Histogram not tightly enclosed by raw-bars as in manual

[Carl Dieker]

yes, it is pretty nasty, this bug, it is easy to clip the highlights when you're so used to watching the right corner of the histogram. sure, i could use zebras, exposure and false color, but if not, the only indicator is the stoplights...

 

[TheBied]

It's important to realize that the histogram is working off of iso corrected data. So it will change based on gamma curve / iso.

If you pull a almost saturated image into RCX and play with regamma2/3/RLF/iso, you'll see a very similar effect.

The manual does need to be updated since it is showing R1 with REDColor set as the gamma curve.

The RAW clip meters and stop lights are metering RAW data, not iso corrected data.

 

[TheBied]

There does appear to be a bug where once we go over the maximum achievable value for a given curve and iso value, the data is clipped and not rendered into the histogram.

 

[Jaakko Rinne]

There does appear to be a bug where once we go over the maximum achievable value for a given curve and iso value, the data is clipped and not rendered into the histogram.

Is this bug the reason there's a gap on the right side of the histogram before the goalpost even when monitoring RAW? ISO doesn't have an effect on the histogram in RAW view, but does gamma curve?

 

[Patrick Grossien]

Thank you, Matthew for the update! After what you said, I still don't understand why, when the RAW data is clipping, the histogram is not. Shouldn't it rather be the other way round? Or is that exactly what you meant with your second post?

 

[TheBied]

Is this bug the reason there's a gap on the right side of the histogram before the goalpost even when monitoring RAW? ISO doesn't have an effect on the histogram in RAW view, but does gamma curve?

RAW view is essentially setting ISO to 320 in whatever curve you are viewing. If you compare RAW in RLF vs RG3 you'll see what I'm talking about. But bottom line is that the histogram represents a purely metadata view of the data.

There are two ways of representing this..

1) show the full width of the histogram which represents every possible bin in any mode (this is what we're doing now) .. but right now once you go over the max value all values go to 0 but the raw clip meters and stop lights indicate clipping.

2) clamp the top of the histogram to the maximum achievable value and show a stretched view of the histogram. (Portraying things in this way may alleviate some of the confusion, but will not be consistent with RCX)

 

[Patrick Grossien]

Ok, I'll check that comparison first thing tomorrow
One more though: Why does the histogram then cut close to the noise bar on the left, in dark situations? Shouldn't this then be the same as with the highlights?

 

[Jaakko Rinne]

1) show the full width of the histogram which represents every possible bin in any mode (this is what we're doing now) .. but right now once you go over the max value all values go to 0 but the raw clip meters and stop lights indicate clipping.

Ah, now I understand, thanks for the explanation.

 

[TheBied]

Ok, I'll check that comparison first thing tomorrow
One more though: Why does the histogram then cut close to the noise bar on the left, in dark situations? Shouldn't this then be the same as with the highlights?

It's actually very similar in the dark. If you put a lens or body cap on the camera try the exact same thing. Shuffle through RLF/RG2/RG3, you'll see that the overall dark point in the histogram moves around with just a curve change. RG3 actually pins the data closest to the left goal post in this case.

... and again, I recommend experimenting with a single bright or dark image in RCX and compare the results to the camera. They should behave very similar.

 

[Patrick Grossien]

Ok, Matthew. Thanks so much for the 101 ! I'll check in the morning!

 

[Carl Dieker]

There does appear to be a bug where once we go over the maximum achievable value for a given curve and iso value, the data is clipped and not rendered into the histogram.

Matthew, just so that we are clear, this bug is the root of this thread.

 

[TheBied]

Matthew, just so that we are clear, this bug is the root of this thread.

