Canon vs CHDK histograms - page 9 - General Help and Assistance on using CHDK stable releases - CHDK Forum  

Canon vs CHDK histograms

  • 88 Replies
  • 4657 Views
*

Offline reyalp

  • ******
  • 13394
Re: Canon vs CHDK histograms
« Reply #80 on: 30 / June / 2021, 01:34:51 »
Advertisements
Haven't had a chance to spend a lot of time checking but have looked at the liveview histo data and it looks ok, ie no strange peaks.

Also, using the Y option, the CHDK histogram looks well matched to the Canon one in both EVF and LCD modes.

Looking good :-)
Thanks for testing. I've checked in the fix in trunk 5977, and also in the ximr branch.  I'll backport it the stable branch later.

This took a while, but it (hopefully) solved some issues that affected many cameras, as well as a couple specific to the M3 ximr builds.

Quote
Ps forgot to say that the under and over warnings looks off and might need ‘calibrating’.
If you can describe or post screenshots/screendumps of what's wrong, that would be helpful.
Don't forget what the H stands for.

Re: Canon vs CHDK histograms
« Reply #81 on: 30 / June / 2021, 12:33:38 »
Quote
If you can describe or post screenshots/screendumps of what's wrong, that would be helpful.

Had another look and things look reasonable. So nothing to add at the moment, but I’ll keep an eye on things  ;)

Re: Canon vs CHDK histograms
« Reply #82 on: 30 / June / 2021, 14:49:17 »
@reyalp

I noticed some strangeness the other day, but in my mind I associated it with the 'wrong build' you sent me. I've now looked at things and there is definitely a strangeness going on.

Specifically, you see a good match to the Canon histo, but if you then under or over expose you see more and more 'quantisation' going on, the attached hopefully illustrates.


*

Offline reyalp

  • ******
  • 13394
Re: Canon vs CHDK histograms
« Reply #83 on: 30 / June / 2021, 16:21:26 »
Specifically, you see a good match to the Canon histo, but if you then under or over expose you see more and more 'quantisation' going on, the attached hopefully illustrates.
That doesn't look right, but it's really hard to tell what's going on from one small fuzzy image. A screen dump of that case might be informative.

One thing I noticed is if you set "ignore boundary peaks" to a large value, and most of the histogram falls outside the boundary, what's displayed gets very noisy because it magnifies whatever else is left. However, your canon histogram doesn't look like much should be outside the boundary.
Don't forget what the H stands for.


Re: Canon vs CHDK histograms
« Reply #84 on: 30 / June / 2021, 17:44:33 »
Quote
but it's really hard to tell what's going on from one small fuzzy image

Sorry

UPDATE:

I had a look at the liveview data and it looks OK, ie when in the histogram you see the quantisation.
« Last Edit: 01 / July / 2021, 02:43:24 by pigeonhill »

*

Offline reyalp

  • ******
  • 13394
Re: Canon vs CHDK histograms
« Reply #85 on: 04 / July / 2021, 03:47:56 »

UPDATE:
I had a look at the liveview data and it looks OK, ie when in the histogram you see the quantisation.
I'm still not sure what is going on, but I notice in the screenshot, auto-magnify is 17x. That could explain the 'quantization', and be consistent with the majority of the histogram being beyond the boundary, but doesn't explain the difference from the Canon histogram.

What is "ignore boundary peaks" set to?

It's unlikely I can tell much more without a screen dump.
Don't forget what the H stands for.

Re: Canon vs CHDK histograms
« Reply #86 on: 04 / July / 2021, 03:55:23 »
Quote
It's unlikely I can tell much more without a screen dump.

I'll do that now. Can you remind me what you wish to see and how I get it  ;)

UPDATE

Ignore that, I remembered.

Plus, after loading @philmoz latest build, the histogram looks reasonable, even if I under or over expose, within limits.
« Last Edit: 04 / July / 2021, 07:16:44 by pigeonhill »

Re: Canon vs CHDK histograms
« Reply #87 on: 04 / July / 2021, 03:58:58 »
@reyalp

Quote
I'm still not sure what is going on

One thing I have noticed, by using log display, is the constant 'noise' in the floor of the histogram.

This is very apparent when there is no light there, ie well under exposed.

I guess this is 'natural' photon noise?

I wonder if it is worth suppressing the 'dark noise'?

Just a thought.

Update: I personally find that, on the M3, which has a Canon histogram, having the CHDK histogram set to log gives me more info. As I say, doing this you see the 'noise floor'.
« Last Edit: 04 / July / 2021, 04:12:51 by pigeonhill »


*

Offline reyalp

  • ******
  • 13394
Re: Canon vs CHDK histograms
« Reply #88 on: 04 / July / 2021, 18:00:19 »
One thing I have noticed, by using log display, is the constant 'noise' in the floor of the histogram.

This is very apparent when there is no light there, ie well under exposed.

I guess this is 'natural' photon noise?
It's normal for the live view to show noise, and since the CHDK histogram is based on the live view, it will reflect that noise if it's operating correctly. Log mode and auto-magnify will naturally tend to make variations in smaller counts more obvious.

I don't know whether this explains what you are seeing or not. In general, it's very difficult for me to draw useful conclusions without seeing the actual data, either by reproducing locally or from a screen dump.

Quote
I wonder if it is worth suppressing the 'dark noise'?
My priority is making sure it's not broken, that is, that the counts in the histogram accurately reflect the values in the framebuffer.
Don't forget what the H stands for.

 

Related Topics