Ad
Ad
Ad
Pages: « 1 ... 6 7 [8] 9 10 »   Bottom of Page
Print
Author Topic: RAW files: 1Ds3 and Phase P30+  (Read 69129 times)
Frank Doorhof
Sr. Member
****
Offline Offline

Posts: 1513


WWW
« Reply #140 on: June 27, 2008, 01:14:29 AM »
ReplyReply

Just shoot a grey background with a graduate and you will see a difference.
At least between the 5D and the Leaf.
On the 5D I see some stepping/rings caused by bitdepth shortage, on the Leaf file I don't see those.
Logged
gss
Full Member
***
Offline Offline

Posts: 124


« Reply #141 on: June 27, 2008, 02:30:05 AM »
ReplyReply

Quote
The sensel size (linear, not area) of the 39 MPix H39 is a hopping 6.8micron.

For comparison:

Canon 5D: 8.2micron
Canon 1DMkIII: 7.4micron
Nikon D3: 11.3micron

There must be something else there.
I think there is something wrong with your numbers here.  D3 is around 8.7 microns, 1Ds3 is roughly the same sensel size as the H3DII-39, in the mid 6 range (6.4), 5D I believe is right, 8.2.  Oops, you said 1DMkIII, not 1DsMkIII; that one has a 7.2 micron sensel size.
One thing you are missing is that the MFDBs use CCDs and don't give up any of their real estate to processing power.  The CMOS chips may be fixing this soon by putting the engines behind the chips, but as of right now not all of a CMOS is dedicated to capturing photons.
« Last Edit: June 27, 2008, 09:35:55 AM by gss » Logged
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #142 on: June 27, 2008, 10:46:09 AM »
ReplyReply

Quote
I think there is something wrong with your numbers here.  D3 is around 8.7 microns
Right; when I looked up the resolution, I picked the wrong number, namely the middle-resolution.

Quote
One thing you are missing is that the MFDBs use CCDs and don't give up any of their real estate to processing power
Yes; however, those chips have microfilters, which increase the utilization of the light falling over that area. This reduces the sharpness, but that is a different issue. I was not disputing, that MFDBs yield better image quality than DSLRs; I was disputing the particular argumentation based on the sensel size; that is plainly incorrect.
Logged

Gabor
ruraltrekker
Jr. Member
**
Offline Offline

Posts: 69


« Reply #143 on: June 27, 2008, 06:00:21 PM »
ReplyReply

Quote
Just shoot a grey background with a graduate and you will see a difference.
At least between the 5D and the Leaf.
On the 5D I see some stepping/rings caused by bitdepth shortage, on the Leaf file I don't see those.
[a href=\"index.php?act=findpost&pid=203926\"][{POST_SNAPBACK}][/a]

Are you judging this from a print or on screen? I would gather that the screen view could be less than accurate in your particular test.

Ken
Logged
skid00skid00
Jr. Member
**
Offline Offline

Posts: 53


« Reply #144 on: June 27, 2008, 08:32:06 PM »
ReplyReply

Quote
One thing you are missing is that the MFDBs use CCDs and don't give up any of their real estate to processing power.  The CMOS chips may be fixing this soon by putting the engines behind the chips, but as of right now not all of a CMOS is dedicated to capturing photons.
[a href=\"index.php?act=findpost&pid=203929\"][{POST_SNAPBACK}][/a]


Except that Canon and Nikon use *lenses* over each pixel to concentrate the photons onto the appropriate bit of real estate...  It's been that way for at least the last 6 years...

