Ad
Ad
Ad
Pages: « 1 ... 4 5 [6]   Bottom of Page
Print
Author Topic: The Missing Factor in Sony F828 Reviews  (Read 36218 times)
BJL
Sr. Member
****
Offline Offline

Posts: 5171


« Reply #100 on: February 07, 2004, 02:58:56 PM »
ReplyReply

Quote
It's so easy for manufacturers to pull the blinds over the average layperson's eyes by quoting very appealing and attractive specs without reference to other very significant and limiting factors.
That is why I like working from the specifications provided by companies like Kodak and Dalsa. The other big sensor seller Sony offers only far less informative data sheets, and since Canon and Nikon are not selling sensors directly, they do not publish such data. These specifications are intended for reading by technically skilled customers, not laypersons, and so play it very straight.

   For example, the full data sheet for the KAF-5101CE sensor in the E-1 at http://www.kodak.com/global....pec.pdf reports reports values like total noise [17 electrons rms] and the consequent dynamic range [67dB] that are measured at a very high maximum "guaranteed temperature of performance" of 60C, and at the maximum rated read rate of 28MHz, to give a guaranteed "worst case" value. Measuring at a more normal 20C or 25C lab. temperature would be an easy and somewhat legitimate way to improve the numbers, by reducing the thermal noise level greatly.

   By the way, from reading various Kodak documents, it seems that the total noise is "measured in the dark including amplifier and system noise", so dark current and read noise are there; everything except photon noise, which does need measurement since it is always the square root of the electron count. So it seems that no "significant and limiting factors" are being glossed over there.
Logged
georgj
Newbie
*
Offline Offline

Posts: 6


WWW
« Reply #101 on: February 09, 2004, 02:27:56 PM »
ReplyReply

Hi all,

this thread is very interesting for me, not only because I'm considering to go for a F828 but because I've heard a lot about the basic physical pirciples in digital imaging.
After reactivating my physics (long, long ago) I've made some calculations to find out, how many photons will be projected onto an sensor's photosite. These figures are important when considering the statistical error (square_root(n)) and the maximum possible dynamic range.
[for example, in bright sunlight a typical 3 µm photosite would be hit by roughly 18,000 photons given the 'sunny sixteen'-rule for ISO100 ( 1/100 exposure time, f/16)].
Is there a general rule to get the number of electrons caused by a single photon (I remember a mean efficiency of about 30%)?
Any hints are appreciated.

Georg
Logged
Howard Smith
Guest
« Reply #102 on: February 18, 2004, 09:00:46 AM »
ReplyReply

Ray, I didn't say Darwin became a Christian.    He simply realized that his theory of evolution was fatllly flawed - was not correct.  One does not have to be a Christain to beleve Darwin was wrong.  I have no idea when in Darwin's life he came to the conclusion he was wrong.  And be careful, not everything you read on the web is true.

Something on the order of 999 out of 1000 mutations are ether fatal to the organism or of no benefit.  So if the organism dies, the mutation is not passed on.  If the mutation os of no benefit, it does not assist in natural selection, and will eventually die out.

So what about the 1 in 1000?  Well, suppose an organism evolved a complete eye on its forehead.  A pretty neat trick in itself since the eye is a rather complex organ.  But unless that eye also has an optical nerve and a site in its brain to process the information and understand what it is seeing, the new eye is worthless.  The mutated eye will disappear in futre generations.  The entire optical system would have to evolve in one step.  Never happen.  It is similar to a tornado going through a junk yard and leaving a fully assembled CAnon 1Ds with a 17-40mm zoom in its path. Never happen.  Nature goes from the organized to the disorganized.  Randomness is always increasing.
Logged
Ray
Guest
« Reply #103 on: February 19, 2004, 05:23:35 PM »
ReplyReply

Quote
There doesn't seem to be enough time or material to make hemoglobin by chance.  You need a lot of external help.  A Designer or Creator.
Now we're really into the thick of it, aren't we! I think you're making the same mistake as in the first example. Just as a complex organ such as the eye does not mysteriously and suddenly appear as a result of a chance mutation, the complex process of haemoglobin production does not magically happen in an instant, against all odds of a million trillion to one.

