of exposure.

I really don't get it, or perhaps, I'm just being too simple-minded.

[clackers]

.

Question: Why does it take rocket science to report this sensor behavior directly to the photographer out in the field in real time?

I really don't get it.

???

I have the warning turned on, and the areas flash obviously.

[wildman]

???

I have the warning turned on, and the areas flash obviously.

The question still remains - does this accurately reflect true RAW sensor behavior? I really don't know.

[clackers]

The question still remains - does this accurately reflect true RAW sensor behavior? I really don't know.

As I understand it, yes, those pixels are maxed out.

The JPEG ones may only have clipped because the algorithm lifted the RAW values.

[wildman]

As I understand it, yes, those pixels are maxed out.

Good then we are half way there.

If, in fact, the blinkies are based on accurate RAW data I don't think it's asking too much, in addition to blinkies and/or in combination with blinkies, to also or only display this data in the form of a histogram.

I say this because, when doing creative work, blinkies can be distracting and misleading and a histogram more direct and intuitive.

[Na Horuk]

No, I think the blinkies are wrong, as well. They are generally right, but sometimes they flash things that can be recovered if you shot raw. At least in my limited experience. I think there were many threads about the histogram inaccuracy some years ago, especially when the uni wb thing and ettr were really popular. You can search back, but I doubt you will find anything that hasn't been mentioned here. Use the Neutral jpeg setting, disable jpeg corrections, use a neutral custom WB (uniwb) and the jpeg will be very similar to the raw.

The problem is that the sensor doesn't know anything, it only senses things. And then there is a path that goes to different parts of interpretation chips, which encode the image or video data in a certain way, turning it into an actual visual file. Only at this point does another part of the camera read this encoded data and displays it on the screen, and then draws histogram and other things.

You will notice that the jpeg thumbnail of the raw file gets "developed" according to the camera settings. Depending on these settings, it can be more or less accurate. Especially vignetting correction, shadow correction will play big role; but things like saturation, contrast and WB will, as well. It probably has to do with buffer and overall camera speed. Making a histogram of a jpeg file or an encoded video stream is much easier than of a big 16MB (or even bigger, with the latest cameras!) file that contains only raw data. Remember that raw data itself is not yet an image, it is just data. It has to be interpreted into an image, and then the interpretation gets analyzed. Interpreting a large file with lots of data is significantly more difficult and slow than a smaller file that is already neatly packaged into jpeg standard.

[BarryE]

Another complication for the camera to reflect highlight problems accurately, is that post processing of raw files can recover highlights where one channel is maxed out by using the other channel data if they are ok. This allows burnt out highlights to be recovered, somewhat.

[wildman]

The problem is that the sensor doesn't know anything, it only senses things.

Sounds very much like a certain candidate that is currently running for the office of President in my country.

And then there is a path that goes to different parts of interpretation chips, which encode the image or video data in a certain way, turning it into an actual visual file. Only at this point does another part of the camera read this encoded data and displays it on the screen, and then draws histogram and other things.

Yes I think I understand that. But I am dealing only with whether any given pixel is or is not over or under saturated and not trying to recreate a image meaningful to human perception. At least that's all I'm purposing.

It's my understanding that each pixel has it's own discrete site with it's own discrete data path. Would it not be possible, if not practical, at the physical pixel level to intercept that data, determine it's luminosity and display that as a simple, but accurate, greyscale (lumenosity) graph of what the sensor is actually doing at the RAW level?

But then perhaps the real problem is we are all speculating above our pay grade.
Where in the hell is the proverbial nerd when we need him?

[BrianR]

No, I think the blinkies are wrong, as well.

I think you're correct and the blinkies also depend on the jpeg settings. At least on my k5iis, the highlight blinky area is noticeably different between two shots that only differ in the contrast setting. I'm not sure if all Pentax models behave the same way.

[Na Horuk]

I'm not sure if all Pentax models behave the same way.

They probably do, and I'm surprised that some other brands might not. Its nice to hear that some other brands are taking more care in this regard. While I really like Pentax and I believe Pentax pays attention to a lot of details, they are known to sometimes be slightly sloppy when it comes to "software" (complaints about focus peaking, CA correction in BW jpeg mode (would add colourful corrections after the photo was turned BW, which is just silly, it should be the other way around), histogram and blinkies being inaccurate, digital SR in video mode, etc.)