I'd suggest reading Roger Clark's excellent site...
Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8812


« Reply #145 on: June 27, 2008, 09:41:38 PM »
ReplyReply

Quote
Right; when I looked up the resolution, I picked the wrong number, namely the middle-resolution.
Yes; however, those chips have microfilters, which increase the utilization of the light falling over that area. This reduces the sharpness, but that is a different issue. I was not disputing, that MFDBs yield better image quality than DSLRs; I was disputing the particular argumentation based on the sensel size; that is plainly incorrect.
[a href=\"index.php?act=findpost&pid=204006\"][{POST_SNAPBACK}][/a]

Gabor,
I would have thought that, in addition to any sharpness benefit flowing from the absence of an AA filter and microlens, the CCD DB, of same pixel pitch as a CMOS sensor, has a larger photon collector (deeper well, or greater fill factor) and can therefore accommodate a greater dynamic range, all else being equal (which it never is, of course).

In fact, when you consider all the factors that the DB has going for it, it's surprising that the DB is not better than it appears to be.

Consider the following which I believe are statements of fact:

(1) Doubling the sensor size results in any image of same FoV and DoF being comprised of double the number of photons, or double the amount of light at the same ISO.

For example, a 50mm lens on the 1Ds3 has, at F8, the same diameter aperture as a 70mm lens on the P45 at F11 (Aperture Diam=FL/F stop). Allowing for slight discrepancies due to different aspect ratios, the Field of View will be the same; the DoF will be the same; but the amount of light received by the P45 sensor will be approximately double at any given ISO.

(2) Due to the greater size of the CCD photodiode within each photosite, base ISO can be lower; for example ISO 50 instead of the usual ISO 100, assuming equal quantum efficiency. The base ISO of 200 on the D3, in place of the usuall ISO 100, is possible due to greater quantum efficiency. If the D3 photodiodes were the same size as its pixel pitch (as a result of all the on-chip processors being on the reverse side of the chip) then the D3 would benefit from improved image quality at ISO 100 and greater dynamic range.

(3) Any image taken at a base ISO of 50, using a 24x36mm DB, will be a product of 4x the amount of light as that used to produce the same image with a 1Ds3 at its base ISO of 100, assuming equal standards of correct exposure are applied.

Perhaps the surprising thing here is just how good the images are from 35mm DSLRs considering they can't manage to use more than 1/4 of the amount of light that a DB can use, when maximum image quality is sought.
Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8812


« Reply #146 on: June 27, 2008, 10:26:50 PM »
ReplyReply

Quote
Except that Canon and Nikon use *lenses* over each pixel to concentrate the photons onto the appropriate bit of real estate...  It's been that way for at least the last 6 years...

I'd suggest reading Roger Clark's excellent site...
[a href=\"index.php?act=findpost&pid=204098\"][{POST_SNAPBACK}][/a]

I believe the P30 also has microlenses and a base ISO of 100, suggesting its photodiodes are smaller than its pixel pitch.

The purpose of the microlens is to ensure that the light falling on the sensor is not wasted. The light, as far as possible, is directed onto the photon collector or diode rather than onto other devices such as on-chip pre-amplifiers.

Without the microlens, a base ISO of 100 would have to be something like ISO 25 for the same image quality. It would be a very inefficient system.

However, the microlens cannot help the inherent dynamic range limitations of the small photodiode. A small bucket can hold only a small amount of water.
Logged
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #147 on: June 27, 2008, 11:06:41 PM »
ReplyReply

Quote
(1) Doubling the sensor size results in any image of same FoV and DoF being For example, a 50mm lens on the 1Ds3 has, at F8, the same diameter aperture as a 70mm lens on the P45 at F11 (Aperture Diam=FL/F stop). Allowing for slight discrepancies due to different aspect ratios, the Field of View will be the same; the DoF will be the same; but the amount of light received by the P45 sensor will be approximately double at any given ISO
The unit is the sensel, not the sensor. The P45+ has twice the area, but 39Mpix vs 21Mpix of the 1DsMkIII.

My opinion is unchanged: one of the most distinctive features of MFDBs is the higher dynamic range, which is mainly due to the different approach of reading the data (perhaps we could call it UniISO); this requires three to four more bits per pixel, much longer processing and recording time. It is not surprising, that MFDBs' frame rate is counted in frames/minute, not in frames/second.

Any DSLR maker could do the same if they wanted to. This is the main issue.

Of course the wide field of view and huge pixel count are important as well. The lack of AA is just that: the lack of a useful feature, as trade-off for another useful characteristic. One can be happy about the added sharpness because of the lack of AA filter, but one needs to keep in eyes: anyone can remove the AA filter from a DSLR. There is a good reason they are there.

This was different with the earlier lower pixel counts, but now the MFDB pixel sizes are very close to those of DSLRs (or even smaller, see my examples above), and moiree is becoming a consideration.
Logged

Gabor
203
Guest
« Reply #148 on: June 27, 2008, 11:22:59 PM »
ReplyReply

Quote
one of the most distinctive features of MFDBs is the higher dynamic range, which is mainly due to the different approach of reading the data [a href=\"index.php?act=findpost&pid=204108\"][{POST_SNAPBACK}][/a]

Does anyone know of a real side-by-side test which demonstrates this? Link? (I am not talking about anecdotal commentary, but an actual test with both cameras in the same room at the same time.) Just curious as to how much difference there is.
Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8812


« Reply #149 on: June 28, 2008, 12:19:11 AM »
ReplyReply

Quote
The unit is the sensel, not the sensor. The P45+ has twice the area, but 39Mpix vs 21Mpix of the 1DsMkIII.

[a href=\"index.php?act=findpost&pid=204108\"][{POST_SNAPBACK}][/a]

A unit is anything you define as such. The final composition with a specific FoV and a specific DoF, seen or printed at a specific size, is also a unit. Many would consider it the most important unit of all, regarding photographic matters.

Such a unit from the P45+, in best image quality mode at ISO 50, will be comprised of approximately 4x the number of photons as the same unit from the 1Ds3 at ISO 100.

The word photography means 'painting with light'. MFDBs give you more light to paint with. It's as simple as that.

However, CMOS based DSLRs can do a better job with the same amount of light, which of course means underexposure for the MFDB.
Logged
dwdallam
Sr. Member
****
Offline Offline

Posts: 2044



WWW
« Reply #150 on: June 28, 2008, 06:02:04 AM »
ReplyReply

Oh I just had to reply to this thread. We had a discussion in the 35mm forum about the 1DS3 and high ISO noise when under exposed. The result was that when properly exposed, and even better ETTR, the 1DS3 make excellent prints all the way to 1600.

On the other hand, and since I believe the op of this thread made sure both images were ETTR this may not be relevant, the further under exposed you went with the 1DS3 and high ISO, the problem of noise and detail degenerated exponentially.

The reason I mention this is that perhaps the MF outperforms when underexposed and in shadow detail?

This shot was made hand held at 1/20 F2.8 at ISO 1600 using a street light only. Virtually noiseless at 12x18 print size.

I for one am interested in this thread. For one, if you could get the performance out of a Canon G9 that you get out of a MF, would you carry the MF around, or use the G9? Things do get miniaturized as time goes. For that reason alone this thread is important.

I guess my main interest for now is how well the detail and noise of the 1DS3 in shadows stacks up against the MF.
« Last Edit: June 28, 2008, 06:05:09 AM by dwdallam » Logged

Ray
Sr. Member
****
Offline Offline

Posts: 8812


« Reply #151 on: June 28, 2008, 08:42:14 AM »
ReplyReply

Quote
Of course the wide field of view and huge pixel count are important as well. [a href=\"index.php?act=findpost&pid=204108\"][{POST_SNAPBACK}][/a]


35m DSLRs usually have a wider maximum field of view because of the availability of lenses. I believe there's no MF equivalent of a 12mm lens for 35m format.

The larger sensor, and larger photoreceptor within each photosite, are the two main factors which allow the DB to use more light at base ISO, for any given scene or composition. Pixel count has little bearing on this issue.

