Ad
Ad
Ad
Pages: [1] 2 »   Bottom of Page
Print
Author Topic: Michael's first impressions of the 5D  (Read 4651 times)
Don Miller
Newbie
*
Offline Offline

Posts: 22


« on: September 04, 2005, 02:16:33 PM »
ReplyReply

But more pixels on a sensor adds noise. It's a complex trade off. I wasn't aware there was a noise problem. But I don't shoot to look at the results at 100% in photoshop. I find the noise/grain too low for certain types of prints. Some shots call for a little grit.
Logged
BernardLanguillier
Sr. Member
****
Online Online

Posts: 7526



WWW
« Reply #1 on: September 05, 2005, 11:24:02 AM »
ReplyReply

Some more pre-prod samples from Japan for those who might not have seen these yet:

http://dc.watch.impress.co.jp/cda/review/2...08/29/2180.html

Regards,
Bernard
Logged

A few images online here!
LeifG
Guest
« Reply #2 on: September 05, 2005, 05:32:25 PM »
ReplyReply

I'm not sure what they mean by on-chip noise reduction. My understanding of dark frame subtraction is that an exposure is taken with the shutter closed, to account for thermal noise, hot pixels etc. The D70 can do this, but of course it requires a second exposure after the 'true' one. By default this feature is off.

They might indeed do some pixel averaging, maybe only in dark areas. I guess this is the sort of thing they want to keep secret, and rely on marketing to pursuade is that it is the best on the market.

Out of interest I do wonder if the RAW file is an exact representation of the image i.e. the Bayer grid readings, or whether the on board chip processes the Bayer readings to create the pixel grid we normally associate with an image e.g. 6MP with each pixel have colour and brightness. If the RAW file is not a direct representation, than I could see that on chip processing might be worth having, as it can work with the true raw data. In other words, Neat Image etc work with processed data.

Leif
Logged
Slough
Guest
« Reply #3 on: September 06, 2005, 07:15:26 AM »
ReplyReply

Quote
I think dark frame subtraction would be done with the digic processor. I believe there is more noise reduction going on than just that logic,  as suggested in the posts above. If the camera understanding the tendencies of a each particular senor, this ability would not only reduce noise but allow more acceptable varience resulting in a higher chip yield.
I just don't want my camera smoothing out solid color areas because someone may be judging quality at 100% in photoshop.
I'm expecting the 5D to be very good. Canon should have learned a lot in the three years since the 1Ds. Think about three years before the 1Ds - 1999. Wasn't the best DSLR at that time the 6mp $20,000 kodak? That camera would be laughable now compared to the rebel or the Nikon equivalent. I don't expect that much improvement to the 5D, but digital camera years are like dog years. Three years is a long, long time.
I don't doubt the truth of Jonathan's info and no doubt all manufacturers do something similar. However what he refers to is systematic noise that can be removed with some level of certainty. At high ISO the signal to noise ratio drops, and in this case random electronic noise is also serious. For obvious reasons it is harder to filter without image loss. I have heard some Canon D2s users comment on plasticy tones from the camera and comment that this is due to aggressive noise removal. I suspect this is the source of Don's concern though I don't know if these comments are valid. Like Don I would prefer post-processing removal of random noise unless the manufacturer is sure that on board noise removal has clear advantages.

As an aside I have noticed that noise from Nikon and Canon cameras has a different texture. The D70 is said to have more noise than the 300D, but the noise is more evenly spread and hence is less offensive at modest ISO. At high ISO the D70 is not so good.

Leif
Logged
LeifG
Guest
« Reply #4 on: September 07, 2005, 01:43:47 PM »
ReplyReply

Quote
Nope, that one doesn't exist either.
Oh poo. I'm sure you know which one I mean despite the failure of my currently feeble mind.

Drum roll ...

The Canon 1Ds.

Leif
Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8812


« Reply #5 on: September 04, 2005, 11:43:00 AM »
ReplyReply

Let me start by saying I think the 5D is a great step forward. Full frame 35mm digital for the masses. A worthy goal and definitely achievable. I fully endorse this move by Canon. It's a move in the right direction. At first I thought it wasn't. I would have preferred a D2X style camera. Then I got talking to BJL and realised that Full Frame is the future.  