Well, the bug I mentioned above and the fact that the histogram isn't tightly coupled are 2 different things. In any case, I'm just trying to explain what the camera is currently reporting to you guys. :smiley:

 

[Alvise Tedesco]

Well, the bug I mentioned above and the fact that the histogram isn't tightly coupled are 2 different things. In any case, I'm just trying to explain what the camera is currently reporting to you guys. :smiley:

Matthew, I don't get it (mostly, I think, because my english is poor).
Can you elaborate on the bug you're talking about. I know something is wrong with my false color tool (It shows yellow, orange and red ONLY from 2500 iso and higher. If I lower the iso by 1 stop and open the lens by the same 1 stop I incomprehensibly loose red, then orange, then yellow. 800 and lower, straw will always be the "hottest" color regardless of how heavily I am overexposing). And I know something is wrong with my histogram, as it starts adherent to the left goal post but ends "black on black" instead of going adherent to the right. And it ends differently if meta iso are 320 or 800, and so on.
No real problem here. Raw noise and stoplights indicators used in conjunction with exposure tool let me have a good idea. And is getting obviously better day by day, as I'm experimenting the camera directly
But the bug you're referring to worried me.
Is it only the exposure tools that need a little tweaking from you're side, or I am actually missing some real bits of latitude of what the camera should be able to let me use??
Cheers
Alvise

 

[Carl Dieker]

Matthew, thanks!

 

[TheBied]

I know something is wrong with my false color tool (It shows yellow, orange and red ONLY from 2500 iso and higher. If I lower the iso by 1 stop and open the lens by the same 1 stop I incomprehensibly loose red, then orange, then yellow. 800 and lower, straw will always be the "hottest" color regardless of how heavily I am overexposing). And I know something is wrong with my histogram, as it starts adherent to the left goal post but ends "black on black" instead of going adherent to the right. And it ends differently if meta iso are 320 or 800, and so on.

The false color meter is acting on metadata corrected data, so it will change based on curve / iso / white balance. But I see what you are saying and it's something we need to look into. Certainly a better measure of over/under exposure can be done by looking at the raw clip meters or using the raw exposure tool.

Is it only the exposure tools that need a little tweaking from you're side, or I am actually missing some real bits of latitude of what the camera should be able to let me use??[/U]
Cheers
Alvise

It's most likely a data representation issue, not an actual sensor latitude issue.

What I am saying is, right now, the full width of the histogram represents the full range of the sensor. HOWEVER, depending on the curve you apply to the data, you may never actually see the full range plotted to the histogram.

BUT, If a linear curve were applied to the data, you would see the full range of the sensor in the histogram

 

[Patrick Grossien]

Hi Matthew, so I tried it out. And I guess it's just the bug which is confusing. As for the blacks it works as you mentioned. Put a lens cap on, and when the screen goes dark black (RG3) , the histogram goes all the way to the left cut of by the bar, which - with the same picture - isn't the case in RLF, where the Histogram ends a bit further in (to the right) as is with RG2. Which makes sense, cause the screen turns grayish.

But with the highlight it behaves differently. Which is probably the zeroing of the curve as you mentioned. It should probably be all the way up in total white situations, what it isn't. But still it's confusing a bit that the total sensor range can be reached with blacks but never with the lights.

 

[TheBied]

There does appear to be a bug where once we go over the maximum achievable value for a given curve and iso value, the data is clipped and not rendered into the histogram.

This bug is now fixed in both 3.0.7 and 3.2.8. Note, the gap will still remain in some situations since the histogram is metering metadata not RAW. But you will now see the maximum values piling up in their final histogram bins rather than disappearing.

 

[Matthew Scott]

I found that switching to RedLog in RCX (not RedLogFilm) seems to be the only one that gives an accurate histogram reading. If we could enable this in camera, it would be solved!

 

[Carmine Scalzi]

Btw also with the new firmware (both 3.0.7 and 3.2.8) same issues here

 

[TheBied]

Btw also with the new firmware (both 3.0.7 and 3.2.8) same issues here

Depending on your settings for gamma, color and iso you will see the histogram change ... and you may observe a gap. You will observe the same histogram behavior in RCX if you change the same settings there.