The P21 with fewer pixels than the 1Ds3, the P25 with about the same number of pixels and the P45 with almost double the number of pixels, can all use about 4x the amount of light as the 1Ds3 at base ISO (assuming that all those backs have a base ISO of 50).

Whether the DB can use that light as efficiently as the 1Ds3 is another matter. This could be determined by comparing the DR of a fully exposed 1Ds3 image at ISO 100 with a P25 image underexposed 2 stops at ISO 50. Assuming the ISOs are accurate and assuming proper technique is used to compare images which are matched with respect to FoV and DoF, the exposure values should be the same for both images.
Logged
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #152 on: June 28, 2008, 11:40:05 AM »
ReplyReply

Quote
I believe there's no MF equivalent of a 12mm lens for 35m format
I doubt, that this lens would come up in a "competition" between FF and MFDB, when the question is image quality.

Quote
The larger sensor, and larger photoreceptor within each photosite, are the two main factors which allow the DB to use more light at base ISO, for any given scene or composition. Pixel count has little bearing on this issue
Here is the secret: the larger pixel count on the same size of sensor implies smaller sensels :-)
« Last Edit: June 28, 2008, 12:52:43 PM by Panopeeper » Logged

Gabor
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #153 on: June 28, 2008, 01:23:21 PM »
ReplyReply

Quote
the further under exposed you went with the 1DS3 and high ISO, the problem of noise and detail degenerated exponentially
This is obvious; the noise depends mainly on the exposure (with a given camera).

Quote
This shot was made hand held at 1/20 F2.8 at ISO 1600 using a street light only. Virtually noiseless at 12x18 print size
The noise in OOF areas is pretty useless in assessing the camera. Plus, processed images reveal nothing. From what I see, most of the noise could have been cut off by blackpoint.
Logged

Gabor
Ray
Sr. Member
****
Offline Offline

Posts: 8812


« Reply #154 on: June 28, 2008, 08:28:56 PM »
ReplyReply

Quote
Here is the secret: the larger pixel count on the same size of sensor implies smaller sensels :-)
[a href=\"index.php?act=findpost&pid=204199\"][{POST_SNAPBACK}][/a]

That's no secret. The secret is the actual size of the photoreceptor, photodiode, photon receiving device (whatever you want to call it) within the sensel. That is, the size of the bucket.

The Sigma 12mm lens might not be too hot, but the Nikkor 14/2.8 is reputed to be excellent. MFDBs have no wide-angle or FoV advantage compared with FF 35mm. DBs are, after all, a cropped format.
Logged
dwdallam
Sr. Member
****
Offline Offline

Posts: 2044



WWW
« Reply #155 on: June 29, 2008, 12:35:04 AM »
ReplyReply

Quote
This is obvious; the noise depends mainly on the exposure (with a given camera).
The noise in OOF areas is pretty useless in assessing the camera. Plus, processed images reveal nothing. From what I see, most of the noise could have been cut off by blackpoint.
[a href=\"index.php?act=findpost&pid=204212\"][{POST_SNAPBACK}][/a]

Yep I agree with your analysis. I actually only posted that image to make the point that at ISO 1600, images from the 1DS3 are quite usable in real world applications, such as print. I have underexposed some images using ISO 800 that were not. Anyway, you're point is well taken. I could post the original RAW if you're interested?
Logged

JeffKohn
Sr. Member
****
Offline Offline

Posts: 1671



WWW
« Reply #156 on: June 29, 2008, 01:14:32 AM »
ReplyReply

Quote
Any DSLR maker could do the same if they wanted to. This is the main issue.
I think Canon really dropped the ball on the 1Ds3 in this regard, by making it the same basic approach as all their other DSLR's (high-speed CMOS, AA filter, etc).

I really hope the supposedly upcoming D3x is built more like an MFDB, with a slower CCD and no AA filter. I would be perfectly happy with 2fps and a base ISO of 50 if it meant more DR and better image quality at base ISO. Unfortunately I think it's unlikely to happen, because the market has decided that high ISO noise is the single biggest criteria by which DSLR's are judged, and if it were to be noisy at ISO 1600 everyone on the internet would criticize it as a failure.
Logged

dwdallam
Sr. Member
****
Offline Offline

Posts: 2044



WWW
« Reply #157 on: June 29, 2008, 01:23:01 AM »
ReplyReply

Quote
I think Canon really dropped the ball on the 1Ds3 in this regard, by making it the same basic approach as all their other DSLR's (high-speed CMOS, AA filter, etc).

I really hope the supposedly upcoming D3x is built more like an MFDB, with a slower CCD and no AA filter. I would be perfectly happy with 2fps and a base ISO of 50 if it meant more DR and better image quality at base ISO. Unfortunately I think it's unlikely to happen, because the market has decided that high ISO noise is the single biggest criteria by which DSLR's are judged, and if it were to be noisy at ISO 1600 everyone on the internet would criticize it as a failure.
[a href=\"index.php?act=findpost&pid=204291\"][{POST_SNAPBACK}][/a]


Well, yes, but having the ability to shoot at ISO 1600 and print 20x30 inch prints that are relatively noise free is a big thing to most people who use the 1DS3 for typical 35mm photography, with the exception of landscape photography. That means you can shoot in light four stops lower than you could originally shoot virtually noise free, which means you can shoot natural light in places you never could before. Or you can at least shoot longer. It is really nice to dial up 400, then 800, and then 1600 and keep shooting like nothing has changed. It also opens up other natural light possibilities that were beforehand impossible w/o gross noise. For studio work, this doesn't apply of course. And you already know all of this.
Logged

203
Guest
« Reply #158 on: June 29, 2008, 10:06:49 AM »
ReplyReply

Quote
I think Canon really dropped the ball on the 1Ds3 in this regard...

I would be perfectly happy with 2fps and a base ISO of 50 if it meant more DR and better image quality at base ISO.
[a href=\"index.php?act=findpost&pid=204291\"][{POST_SNAPBACK}][/a]

Hi Jeff, what are your issues with the image quality?
Logged
KevinA
Sr. Member
****
Offline Offline

Posts: 883


WWW
« Reply #159 on: June 29, 2008, 05:08:22 PM »
ReplyReply

Quote
I think Canon really dropped the ball on the 1Ds3 in this regard, by making it the same basic approach as all their other DSLR's (high-speed CMOS, AA filter, etc).

I really hope the supposedly upcoming D3x is built more like an MFDB, with a slower CCD and no AA filter. I would be perfectly happy with 2fps and a base ISO of 50 if it meant more DR and better image quality at base ISO. Unfortunately I think it's unlikely to happen, because the market has decided that high ISO noise is the single biggest criteria by which DSLR's are judged, and if it were to be noisy at ISO 1600 everyone on the internet would criticize it as a failure.
[a href=\"index.php?act=findpost&pid=204291\"][{POST_SNAPBACK}][/a]

The type of work most suited to 35mm style cameras does not go well with not having a AA filter, 35mm shooters often shoot greater quantities than MF shooters. If it's wedding photography dealing with moire will seriously restrict workflow.
The 1DsmkIII has more moire than the smkII, that will be why it has an AA. not that it will show in most images. Having had to deal with moire on the Kodak SLR/n it ain't a joke.
If the big Nikon came around without an AA it would kill the camera  for the type of work it should excel at.
MF is MF it has advantages and disadvantages over 35mm, depending on what and how you shoot will determine wether the advantages or disadvantages come to the front.
If slower CCD and no AA filter fit with your photography then a MF is the way to go.
MF has more to offer than extra MP's, it is not however best suited to the type of work I do..... more the pity.

Kevin.
Logged

Kevin.
Pages: « 1 ... 6 7 [8] 9 10 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad