Find brightest pixel - page 2 - LUA Scripting - CHDK Forum supplierdeeply

Find brightest pixel

  • 20 Replies
  • 5022 Views
*

Offline reyalp

  • ******
  • 12159
Re: Find brightest pixel
« Reply #10 on: 28 / December / 2013, 22:42:17 »
Advertisements
ps amp glow comes from an electric field so it will be smooth in appearance, in one particular area, also I don't expect it in a modern sensor, and it might only apply to CCD (just some rules of thumb off the top of my head).
It definitely does apply to both CCD and CMOS, and appears on many CHDK cameras. I've never heard of a CHDK camera that didn't suffer from it, but I guess people wouldn't usually post about that. The appearance varies depending on the sensor and surrounding electronics. It would surprise me a lot if SX50 didn't suffer from it to some degree, even if it isn't the only problem in your image. The star pattern in the upper blob is a bit weird but the lower one looks pretty typical.

Here https://app.box.com/s/scdr4xgm1agnqlf9qrgu is an example from my d10. The elph130 has less, but it's still visible in a 2 minute exposure.

The best way to avoid it is to power down the amp while the exposure is in progress, but considering that Canon typically limits the exposure to 15 seconds and uses a dark frame over 3 seconds, they have little reason to expend engineering on it.

If you are feeling brave, at least one users has successfully have hacked in an amp-off circuit: http://chdk.setepontos.com/index.php?topic=8949.0  8)

Quote
I wouldn't need to take dark frames at the time, and can even go back and clean up some night shots I've previously made with the same camera.
You can do this by creating dark frames after the fact too.
Don't forget what the H stands for.

Re: Find brightest pixel
« Reply #11 on: 28 / December / 2013, 22:58:09 »
FWIW,  radioman193 claimed that most amp glow could be solved by shielding the sensors.  He posted pictures of the mods that he did on one of his cameras on the forum using aluminum foil to wrap the backside of the sensor.  He claimed it worked really well.  I'm not sure what he was actually shielding,  but based on this I'm pretty sure it wasn't the sensor's high gain amplifier as that would be on the same side of the foil as the sensor cells.

Unfortunately,  radioman193 was not to good at keeping backups and the linked images he posted are long gone.  And many people on other forums take his posts with a grain of salt - I'll spare you the details.
Ported :   A1200    SD940   G10    Powershot N    G16

*

Offline reyalp

  • ******
  • 12159
Re: Find brightest pixel
« Reply #12 on: 28 / December / 2013, 23:18:39 »
I'm pretty sure it wasn't the sensor's high gain amplifier as that would be on the same side of the foil as the sensor cells.
Good point. What I'm referring to as  "amp glow" could really be any radiation from camera electronics accumulating charge in the sensor. In a tightly packed P&S, that could come from a lot of stuff.
Don't forget what the H stands for.

Re: Find brightest pixel
« Reply #13 on: 28 / December / 2013, 23:22:23 »
Good point. What I'm referring to as  "amp glow" could really be any radiation from camera electronics accumulating charge in the sensor. In a tightly packed P&S, that could come from a lot of stuff.
I've googled this many times and never come up with a definitive answer about whether its light, EMI, or heat.  With a CCD sensor, I can see any of those building up charge in the cells.  With CMOS, not so much.  The link I posted says it's light (http://darkerview.com/CCDProblems/ampglow.php).   

There is probably a definitive answer to this but with all the "experts" on the interweb, who knows which one it is.
Ported :   A1200    SD940   G10    Powershot N    G16


Re: Find brightest pixel
« Reply #14 on: 29 / December / 2013, 00:44:04 »
Update,
I tested combinations of DNG and they are working now.  The problem I found before showed a slanted garbage on the DNG.  I've updated my firmware since then so perhaps it's fixed. 

Therefore, I've been able to save my dark frames as DNG.  They still show no hot pixels.  I still have to make new tests with the shutter closed to see if there's light leak.  Making the cold photos takes a very long time as I put the camera in the freezer for a while, then insert a fresh battery, and let it expose 4 shots at 16 minutes each.  The temp stabilizes by the last shot.

Edit: just got the close shutter to work, you have to enable native calls in the misc. menu

Edit: And now I have results from both a lens cap in light and shutter closed in light.  They both show an image!

1 lens cap (16min exposure)
2 shutter closed (16min exposure)

Edit: I also have results for the sensor in saturation.  The brightest pixel is 3968.  The dynamic range is 3844, or 11.9 bits, at ISO 80.

Conclusions
-There's definitely no hot pixels.  This sensor is a pleasure to work with.
-I confirmed there's light leak from outside the camera to inside, no matter if there's a lens cap or the shutter is closed
-I am still not absolutely convinced that the glow is light leak or amp glow.

My next idea is to take a picture of the camera itself from another camera to catch any internal glow :)
I also need to verify my script to see if there's a double exposure (like the shutter is open for a moment then closed in the script).  If there is, that's a good thing, because you can do some neat effects with that.

Also not all pixels saturated to the same amount.  I can see a possibility here and extracting a bit more quality in highlights by using every pixel to it's fullest, knowing exactly where each one saturates.
« Last Edit: 29 / December / 2013, 17:03:00 by jmac698 »

*

Offline reyalp

  • ******
  • 12159
Re: Find brightest pixel
« Reply #15 on: 29 / December / 2013, 19:14:15 »
I'm a bit confused
Quote
Edit: And now I have results from both a lens cap in light and shutter closed in light.  They both show an image!
"lens cap.jpg" and "shutter closed.jpg"  are very different. Are these crops from different parts of the image? Can you post original DNGs on a file share site like box or google drive?

Quote
-There's definitely no hot pixels.  This sensor is a pleasure to work with.
There are a lot of very bright points in your images, but without context I'm not sure what they are showing.
Don't forget what the H stands for.

Re: Find brightest pixel
« Reply #16 on: 31 / August / 2014, 03:13:20 »
Back again after a long time, but I may as well reply.

One image was taken with a lens cap on, underneath a bright lamp.
The other image was taken with a script that closed the shutter, under a bright lamp.
From looking at the images, I believe they both show light leak from the lamp.  To put it another way, they look different in pattern than a picture taken in true darkness.
Based on that, I'm concluding that nothing but external darkness can be relied on to eliminate light leak.
Even given that my "dark" condition was in a dark box, there's still led's and lcd screen on, and the box had a reflective bright white surface, so those could potentially show up in a picture.

To be really convinced it's amp glow, I'll have to make sure the display mode turns the screen off, and see if there's a visible difference (showing that the lcd on/off can exposure light).

Anyhow, at this point I'm willing to say it could be amp glow, it's just weird that there's those starburst patterns.

One more thing I forgot to mention is that my dynamic range of 11.9 ISO fits closely to what DxOmark reports for this model.

Re: Find brightest pixel
« Reply #17 on: 31 / August / 2014, 10:02:23 »
Anyhow, at this point I'm willing to say it could be amp glow, it's just weird that there's those starburst patterns.
Maybe you are seeing both?
Ported :   A1200    SD940   G10    Powershot N    G16


Re: Find brightest pixel
« Reply #18 on: 31 / August / 2014, 15:47:19 »
Now that lapser made a patch to turn off the LCD and do histograms, as I was originally wanting, perhaps I'll be able to do one final experiment to prove which is light leak and which is amp glow.  Besdies, I'd like to get back to my experiment to map black level to temperature for every pixel.  Luckily he has the same cameras and some interests as me :)

Re: Find brightest pixel
« Reply #19 on: 31 / August / 2014, 15:56:48 »
Now that lapser made a patch to turn off the LCD
That wasn't actually lapser's patch - and it's a released part of CHDK now : http://chdk.wikia.com/wiki/Script_commands#set_lcd_display.28.29

Quote
and do histograms, as I was originally wanting, perhaps I'll be able to do one final experiment to prove which is light leak and which is amp glow.  Besdies, I'd like to get back to my experiment to map black level to temperature for every pixel.  Luckily he has the same cameras and some interests as me :)
reyalp has added "hooks" to let you do the equivalent of what lapser was doing via CHDK mods using just Lua script functions.

http://chdk.wikia.com/wiki/Script_Shooting_Hooks

http://chdk.setepontos.com/index.php?topic=11081.msg108530#msg108530

I'm sure there is one more I'm missing that was specific to histograms but the forum seach engine really draws a huge vacuum.
Ported :   A1200    SD940   G10    Powershot N    G16

 

Related Topics