fredmiranda.com
Login

Moderated by: Fred Miranda
  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Canon Forum | Join Upload & Sell

1       2       3              13              end
  

Archive 2007 · 1D3 14bit sample

  
 
Paul Gardner
Offline
• • •
Upload & Sell: Off
p.14 #1 · 1D3 14bit sample


Looking at Larry's dark frame I would suggest that the "notch" is a component error in the circuit. I would send the camera to Canon with the data showing the notch.


Jul 11, 2007 at 12:11 AM
alundeb
Offline
• • • • •
Upload & Sell: Off
p.14 #2 · 1D3 14bit sample


Very nice.

In order to compare the two histograms, you need to adjust the 1D3 intensity values from 14 bit to 12 bit scale. Assuming that the 2 bits are added in the lower end, you need to first subtract 127 (or 128) then divide by 4, and add 127 again. The missing value will then be smoothed out in the average of 4 values.

The standard deviations of the two dark frames are similar in terms of binary codes, so they should be different in terms of DR. Now see what camera you want to throw away



Jul 11, 2007 at 12:54 AM
Pondria
Offline
• • • • • •
Upload & Sell: Off
p.14 #3 · 1D3 14bit sample


Paul Gardner wrote:
Looking at Larry's dark frame I would suggest that the "notch" is a component error in the circuit. I would send the camera to Canon with the data showing the notch.


It is hard to tell what may be wrong if something is wrong. If it is a component failure in the circuit, I think the notch position would change as the ISO changes ( ISO is analog gain set ). My previous ISO testing shows that the notch position is the same @127 ?!?!



Jul 11, 2007 at 08:29 AM
alundeb
Offline
• • • • •
Upload & Sell: Off
p.14 #4 · 1D3 14bit sample


It is simply not possible, to have an analog signal out of an amplifier that NEVER hits a certain level. The notch cannot origin from the analog chain.

If the notch is introduced after the A/D converter, what intensity should the processor substitute for 127 for those pixels? There is no other intensity in the histograms that could take the place for the missing 127.

The raw format from 1D3 is stated from Canon to be linear (without compression). There is no evidence in the data presented here that it's not.

The only place they can theoretically origin, is in the A/D converter itself.

What do the missing values mean in practice? I would say nothing. They are enough separated that the tonal graduation will not be reduced more than 1 stop (from 14 to 13 bits) for each 20-30'th level. That's well into the noise.

I think those missing values are a spinal tap in this topic, and we should concentrate on why the raw converters don't work well for the lowest intensity values.

Canon have admitted a problem i DPP 3.0 and have released a fix in DPP 3.0.2

From their download site:
2. Changes from Digital Photo Professional 3.0
...
Fixed problem
2-2. When significantly adjusting the tone curve for RAW images taken with specific models* and executing [Convert and Save] or [Batch Save], gradation was partially lost; but this problem has been fixed.
* EOS 5D, EOS-1D Mark II N, EOS 30D, EOS Kiss Digital X/EOS DIGITAL REBEL XTi/EOS DIGITAL 400D, EOS-1D Mark III


I downloaded it, and tried it on David's raw file. The triangular shape in the lower end is now gone, but it still doesn't look right. The slider snaps into -11 or -8, nothing in between.



Jul 11, 2007 at 02:30 PM
Pondria
Offline
• • • • • •
Upload & Sell: Off
p.14 #5 · 1D3 14bit sample


Adding to the mystery...

I looked into several more images. For normal shots, there is no notch @ 127. For the shots with large area is underexposed ( like Dark Frame ), The notch exists. And the notch value is not exactly 0. Very small numbers are there like 24 while the neighbor values are in few millions order.



Jul 12, 2007 at 08:28 AM
alundeb
Offline
• • • • •
Upload & Sell: Off
p.14 #6 · 1D3 14bit sample


That's interesting, Pondria.
In normal frames, how many pixels are there at the 127 level, and how many are there around 125 - 130?
Are there enough to say that a notch "should have been there?" I would believe there were hardly any at all.



Jul 12, 2007 at 10:11 AM
Pondria
Offline
• • • • • •
Upload & Sell: Off
p.14 #7 · 1D3 14bit sample


alundeb wrote:
That's interesting, Pondria.
In normal frames, how many pixels are there at the 127 level, and how many are there around 125 - 130?
Are there enough to say that a notch "should have been there?" I would believe there were hardly any at all.


I can post the numbers in the vicinity tonight. THere is no noticeable notch at all. I could visualize the smooth curve.



Jul 12, 2007 at 10:47 AM
Pondria
Offline
• • • • • •
Upload & Sell: Off
p.14 #8 · 1D3 14bit sample


Without the camera ( 1D3 ) in my hands, I really can't produce more data to share As you get 1D3, please join the FM DR measure data base.
https://www.fredmiranda.com/forum/topic/543552/25



Jul 15, 2007 at 10:35 AM
Paul Gardner
Offline
• • •
Upload & Sell: Off
p.14 #9 · 1D3 14bit sample


Here is a very good anlyisis of the DR of sensors. This should clear up a lot of confusion.

http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary/




Jul 18, 2007 at 11:48 AM
fourfa
Offline
• • • •
Upload & Sell: Off
p.14 #10 · 1D3 14bit sample


oh god

round and round and round we go



Jul 18, 2007 at 12:32 PM
Paul Gardner
Offline
• • •
Upload & Sell: Off
p.14 #11 · 1D3 14bit sample


This is a technical discussion. why the negative comment? Just add something of value to it. God isn't going to help.


Jul 18, 2007 at 02:08 PM
fourfa
Offline
• • • •
Upload & Sell: Off
p.14 #12 · 1D3 14bit sample


it's been discussed to death already. did you read the thread?


Jul 18, 2007 at 03:20 PM
Paul Gardner
Offline
• • •
Upload & Sell: Off
p.14 #13 · 1D3 14bit sample


Yes! and when Canon listens and starts using 16 bit ADs I'll shut up.


Jul 18, 2007 at 08:03 PM
Paul Gardner
Offline
• • •
Upload & Sell: Off
p.14 #14 · 1D3 14bit sample


Here is the web page for ADI. they are the top ADc maker. Lot of information on ADs.
http://www.analog.com/en/?sourceid=goo_corporate+search_branded_adi_hp_ptab

http://www.analog.com/en/prod/0,,AD7982,00.html



Jul 25, 2007 at 10:28 AM
Pondria
Offline
• • • • • •
Upload & Sell: Off
p.14 #15 · 1D3 14bit sample


We do not know if 14bit ADC makes difference over 12 bit yet. Too early to talk about 16 bit



Jul 25, 2007 at 10:51 PM
Paul Gardner
Offline
• • •
Upload & Sell: Off
p.14 #16 · 1D3 14bit sample


To paraphrase another thread: If you only have dollars for currency and you buy a 50 cent item, you receive no change. If the item is $1.25 you have to pay two dollars and receive no change. If you have pennies to work with you can receive your correct change. Which would you prefer? Canon can choose to produce 16 bit files as easily as 12 bit files. We just have to stop pussyfooting around and start demanding it. The cost is negligable. They are just milking the updates. rant rant lol


Jul 25, 2007 at 11:02 PM
Pondria
Offline
• • • • • •
Upload & Sell: Off
p.14 #17 · 1D3 14bit sample


I think it's worthwhile to revive this discussion.



Jul 14, 2008 at 10:22 AM
1       2       3              13              end




FM Forums | Canon Forum | Join Upload & Sell

1       2       3              13              end
    
 

Welcome back
Log in to your account