July 23, 2014, 01:58:59 AM

Author Topic: T3 Sensor - question...  (Read 19398 times)

Woody

  • 6D
  • *****
  • Posts: 515
    • View Profile
Re: T3 Sensor - question...
« Reply #30 on: February 13, 2011, 07:04:51 PM »
The different would be much more significant and noticeable if Canon had applied all of the newer tech while holding MP's steady as Nikon has been doing.


Errr.. you are out of touch with the current state of affairs. Clearly, you have NOT noticed the Nikon D3100 is 14 MP and the D7000 is 16 MP. On the other hand, the 600D has the same pixel count as the 550D, 60D and 7D. So, Canon has been holding the pixel count steady while Nikon is increasing theirs... and soon Nikon/Sony will be going for 24 MP.

Addendum:
In case you think Nikon improved on the Sony sensor technology in their D90 and D5000 as compared to the D300, that is wrong also. There is definitive proof that all Nikon did was to introduce destructive (albeit sneaky) in-camera RAW NR. I have the link at home and can post it later. This is one of those things NOT captured by the limited DXOMark tests.

Most interestingly, the D300s (released after the D90 and D5000) does not have the same in-camera RAW NR. The postulate is the processor at that time cannot cope with the relatively high fps of the D300s (as compared to D90/D5000). It appears that sneaky destructive in-camera RAW NR is practised by everyone these days EXCEPT Canon who wants to continue their support for the small astro-photography crowd. Sony did the same with their RAW files at one point but seemed to have stopped the practice after the barrage of complaints.

Addendum 2:
Proof of in-camera RAW NR on D90:
http://forums.dpreview.com/forums/read.asp?forum=1032&message=32401883
Not sure if the pics will load 'cos the poster, Gabor, has passed away. :(
« Last Edit: February 14, 2011, 10:34:04 AM by Woody »

canon rumors FORUM

Re: T3 Sensor - question...
« Reply #30 on: February 13, 2011, 07:04:51 PM »

WarStreet

  • Guest
Re: T3 Sensor - question...
« Reply #31 on: February 13, 2011, 07:32:09 PM »
Actually I don't have to do a thing, because sites like DxOMark.com already are.  :)  Go fire up a comparison between the Canon 60D, 50D, and 40D, click on the SNR tab and PRINT size comparison and you'll see that they all resolve practically the SAME amount of detail. 

link


Using DXO comparing 40D, 50D and 7D (60D not available for resolution measures) it shows that there is about 3 to 5 lp/mm difference between these cameras even with a low performance lens such as 28-135 wide open. The better the lens, the bigger the gap. One has to convert these numbers into the actual print size.

For example comparing the 7D vs 50D with the 28-135mm @28mm F5.6, the difference is about 5 lp/mm at the center and compared to the center measurement of the 50D which is about 45 lp/mm it means you have more than 10% larger print or cropping ability. 

The 18MP sensor doesn't resolve noticeably more detail than even the 10MP.  The resolution limits of an APS-C sensor with today's lenses has already been reached at about the 12MP level, so what's a 24MP APS-C sensor gonna do?



As mentioned above DXO shows that 7D has better resolving power than a 50D. Sensor resolution and optics, do not limit each other but compliment each other. That's why I used a low performance lens in the comparison. Eventually there is a limit, where more resolution won't improve lp/mm, but it seems we are far from that. 

Why is it that when you jump up to a full-frame 12MP sensor like in the D700, that the 12MP sensor will resolve MORE detail than the 18MP APS-C?  A 21/24MP full-frame sensor will resolve a crap ton more detail than a 24MP APS-C sensor. 


An FF camera has the advantage of using all the lens glass instead a fraction of it on an APS-C, and optics compliment the lp/mm. Using DXO, it shows that 12mp FF can compete with a 18MP APS-C. When I compared the 7D with the D700, with similar category lenses, their lp/mm is not far from each other, with usually a slight edge for the 7D. On the other hand a 21mp FF will have a significant lp/mm over a 12mp FF.


It's why their full-frame sensors with only 12MP are such hot performers.  100% crops from these cameras look spectacular, but every single 100% crop I've seen from the Canon 18MP APS-C sensors have looked soft and fuzzy to me.


If both are compared at 100% crop, the APS-C 18mp for sure will be softer. You need to compare with the same print size. the 18mp sensor will represent a bigger print than a 12mp sensor, when compared at 100%. If you compare with the same print size as DXO do, you should see similar sharpness, as showed from DXO comparing 12mp FF vs 18mp APS-C. Also, the noise gap decreases too, although FF or lower res APS-C cameras, will always win in this department !
« Last Edit: February 13, 2011, 07:39:28 PM by WarStreet »

Rocky

  • 6D
  • *****
  • Posts: 570
    • View Profile
Re: T3 Sensor - question...
« Reply #32 on: February 13, 2011, 10:28:21 PM »
Last night I was looking at the noise performance (only) of 40D, 7D and 60D with DXOmark. At ISO1600 and 3200, 40D is the best performer by  a slight margin. This says a lot about the noise and pixel size.

Sorry, I was looking at the raw data(screen mode). The print mode does give the 18Mp sensors(7D and 60D)an edge on the high ISO noise. Thanks for all the advices of using DXOMark.

S P

  • Guest
Re: T3 Sensor - question...
« Reply #33 on: February 14, 2011, 02:14:57 PM »
Actually I don't have to do a thing, because sites like DxOMark.com already are.  :)  Go fire up a comparison between the Canon 60D, 50D, and 40D, click on the SNR tab and PRINT size comparison and you'll see that they all resolve practically the SAME amount of detail.


WRONG.

The PRINT tab in DXOMark shows images from the various cameras DOWN rez to 8 MP. Take a good look at this site which performed controlled testing on various APS-C cameras and compared their RAW output by UP rez them to 40 MP:
http://pixinfo.com/cikkek/dslr_evolution.3
The 18 MP 7D destroys all other APS-C cameras in that comparison (from Canon 10D to 50D, from Nikon D100 to D300) for image quality at all ISO levels.

Not from what I've seen. 

Just because DxO downsizes to to a standard print size doesn't invalidate their methodology.  What you're suggesting is that the downsizing doesn't show the difference of higher MP count sensors, when in fact it still does.  Compare a D90 to a 60D to a Nikon D3x.  The D3x scores a 43.9dB SNR at base ISO, vs 39dB for the 60D and 38.4dB for the D90.  The extra 6MP from the Canon's sensor buys you a whopping 0.6dB over the D90.  Wow.  The reason the 60D isn't closer to the D3x and its resolution doesn't scale linearly is because it's approaching the physical limits of resolution of that sensor size whereas the D3x isn't quite at the same limits.

As for the comparison you linked, I have no idea what to make of that and have never seen such blurry D200 and D300 images.  It almost looked like they were slightly out of focus or didn't have the same sharpening applied.  And if you look at the other image samples you can even see lighting change from shot to shot, so not exactly the most scientific setup.  I even see camera company advertising directly on the website, which gives me further reason to not trust it.  I've seen other comparison between the Canon 18MP APS-C and the Nikon 12MP APS-C and some full-frame sensors from trusted and camera manufacturer independent sources that show pretty much what DxO's data illustrates.  Basically no difference between 12 and 18 on an APS-C, or if the 18MP does show a little bit more detail it's nothing to write home about when you compare it to a full-frame sensor.  It makes it plain as day obvious that if ultimate resolving power is what you want full-frame is the solution, not more and more pixels crammed into APS-C.  Hence my preference for more megapixel constrained sensors esp on APS-C that focus on other image quality attributes such as dynamic range and color accuracy, etc.

Rocky

  • 6D
  • *****
  • Posts: 570
    • View Profile
Re: T3 Sensor - question...
« Reply #34 on: February 14, 2011, 06:48:23 PM »
After reading more on different photo sites and forums here. I have the following thought:
1. There is s "slight" improvement between the 18MP and 10MP. But is the improvement worth the files being twice as big.
2. The improvement in the 18MP over 10Mp is it due to the sensor itself or due the processor and  14 Bit vs. 12 bit color  depth? Also the improvement shows after the 18MP is down sized to be 10MP.
3. Everything that I have read says the existing lenses (even the L lenses) cannot resolve more than 14 MP on APS_C. How true is it???
4. Based on the DXOMark data, the lens/sensor combination can only resolve about 50 lpmm. That translates to be about 3.4 MP. What is wrong with this picture??
5. If #4 is true, then the debate between 18MP and 10 Mp is almost a waste of time in terms of resolution.
6. One of the site compares different sensor by scaling them all up to 40MP. This is not  a fair test. The 18MP will always MUCH better than the 10 MP in terms of sharpness.
I am not trying to take side. Just hope someone in this forum can shine some light on the above points.

CR Backup Admin

  • Administrator
  • 1D Mark IV
  • *****
  • Posts: 751
    • View Profile
Re: T3 Sensor - question...
« Reply #35 on: February 14, 2011, 07:00:27 PM »
Resolution of a lens / sensor combination is going to always be much lower than a person would expect.  The system resolution is equal to the product of the MTF of the individual components.

If your lens has a mtf of .7 and your camera 0.7, the result is 0.49, much lower than either one.  The way to increase the system resolution is to increase the mtf of the components.

Increasing the megapixel count in a camera increases its mtf.  noise reduction, noise, and AA filters as well as other factors including the firmware also affect the mtf of the body.

This was exactly the same issue with film, each layer of the emulsion has a mtf and the total mtf of the film is the product of the three, then multiply that by lens mtf, and very quickly, the system mtf drops off sharply.

A printer or a computer monitor screen drops it a whole lot more.  Its a struggle to keep the resolution up.

S P

  • Guest
Re: T3 Sensor - question...
« Reply #36 on: February 15, 2011, 02:54:55 PM »
The different would be much more significant and noticeable if Canon had applied all of the newer tech while holding MP's steady as Nikon has been doing.

Errr.. you are out of touch with the current state of affairs. Clearly, you have NOT noticed the Nikon D3100 is 14 MP and the D7000 is 16 MP. On the other hand, the 600D has the same pixel count as the 550D, 60D and 7D. So, Canon has been holding the pixel count steady while Nikon is increasing theirs... and soon Nikon/Sony will be going for 24 MP.

Nikon's pro bodies both APS-C and full-frame are all still at 12MP, with the exception of the D3x which few actually buy.  So I really don't think it's accurate to say that Nikon is cranking up their MP count while Canon is "holding steady".  Yes the D3100 and 7000 have recently gone up to 14 and 16, but that's still less than what Canon has been at for awhile - 18MP.  A prominent Nikon analyst/shooter thinks the only reason they went up that much in the first place was merely to "keep up appearances" with Canon in the whole MP race, and would have been more than happy to stay lower except for that.

Addendum:
In case you think Nikon improved on the Sony sensor technology in their D90 and D5000 as compared to the D300, that is wrong also. There is definitive proof that all Nikon did was to introduce destructive (albeit sneaky) in-camera RAW NR. I have the link at home and can post it later. This is one of those things NOT captured by the limited DXOMark tests.

I've owned a D90 and shot it side by side with a D300 and have seen plenty of comparisons elsewhere, and the D90 is definitely better.  Resolving power is roughly the same, but the D90 has much improved dynamic range which is what DxOMark shows, and what I've seen with my own two eyeballs looking at images that I've taken.  The D90/5000 sensor IS IMPROVED over the D300 in those areas.

canon rumors FORUM

Re: T3 Sensor - question...
« Reply #36 on: February 15, 2011, 02:54:55 PM »

match14

  • Guest
Re: T3 Sensor - question...
« Reply #37 on: February 26, 2011, 09:36:21 AM »

Digic's don't contain "software algorithms". That is in the firmware.  Ignoring for a minute that software doesn't live in hardware.  Amplification is not handled by the Digic either, it's handled by the onboard pixel level signal amps.  Digic are ARM core based (I believe) processors with SIMD functionality.  Digic's are just basic logic cores to control the machine coupled with some simd/matrix multiplication support circuitry.  I would *strongly* suggest you research what ASICs are, and the functionality they general contain before you make any more statements about what the Digic does or does not do.

Although Digic 4 is a processor system, and firmware is in a separate memory chip, the net result is more speed and the ability to process more complex algorithms, and thus improve images over Digic III.

Its just a matter of nitpicking the wording that is in Canon's white paper quoted below, a person does not need to be a engineer and research ASIC's and firmware to know that cameras with Digic 4 produce images with less noise.

From Canon.

DIGIC 4
The DIGIC 4 processor first appeared in the EOS 50D camera launched in Autumn 2008 and is also featured in the EOS 5D Mark II. Compared to DIGIC III and the previous incarnations of DIGIC, the DIGIC 4 offers several advancements.

A more powerful processor can carry out more processing in the same time and therefore perform more advanced processing functions. The DIGIC 4 is actually about 1.3 times faster at signal processing when compared to a DIGIC III and this has several effects. It produces lower image noise for smoother images at every ISO setting and it also allows the extreme expansion (ISO) settings of H1 (12,800) and H2 (25,600) on the EOS 5D Mark II.

The extra processing speed also helps if high ISO noise reduction is enabled. The cameras can retain their burst rate and shooting speed in all but the highest level of noise processing.

   Enlarge imageThe DIGIC 4 processor first appeared in the EOS 50D.

Much of the DIGIC 4 processing is similar to the DIGIC III but it has also added some new features such as Peripheral Illumination Correction to counter the possibility of corner shading, Face Detection AF in Live View mode, and UDMA class 6 card compatibility.

In the EOS 5D Mark II it has added the ability to shoot full HD 1080p movie with sound because the processor can deal with the data fast enough to process it out to the memory card without bottle necks.

Overall, DIGIC 4 means that the cameras capture images with visibly better image quality.

The EOS 7D and EOS-1D Mark IV both feature Dual DIGIC 4 processors. The benefits are the same as having a single processor in the EOS 5D Mark II, but with two processors the processing speed is faster – this allows these cameras to deal with high resolution images at 8fps and 10fps respectively.

The extra processing power has also allowed 50 and 60fps HD movie recording at 720p as well as higher ISO capabilities, with the EOS-1D Mark IV featuring ISO expansion up to ISO 102,400.

Yes there is more to Digic 4 processing than just the processor.  Thats why it is called Digic 4 processing not processor, as it is processing the data so I still stand by my original claim that Digic 4 vs Digic 3 is one of the main differences for the ISO performace of the T3 and XSi.

CR Backup Admin

  • Administrator
  • 1D Mark IV
  • *****
  • Posts: 751
    • View Profile
Re: T3 Sensor - question...
« Reply #38 on: February 26, 2011, 11:48:13 AM »

Yes there is more to Digic 4 processing than just the processor.  Thats why it is called Digic 4 processing not processor, as it is processing the data so I still stand by my original claim that Digic 4 vs Digic 3 is one of the main differences for the ISO performace of the T3 and XSi.

We do not yet know if there is any difference in sensor performance, just that jpeg images look better.  Once there are production cameras and raw files can be viewed and tested, then we can see if the sensor is better. 

A better jpeg image due to a more powerful processor is one thing, but if RAW performance is better, than the sensor or AA filter, etc is likely changed.

You cannot tell with the data existing today, only make a guess.

WarStreet

  • Guest
Re: T3 Sensor - question...
« Reply #39 on: February 26, 2011, 03:51:25 PM »
A more powerful processor can carry out more processing in the same time and therefore perform more advanced processing functions. The DIGIC 4 is actually about 1.3 times faster at signal processing when compared to a DIGIC III and this has several effects. It produces lower image noise for smoother images at every ISO setting and it also allows the extreme expansion (ISO) settings of H1 (12,800) and H2 (25,600) on the EOS 5D Mark II.

Scalesusa, could you clarify if the noise improvement you are referring by an improved processor, is acting before the raw file has been created during the analog to digital conversion, or after the raw file has been created ?

I am assuming that this improvement is done before the raw file has been created otherwise any pc will get a better image. Also, if this is so, There must be a limit on how much processing power the perfect algorithm needs, so once reached, a better processor won't give any improved noise, and only a better sensor can improve it. I am curious how far we are from such a limit. Is this correct or not ?

« Last Edit: February 26, 2011, 03:55:32 PM by WarStreet »

CR Backup Admin

  • Administrator
  • 1D Mark IV
  • *****
  • Posts: 751
    • View Profile
Re: T3 Sensor - question...
« Reply #40 on: February 26, 2011, 05:27:37 PM »

Scalesusa, could you clarify if the noise improvement you are referring by an improved processor, is acting before the raw file has been created during the analog to digital conversion, or after the raw file has been created ?

I am assuming that this improvement is done before the raw file has been created otherwise any pc will get a better image. Also, if this is so, There must be a limit on how much processing power the perfect algorithm needs, so once reached, a better processor won't give any improved noise, and only a better sensor can improve it. I am curious how far we are from such a limit. Is this correct or not ?
[/quote]

This was a quote from a Canon white paper explaining the improvements in the Digic IV over Digic III.

Raw images do not have in camera processing, the data from the sensor is processed in your pc.  This is a simplification, but basically true.

The processing part is where the camera produces jpeg images from the Raw data and saves a jpeg image to the flash card , or perhaps in addition to RAW data.  That in camera processor is nowhere near as powerful as a pc, and, since it has to process the images on the fly, it is limited as to what things it can do.

By processing RAW images in your PC using some of the high powered software available, you can apply selective noise reduction to parts of the image, have a huge range of control over colors, more comtrol over brightness, contrast, and a host of controls to opptimize your image.  you can work on one image at a time, or set the process to run on all images.

The software keeps getting more powerful and continues to keep even high powered pc's loaded down.  I do not expect to ever see the processor in a Camera keep up with what a PC can do, but, as in camera processing improves, so do jpeg in-camera images.

WarStreet

  • Guest
Re: T3 Sensor - question...
« Reply #41 on: February 26, 2011, 06:09:52 PM »

Raw images do not have in camera processing, the data from the sensor is processed in your pc.  This is a simplification, but basically true.

I thought Canon white paper was referring to some type of processing at the analog to digital conversion of the bayer sensor. Since they were referring to the standard processing of the raw file, we can say that the camera processor has nothing to do with the noise performance of the raw file. I knew this, but started to get in doubt after reading  several posts of improved noise with better processors. They were just referring for the processed raw and not for some processing before the raw creation. Thanks for the clarification.


« Last Edit: February 26, 2011, 06:11:25 PM by WarStreet »

CR Backup Admin

  • Administrator
  • 1D Mark IV
  • *****
  • Posts: 751
    • View Profile
Re: T3 Sensor - question...
« Reply #42 on: February 26, 2011, 06:43:04 PM »

Raw images do not have in camera processing, the data from the sensor is processed in your pc.  This is a simplification, but basically true.

I thought Canon white paper was referring to some type of processing at the analog to digital conversion of the bayer sensor. Since they were referring to the standard processing of the raw file, we can say that the camera processor has nothing to do with the noise performance of the raw file. I knew this, but started to get in doubt after reading  several posts of improved noise with better processors. They were just referring for the processed raw and not for some processing before the raw creation. Thanks for the clarification.

Like i said, that is the simple story.    Depending on Camera manufacturer, you can change settings to add in noise reduction.  Canon, for now, lets you turn it off, other manufacturers may not.

It becomes very complex, so just look at it simply, you can spend a lot of time and merely discover that no one except Canon knows exactly what modifications are applied to a RAW image in camera, but it is thought to be very minimal.

canon rumors FORUM

Re: T3 Sensor - question...
« Reply #42 on: February 26, 2011, 06:43:04 PM »