There's just one thing that worries me. Larger pixels are supposed to have less noise[/b]. When are we going to see this in Canon DSLRs? I hope Michael looks at this issue carefully.
Logged
Ben Rubinstein
Sr. Member
****
Offline Offline

Posts: 1710


« Reply #6 on: September 04, 2005, 12:44:34 PM »
ReplyReply

The 1Ds has noticeable noise at iso 100 in the skies, it's not just the pixel size, it's what you do with it!
Logged

Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #7 on: September 04, 2005, 04:14:53 PM »
ReplyReply

You can always add noise in Photoshop if you like, but there's no excuse for demanding that the camera add it indiscriminately to every shot you take. If you're a real stickler for a particular type of film grain, you can always shoot a gray card with film, scan it, and overlay that over your images in luminance blending mode and set layer opacity to dial in the exact amount of grain you want. Noise/grain is a creative distortion effect, and like any other effect, works well for some images and is inappropriate for others.
Logged

Jack Flesher
Sr. Member
****
Offline Offline

Posts: 2595



WWW
« Reply #8 on: September 05, 2005, 10:43:47 AM »
ReplyReply

Okay, got it...

But I never heard the expression NO NOISE associted with the 20D at ISO 400 when viewed at the PIXEL level.  Yet MR did say that re the 5D.  

Cheers,
Logged

Ben Rubinstein
Sr. Member
****
Offline Offline

Posts: 1710


« Reply #9 on: September 05, 2005, 12:38:48 PM »
ReplyReply

The amount of detail in those nighttime jpgs when opened up using levels is pretty incredible.
Logged

Don Miller
Newbie
*
Offline Offline

Posts: 22


« Reply #10 on: September 05, 2005, 04:28:11 PM »
ReplyReply

I'm concerned that if we push camera manufacturers too hard on noise they will smooth files in-camera in ways we have no control. I would rather have the choice of handling too much noise in post processing. I wonder what Canon means by "on chip noise reduction"? Does that mean that if a pixel is different thatn its neighbors that they replace it's value with the average of the surrounding values? What I want is an accurate camera. A camera that is electronically designed to minimize significantly errant per pixel values. This type of camera will produce the best real resolution. I don't want a camera designed to smooth out even toned areas.
It is possible to design a camera that does a lot of smoothing yet is smart enough to show high resolution when shooting a resolution chart. In the real world of odd shapes and variable tones a camera such as this would have lower real resolution. Be careful what you wish for, you may get it.
Logged
Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #11 on: September 05, 2005, 04:52:45 PM »
ReplyReply

Quote
I'm concerned that if we push camera manufacturers too hard on noise they will smooth files in-camera in ways we have no control. I would rather have the choice of handling too much noise in post processing. I wonder what Canon means by "on chip noise reduction"?
Most likely not. Noise reduction on-chip generally has to do with dark-frame and other fixed pattern noise that is consistent from frame to frame and can be fairly easily removed by the camera firmware without sacrificing detail. It's a whole different class of "noise" than what is handled by Neat Image. Given the level of detail evident in the samples and Canon's prior track record, I'd say your concern is unfounded.
Logged

Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #12 on: September 05, 2005, 06:19:48 PM »
ReplyReply

Canon has indicated that the on-chip processing is related to dark-frame type fixed pattern removal. Some pixels are more sensitive than others, and will consistently output higher values than their neighbors. In addition, the outputs of multiple readout channels need to be balanced out to comparable levels or you'll get a striped pattern across the image. Analyzing the fixed-pattern characteristics of the chip makes it possible for the camera firmware to remove them from the RAW data before writing the file. Dark frame subtraction builds on this to account for variations in the fixed pattern due to ISO setting and temperature.
Logged

Don Miller
Newbie
*
Offline Offline

Posts: 22


« Reply #13 on: September 05, 2005, 11:24:17 PM »
ReplyReply

I think dark frame subtraction would be done with the digic processor. I believe there is more noise reduction going on than just that logic,  as suggested in the posts above. If the camera understanding the tendencies of a each particular senor, this ability would not only reduce noise but allow more acceptable varience resulting in a higher chip yield.
I just don't want my camera smoothing out solid color areas because someone may be judging quality at 100% in photoshop.
I'm expecting the 5D to be very good. Canon should have learned a lot in the three years since the 1Ds. Think about three years before the 1Ds - 1999. Wasn't the best DSLR at that time the 6mp $20,000 kodak? That camera would be laughable now compared to the rebel or the Nikon equivalent. I don't expect that much improvement to the 5D, but digital camera years are like dog years. Three years is a long, long time.
Logged
jani
Sr. Member
****
Offline Offline

Posts: 1603



WWW
« Reply #14 on: September 07, 2005, 04:09:18 PM »
ReplyReply

Quote
Quote
Nope, that one doesn't exist either.
Oh poo. I'm sure you know which one I mean despite the failure of my currently feeble mind.