Mutation and chance are ingredients in the process, but the primary mechanism in Evolution is "Survival of the fittest", which is another term which has been grossly misrepresented by the religious fraternity.

No two creatures or organisms within a species are identical. Not even two ball-bearings are identical, and the wide variation in human characteristics is patently obvious. The variation amongst ants or bacteria is less obvious. But it is this variation, when pitted against the environment by the driving force of survival, that gives some members of a species a slight advantage. They survive because they are 'more suited' to the environment they find themselves a part of, are more able to surmount the trials and tribulations of existence. Change the environment rapidly through heavy meteorite bombardment or volcanic activity, as as happened at various intervals throughout the history of the Earth, then you get mass extinctions of species that cannot adjust quickly enough to the changed conditions.

You know I'm right!  Cheesy
Logged
Ray
Guest
« Reply #104 on: February 20, 2004, 05:44:50 AM »
ReplyReply

Quote
Quote
For information on the moth myth, see "Icons of Evolution" by Jonathan Wells

Thanks for that, Howard. I wasn't aware there was a controversy over this classic example that appears in all introductory text books on Evolution.

Here's a critique of Jonathan Wells' book at http://www.ncseweb.org/icons/icon6moths.html

It's fascinating reading. I think you'll find the 'peppered moth' story is not entirely bunk  Cheesy . There's certainly no evidence of fraud, although Kettlewell's methodology might have been flawed. But that's often the case with Science. There are always things we are simply not aware of. Darwin was not aware of the existence of genes and Kettlewell, who did the research on the moths, was not aware that birds can see into the ultraviolet part of the spectrum. He assumed that a moth that looked camouflaged to him would also look camouflaged to the birds.

Quote
Atheism is quite foolish.  ..........I claim that the atheist cannot prove his position, but must take that position on faith - just like a believer in God.

So the believer in God is just as foolish as the Atheist.  Cheesy

Actually I'm closer to being an Agnostic, but like to be a bit provocative now and again and call myself an Atheist, especially when speaking to religious people.  Smiley
Logged
Scott_H
Sr. Member
****
Offline Offline

Posts: 331


« Reply #105 on: January 23, 2004, 06:11:15 AM »
ReplyReply

Maybe the issue is worse than I have had experience with on other digicams.  Either way, I don't think I would base my purchase decision on overexposed images of tinfoil.

Looking at the shots on dpr, there are some where it could be a problem.  The car and the building interior both have visible CA, but to me look like the exposure is about where it should be.

Most of the other shots are deliberatley overexposed to create CA.  Maybe some cameras wil handle that better than the 828, but to me the image is overexposed not acceptable anyway.  That would be my fault, or the fault of lighting conditions at the time I took the shot, and not the fault of the camera.

I still think people are making way too big a deal out of this.  My equipment is always going to be a big compromise.  I can't afford to spend 20K on a camera system.  I can understand the limitations of the equipment I can afford, and learn to deal with it.
Logged

Fabio Riccardi
Guest
« Reply #106 on: January 26, 2004, 03:48:20 PM »
ReplyReply

Quote
I would expect the performance of the Zeiss zoom to be best at around f2.8. But it would have to be significantly better than at f8 to make f8 an unusable solution to PF, especially considering the tremendous DoF advantage at f8, equivalent to almost f32 on a 35mm camera.
In my tests with my F828 (shooting a PIMA/ISO 12233 target) the lens appears to be sharpest around f/5.6.

There is no significant sharpness degradation at f/8.

DOF is significantly deeper than my 10D, although you can get nice OOF backgrounds when shooting at the long end of the zoom.

BTW: IMHO the F828's lens has really excellent bokeh compared to other digicams I've seen.

 - Fabio
Logged
Ray
Guest
« Reply #107 on: January 26, 2004, 11:36:34 PM »
ReplyReply

Howard,
The maths seems to say you are essentially right. The bigger F stop number of the larger format produces a larger CoC. The smaller F stop number of the smaller format (for equivalent DoF and focal length) produces a proportionally smaller CoC.

However, the smaller format has to be enlarged more to produce equal sized prints to the same degree that the CoC is smaller.

Result: back to square one. I can only conclude that the smaller format has no ultimate DoF advantage, but the bigger format does, in terms of a potential shallower DoF if required.

