Dynamic Range & CHDK RAW Bits? - page 2 - RAW Shooting and Processing - CHDK Forum supplierdeeply
« previous next »
Pages: 1 [2] 3   Go Down

Dynamic Range & CHDK RAW Bits?

  • 24 Replies
  • 13059 Views
*

Offline jmac698

  • ***
  • 138
Re: Dynamic Range & CHDK RAW Bits?
« Reply #10 on: 23 / November / 2012, 05:23:25 »
Advertisements
I can say a few things off the top of my head about dynamic range.  First of all, the sensor itself has a limit in dynamic range.  In the bright end, each pixel was a well depth, the maximum photons before it gives the highest output voltage.  In the dark end, there's the dark current which magically appears as a function of shutter time and ISO and temperature.  The lower temperature, ISO, and shutter, the less noise.
The ADC should be set to a scale to read between these two extremes in voltage.  No ADC is perfect and there's ADC's which lose 2 bits of ideal performance due to imperfections.  It's because they aren't a straight line to voltage; but bent (usually in the middle), so at first a change in value of +2 is a voltage of +.002 then in the middle a code of +2 is a voltage of +.004 for example, that means only the higher bits show the real voltage.
You're also limited by readout noise.
The total noise is dark current plus readout as additions and a multiplication factor per pixel for variations in pixel size etc. (but this is more fixed pattern).
Another important point is that more bits doesn't make a huge difference, because the sensor responds linearly to light but our eyes respond to a power of 2.2, so each bit is just a bit finer shading.
So in the end I think the sensor not the bits affects dynamic range, but the bits makes the range more usable by having less banding (especially in shadows).
Anyhow, just shoot HDR if it bothers you :)
ps the RAW may contain 10 bits from a 12 bit ADC because as I explained the lower 2 bits can be useless.  Maybe they just threw them away.  It may be possible to use them if you could do some kind of careful calibration.
All this is off the top of my head so don't quote me, but if you want to learn more, read astrophotography sites, maybe cloudyskies
Logged
*

Offline jmac698

  • ***
  • 138
Re: Dynamic Range & CHDK RAW Bits?
« Reply #11 on: 23 / November / 2012, 13:22:41 »
Here's a procedure and test results of dynamic range for various cameras
http://www.dpreview.com/news/2000/11/17/dynamicrange

What I've newly learned is the S surve at the ends.  What these measurements are testing is the sensor itself, basically "turning up the volume" until there's clipping and turning it down until there's noise.
Logged
*

Offline KenO

  • ***
  • 122
Re: Dynamic Range & CHDK RAW Bits?
« Reply #12 on: 23 / November / 2012, 14:12:15 »
jmac698,

Thanks for the comments and link!

"ps the RAW may contain 10 bits from a 12 bit ADC because as I explained the lower 2 bits can be useless.  Maybe they just threw them away.  It may be possible to use them if you could do some kind of careful calibration.
All this is off the top of my head so don't quote me, but if you want to learn more, read astrophotography sites, maybe cloudyskies"

Did a quick look at cloudyskies http://www.cloudynights.com/category.php?category_id=138&pr=2
had hoped they might have some of the newer P&S Canons but the Canon PowerShot SD790IS Digital Camera
by Brian Craig 01/06/10 was the last I found.

Ken



Logged
*

Offline jmac698

  • ***
  • 138
Re: Dynamic Range & CHDK RAW Bits?
« Reply #13 on: 26 / November / 2012, 00:03:39 »
Yes that place, you can find some very technical articles there. 

Something else I forgot to mention, is the black point.  The black of RAW is not defined as zero because your noise statistics would be lopsided.  So the level is set at say 1000 and noise averages to this 1000 which is defined as black.  Without less than 1000 values pulling down the average, a stacking would just end up bringing the black to a grey.  If that makes sense.
So this point is set to contain all possible noise variation.  It is estimated from the extra non-imaging pixels on the border of the CCD.  So your resolution is reduced, say 1000-4096 or you lose .25 bits from that.  I'm not fully satisfied this explains losing 2 bits though.

Here's the technical article I saw:
Signal to Noise: Understanding it, Measuring it, and Improving it
Part 3 - Measuring your Camera
http://www.cloudynights.com/item.php?item_id=2001
« Last Edit: 26 / November / 2012, 00:06:42 by jmac698 »
Logged
*

Offline philmoz

  • *****
  • 3450
    • Photos
Re: Dynamic Range & CHDK RAW Bits?
« Reply #14 on: 26 / November / 2012, 00:11:15 »
Quote from: jmac698 on 26 / November / 2012, 00:03:39
Yes that place, you can find some very technical articles there. 

Something else I forgot to mention, is the black point.  The black of RAW is not defined as zero because your noise statistics would be lopsided.  So the level is set at say 1000 and noise averages to this 1000 which is defined as black.  Without less than 1000 values pulling down the average, a stacking would just end up bringing the black to a grey.  If that makes sense.
So this point is set to contain all possible noise variation.  It is estimated from the extra non-imaging pixels on the border of the CCD.  So your resolution is reduced, say 1000-4096 or you lose .25 bits from that.  I'm not fully satisfied this explains losing 2 bits though.