Drum roll ...

The Canon 1Ds.
Well, you could have meant the Canon 1Ds MkII and mistakenly named it the 2Ds because of the "MkII" part, or even the Nikon D2X, for all we knew.

However, the description of "plasticky tones" doesn't match anything I can recall hearing about any of these cameras (including the original 1Ds). Aggressive noise removal and "plasticky tones" was/is, however, something that the Kodak SLR/c and SLR/n have been accused of. For the aggressive noise removal, see e.g. the DPReview report on the SLR/c.
Logged

Jan
Ray
Sr. Member
****
Offline Offline

Posts: 8812


« Reply #15 on: September 09, 2005, 06:12:36 PM »
ReplyReply

Quote
Quote
first, thanks for the entertaining fantasy about where you got your faith that 24x36mm sensor format is the future.

BJL,
It wasn't a fantasy. Arguing a point can sometimes allow one to see more clearly the basis of the point one is trying to get across. So I have to thank you for that  Cheesy .

Quote
the well capacity has not grown nearly as much as the available sensor area per photosite, due to a continuing disadvantage of CMOS sensors than a larger proportion of sensor area is taken up by other circuitry.

I think this is very likely to be true. We always get information about pixel pitch but rarely about pixel size. I remember on a previous thread some time ago, Michael let slip the fact (unless it was a misprint) that the size of the D30's pixel is a mere 5.25 microns. This compares with a pixel pitch of 10.1 microns. In other words, fill factor is hardly greater than 25%. That of course leaves a lot of room for on-chip processing, which is to be expected because that's one of the touted advantages of the CMOS sensor...... which of course brings us nicely to the question of what is the practical limit to increasingly smaller photosites. The smaller the pixel pitch, the less room there is for on-chip processing, and on-chip processing has its own advantages.
Logged
jani
Sr. Member
****
Offline Offline

Posts: 1603



WWW
« Reply #16 on: September 06, 2005, 12:34:33 PM »
ReplyReply

Quote
I have heard some Canon D2s users comment on plasticy tones from the camera and comment that this is due to aggressive noise removal.
Excuse me for asking, but is this a typo, or is there a Canon "D2s"? If so, what kind of camera is that?
Logged

Jan
jani
Sr. Member
****
Offline Offline

Posts: 1603



WWW
« Reply #17 on: September 07, 2005, 09:14:06 AM »
ReplyReply

(Regarding the existance of a Canon 2Ds.)

Quote
Nope, that one doesn't exist either.
But we sure wish it did! With sensible mirror lock-up and the sensitivity value displayed in the huge, bright viewfinder!

:cool:
Logged

Jan
Jack Flesher
Sr. Member
****
Offline Offline

Posts: 2595



WWW
« Reply #18 on: September 04, 2005, 12:16:07 PM »
ReplyReply

Quote


There's just one thing that worries me. Larger pixels are supposed to have less noise[/b].
I'm a bit confused by your comment -- MR came right out and said that there was zero noise in the 5D files:

"There is no noise visible. None, even at 100% magnification."
Logged

Ray
Sr. Member
****
Offline Offline

Posts: 8812


« Reply #19 on: September 04, 2005, 09:56:37 PM »
ReplyReply

Quote
I'm a bit confused by your comment -- MR came right out and said that there was zero noise in the 5D files:

"There is no noise visible. None, even at 100% magnification."
Yes he did, but I understood that comment referred to an ISO 400 image. I believe similar comments have been made regarding noise at ISO 400 with the 20D. I'm really referring to noise at the pixel level.

Since the 5D has 50% more pixels than the 20D, any image viewed at the same size as a comparable 20D image would appear to have less noise than the 20D even though each individual pixel exhibited the same degree of noise in both cameras.

Since the 5D has significantly larger pixels than the 20D, and since larger pixels are supposed to exhibit less noise than smaller pixels, I'd really like to know if this advantage of the larger pixel has at last been achieved in the 5D. It wasn't with the 1Ds which was slightly noisier than the earlier D60 on a pixel for pixel basis.
Logged
Pages: [1] 2 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad