SX510 - page 19 - CHDK Releases - CHDK Forum supplierdeeply

SX510

  • 213 Replies
  • 68847 Views
*

Offline reyalp

  • ******
  • 13715
Re: SX510
« Reply #180 on: 26 / November / 2015, 15:39:06 »
Advertisements
OK, I had some free time today to do some of the test scripts included in 1.4

isobase didn't create a log but this is the result I got from screen:

isobase.lua
BASE=200

Thanks. This is set correctly in the port.
Quote
I have a question about that isobase, does that mean that's the camera's "native true ISO"? like NIKON's that have 200 as base also? Or nothing to do with that?
From a CHDK POV, it's an arbitrary constant that is require to correctly override the ISO setting. The Canon auto ISO works by adding a (positive or negative) offset to a base value. CHDK (ab)uses this for ISO overrides.

Quote
And do you know what is the true basic ISO in Powershots? 80? 100?  ???
It almost certainly varies between cameras, Canon users many different sensors, and the minimum ISO available in the UI varies between models. ISO override is a complicated subject, the Canon firmware uses two different units (which we refer to as "real" and "market") http://chdk.setepontos.com/index.php?topic=10341.0

See also http://chdk.setepontos.com/index.php?topic=12165.50
Don't forget what the H stands for.

Re: SX510
« Reply #181 on: 02 / December / 2015, 00:24:36 »
Today I noticed a strange behavior with the histogram and higher ISO's. If I set my ISO from 80 to 400 the histogram seems to show true results compared to the preview in playback mode. If I raise my ISO above 400 (800,1600 and 3200) and adjust my ET I see the histogram shows a lower value than the true value written in the file, photos can come overexposed, even the live preview shows a brighter image, but:

I then configured histogram to show all the time for further testing, so when not half pressing at the higher ISO's it seems to show the real value but when I half press it goes down lower and if I shoot the recorded preview shows me what the histogram originally showed without half pressing.

But the thing here is that I started camera without CHDK to see if something similar happens at these high ISO's, I don't have a live histogram from the Canon FW but I did notice that the live preview becomes slightly darker when I do the half press in those ISO's, doesn't do it in ISO 400 below, so it's really not a CHDK issue, it's something to do with the ISO's in the camera. Could this be the result of expanded or not true ISO's? I can live with it if I use the histogram all the time and rely on it only before half pressing the shutter, but I wanted just to let you know.

Thanks!
---------------------------------
Never tie a ship to a single anchor
nor life to a single hope

Re: SX510
« Reply #182 on: 10 / December / 2015, 18:48:57 »
Today I had a few crashes using ND filter state "IN" with 1.4, here's the ROMLOG I got. This did not happen in previous 1.3 since the ND filter was something I have used a lot.

Thanks :)

---------------------------------
Never tie a ship to a single anchor
nor life to a single hope

Re: SX510
« Reply #183 on: 10 / December / 2015, 19:00:14 »
Today I had a few crashes using ND filter state "IN" with 1.4, here's the ROMLOG I got. This did not happen in previous 1.3 since the ND filter was something I have used a lot.

Thanks :)

I'm thinking maybe it has probably something to do with "Manual bad pixel removal" I was able to add my sensor size to the program and compile it, it's working great but maybe extra memory going to use with that function that might cause this crash? I disabled it and it's enabled again no crashes so far with ND filter for now, memory seems pretty good with and without ND filter states and with bad pixel removal on, I'll keep testing and see....
---------------------------------
Never tie a ship to a single anchor
nor life to a single hope


*

Offline reyalp

  • ******
  • 13715
Re: SX510
« Reply #184 on: 10 / December / 2015, 20:35:29 »
I'm thinking maybe it has probably something to do with "Manual bad pixel removal" I was able to add my sensor size to the program and compile it, it's working great but maybe extra memory going to use with that function that might cause this crash?
How large is your badpixel file?
Quote
I disabled it and it's enabled again no crashes so far with ND filter for now, memory seems pretty good with and without ND filter states and with bad pixel removal on, I'll keep testing and see....
The ND override shouldn't take any noticeable amount of memory.

The romlog shows a crash in BrtMsrTask. CHDK does not modify this task, but I suspect this task has something to do with exposure calculation, so ND override might have some effect on it.

I would be somewhat surprised if any of this was related to 1.4.
Don't forget what the H stands for.

Re: SX510
« Reply #185 on: 10 / December / 2015, 21:30:14 »
Quote
How large is your badpixel file?
It's 130.7 Kb over 10,000 lines, is there actually a way to trim it if it's too large? I understand that only the first 80 Kb's of it are used, it takes out the most noticeable stuck bad pixels (about five or six) I see in my JPEGS, I'm not sure if all those too many lines are overkill for the memory or is this a normal size? I'll attach it if you are interested in seeing it

Quote
The romlog shows a crash in BrtMsrTask. CHDK does not modify this task, but I suspect this task has something to do with exposure calculation, so ND override might have some effect on it.

Correct. I assume this has something to do with the other issue I was having with the histogram and ISO I posted above in reply #181, something quirky with ISO's above 400 and the way the display shows me the exposure...

Thanks as usual! :)
---------------------------------
Never tie a ship to a single anchor
nor life to a single hope

*

Offline reyalp

  • ******
  • 13715
Re: SX510
« Reply #186 on: 10 / December / 2015, 22:13:15 »
Quote
How large is your badpixel file?
It's 130.7 Kb over 10,000 lines, is there actually a way to trim it if it's too large? I understand that only the first 80 Kb's of it are used, it takes out the most noticeable stuck bad pixels (about five or six) I see in my JPEGS, I'm not sure if all those too many lines are overkill for the memory or is this a normal size? I'll attach it if you are interested in seeing it
That seems like a very large number, especially if you only see handful of hot pixels in normal images. I'd guess there was something wrong with the way it was created.

If you post the modified code and image you used, I may be able to tell what went wrong.

You can also use chdkptp to make a hotpixel list from a DNG, see https://www.assembla.com/spaces/chdkptp/wiki/DNG_Processing at the end of "Generating badpixel lists"

Quote
I assume this has something to do with the other issue I was having with the histogram and ISO I posted above in reply #181, something quirky with ISO's above 400 and the way the display shows me the exposure...
I don't have much idea about this issue. It's normal for the firmware to brighten up the live view display in low light when you are not half pressing. When you half press, it should preview the actual exposure. The CHDK histogram just reads the buffer used for the LCD display.
Don't forget what the H stands for.

Re: SX510
« Reply #187 on: 10 / December / 2015, 22:55:35 »
Quote
That seems like a very large number, especially if you only see handful of hot pixels in normal images. I'd guess there was something wrong with the way it was created.

If you post the modified code and image you used, I may be able to tell what went wrong.

You can also use chdkptp to make a hotpixel list from a DNG, see https://www.assembla.com/spaces/chdkptp/wiki/DNG_Processing at the end of "Generating badpixel lists"

Well, I deleted the original RAW I used but just created a new one with new badpixel file, I think I previously used ISO 400, this time it was done with ISO 200 and both were 10sec exposures, so badpixel file is much smaller and still is functional at removing the visible ones. Here's a link to the files I used now and my edition of the source, line 32 would be the one I added, I tried before without the

Code: [Select]
black_level<<=2; threshold<<=2;
values but it didn't remove anything so I based it on the one on line 35 since it's also a 12 bit sensor, I'm not sure if that was correct to do or not.

Here's the link to dropbox:
https://www.dropbox.com/sh/blkrgjki0fuwfhl/AACajL7T8xPxkvu7xkqLkZgDa?dl=0

I really appreciate your help on that!

Quote
You can also use chdkptp to make a hotpixel list from a DNG, see https://www.assembla.com/spaces/chdkptp/wiki/DNG_Processing at the end of "Generating badpixel lists"

This sounds cool! I wasn't aware of chdkptp! I see I can even specify the crop sensor area with this,will give it a try later and see, really helpful!

Thanks so much! :)
« Last Edit: 10 / December / 2015, 22:59:26 by joegiampaoli »
---------------------------------
Never tie a ship to a single anchor
nor life to a single hope


Re: SX510
« Reply #188 on: 10 / December / 2015, 22:57:35 »
Sorry that link wasn't right!

https://www.dropbox.com/sh/blkrgjki0fuwfhl/AACajL7T8xPxkvu7xkqLkZgDa?dl=0

Updated in original post also...
« Last Edit: 10 / December / 2015, 23:02:34 by joegiampaoli »
---------------------------------
Never tie a ship to a single anchor
nor life to a single hope

*

Offline reyalp

  • ******
  • 13715
Re: SX510
« Reply #189 on: 10 / December / 2015, 23:35:53 »
values but it didn't remove anything so I based it on the one on line 35 since it's also a 12 bit sensor, I'm not sure if that was correct to do or not.

Thanks, that part looks right to me.

In your sample raw, it looks like there is a column of bright pixels on the left from x=80 to 88 roughly. This is responsible for most of the pixels in your file. This is outside of the active area, which starts at x=96 according to platform_camera.h, so you could just delete everything with a lower X value.

Quote
This sounds cool! I wasn't aware of chdkptp! I see I can even specify the crop sensor area with this,will give it a try later and see, really helpful!
Yes, it can give you just the active area, which ought to give you a reasonable size file.

Note my example in that link sets the threshold to 4095 which would only catch pixels at the brightest possible values. You might want to use something lower. You can use the dnglistpixels -fmt=count option or the dnghist command to see how many pixels are in a given range of values.
Don't forget what the H stands for.

 

Related Topics