Here's the technical article I saw:
Signal to Noise: Understanding it, Measuring it, and Improving it
Part 3 - Measuring your Camera
http://www.cloudynights.com/item.php?item_id=2001


Black point is ~31 for 10 bit sensors and ~127 for 12 bit sensors.

Phil.
Logged
CHDK ports:
  sx30is (1.00c, 1.00h, 1.00l, 1.00n & 1.00p)
  g12 (1.00c, 1.00e, 1.00f & 1.00g)
  sx130is (1.01d & 1.01f)
  ixus310hs (1.00a & 1.01a)
  sx40hs (1.00d, 1.00g & 1.00i)
  g1x (1.00e, 1.00f & 1.00g)
  g5x (1.00c, 1.01a, 1.01b)
  g7x2 (1.01a, 1.01b, 1.10b)
*

Offline jmac698

  • ***
  • 138
Re: Dynamic Range & CHDK RAW Bits?
« Reply #15 on: 26 / November / 2012, 15:11:03 »
Hi,
I thought someone would mention that :) But according to this reference, for cameras other than Canon it can vary 100-1000:
http://darkerview.com/CCDProblems/CCDProblems.php

Very interesting website, this person was a sensor engineer previously, and heavily into astronomy as well.
Logged
*

Offline reyalp

  • ******
  • 14037
Re: Dynamic Range & CHDK RAW Bits?
« Reply #16 on: 27 / November / 2012, 00:16:21 »
Quote from: jmac698 on 26 / November / 2012, 15:11:03
I thought someone would mention that :) But according to this reference, for cameras other than Canon it can  100-1000:
Phil's post gives the correct answer for cameras supported by CHDK (except the G1 X, I guess). Since the subject of this forum is CHDK RAW, that's likely to be more relevant value for users here.
Logged
Don't forget what the H stands for.
*

Offline KenO

  • ***
  • 122
Re: Dynamic Range & CHDK RAW Bits?
« Reply #17 on: 01 / December / 2012, 11:23:04 »
jmac698,

Thanks for Part 3 - Measuring your Camera  http://www.cloudynights.com/item.php?item_id=2001  link!

What are you using for camera(s)?

Also have you made any tests mentioned in Part 3?

Seems the technology is gradually improving.  Recently saw some previews for Sony Cyber-shot DSC-RX1 Full Frame Compact Digital Camera  which has a  14-bit RAW Image Capture. 



Logged
*

Offline lapser

  • *****
  • 1093
Re: Dynamic Range & CHDK RAW Bits?
« Reply #18 on: 01 / December / 2012, 13:10:31 »
Quote from: reyalp on 03 / August / 2012, 23:11:35
Quote from: KenO on 03 / August / 2012, 12:40:10
Was there a posting for "ewavrs measurements" that discussed what are considered "Good & Bad" measurements for 12 bit cameras?
I don't really understand what there is to discuss. ewavr just made a histogram of the raw values, and saw that even though the raw is 12 bit, there are only ~1000 distinct values (equivalent to a 10 bits)

As I said before, this observation was made for one camera, what other cameras do is unknown until someone cares enough to measure them.
Even if all the values are represented in newer cameras, the actual dynamic range depends on the amount of noise in the sensor (more accurately, signal to noise ratio). Maybe the standard deviation of pixel values in a dark frame shot would tell you something?

In researching info on the "active area" of the sensor, I discovered that it is located on the right side of the sensor, not in the middle. I think this is because the left side contains a masked area, with pixels present but completely shaded from light. It seems to me that the average, or maybe the median, of the values in this area would be the actual "black level" of the sensor, and the standard deviation would be related to noise level. The dynamic range would be the white level minus this black level, and multiplied by the standard deviation over a constant or something.

I'm sure this has all been researched, and there must be a dynamic range formula. I assume that the larger sensor and lower noise CMOS of the G1X would be required to get close to 14 bits. I have the G1X, and the low light capability is pretty incredible compared to my sx260.
Logged
EOS-M3_120f / SX50_100b / SX260_101a / G1X_100g / D20_100b
https://www.youtube.com/user/DrLapser/videos
*

Offline reyalp

  • ******
  • 14037
Re: Dynamic Range & CHDK RAW Bits?
« Reply #19 on: 01 / December / 2012, 17:52:22 »
Quote from: lapser on 01 / December / 2012, 13:10:31
In researching info on the "active area" of the sensor, I discovered that it is located on the right side of the sensor, not in the middle.
These locations depend on the actual sensor in question. As you say, there is usually a masked which is presumably used for something. This thread http://chdk.setepontos.com/index.php?topic=6114.0 has some information, but I'm not sure of the accuracy.
Logged
Don't forget what the H stands for.
Pages: 1 [2] 3   Go Up
« previous next »
 

Related Topics