But there is another factor, isn't there! Michael's test shots comparing the F828 and 10D were all taken at f5.6, including the Canon L zooms, and resolution was considered to be on a par. I would be very surprised if those Canon zooms showed equal sharpness at F5.6 and f13.

Can we therefore conclude that the F828 has the potential to deliver higher resolution than the 10D with regard to those areas that are exactly in focus, even though the areas that are out of focus might have the same size CoC's after enlargement to equal size prints?
Logged
Ray
Guest
« Reply #108 on: January 28, 2004, 06:18:18 PM »
ReplyReply

If they are not the same in terms of noise. If one camera or the other has more objectionable noise, at equal shutter speeds and equivalent DoF, (although similar resolution) then at least we shall know which camera has the advantage in the nightclub situation and can avoid much fruitless debate.  Cheesy
Logged
Ray
Guest
« Reply #109 on: January 29, 2004, 06:32:21 PM »
ReplyReply

Quote
Digital is a baby, and a remarkable baby at that.  But just as film was not the ultimate, I doubt digital is either.  
We should not forget that digital cameras, like modern hi fi sets attached to CD and DVD players, are a mixture of digital and analogue processes. In fact, the most significant part of a digital camera, the digital sensor, is still in large part analogue. (If I've understood the process correctly  Cheesy )

We have a photosite or photodetector which accumulates an electrical charge in proportion to the number of photons that have impinged upon it. That's analogue.

As with all analog signals, the accuracy of the match between the signal and what it represents, is degraded by real world interference - in short, noise from whatever source, including noise introduced in the process of converting those millions of different electrical charges to digits.

Now I know you're not so old-fashioned that you don't appreciate the significance of a true digital process  . If that picture information could be digitised at the precise moment the photons impinge upon the photodetector, we'd have a true (and perhaps the ultimate) digital camera.

How this could be possible in practical terms, I don't know, but it's not difficult to imagine a small sensor, say 2/3rds", with literally billions of photosites. Instead of accummulating varying degrees of electrical charge subject to severe degradation, each photosite would be in a state of 'on' or 'off' depending on whether or not it had received a specified minimum number of photons, say 6 being the minimum quantity to rise above the noise floor.

Such a system would probably be comprised of 3 sensors, one for each primary colour. The resulting image would consist of nothing but the presence or absence of 3 dots, red, blue and green, which is all you need - a bit like your computer monitor really.

Of course, it goes without saying that current computers couldn't handle all that information. We'll probably have to wait until quantum computers become a reality. Even at the present time, if Canon were to manufacture a full frame 35mm sensor with the same pixel density as the F828 (which I think would be feasible), it would be aprox. a 130MP sensor producing 390MB images in 8 bit and 780MB images in 16 bit.
Logged
Howard Smith
Guest
« Reply #110 on: January 31, 2004, 09:30:45 AM »
ReplyReply

The resolution of an image is found mathematically by:

1/(image resolution) = 1/(lens resolution) + 1/(film resolution)

So you can see that as film resoltion becomes infinte, the image resolution becomes the lens resolution.  Same foe a perfect lens.

This makes sense if you assume a perfect lens.  The image cannot be "sharper" the the film it is on.  Likewise, an image on perfect filmcannot be sharper than the lens can project.
Logged
Ray
Guest
« Reply #111 on: January 31, 2004, 06:23:30 PM »
ReplyReply

Quote
The resolution of an image is found mathematically by:

1/(image resolution) = 1/(lens resolution) + 1/(film resolution)

So you can see that as film resoltion becomes infinte, the image resolution becomes the lens resolution.  Same foe a perfect lens.

This makes sense if you assume a perfect lens.  The image cannot be "sharper" the the film it is on.  Likewise, an image on perfect filmcannot be sharper than the lens can project.
Howard,
With all due respect, you appear to have missed my point  Smiley . All the above you've written is true. However, the point I am making is this: The current resolution advantage that large format film cameras have (and have always had)over small format film cameras is NOT necessarily due to the superior resolving capability of the LF lens, but due to the 'effective' superior resolving capability of the larger piece of film.

But you clearly know this already because it was you who first referred recently to that fairly old Photodo experiment comparing a high quality 35mm lens at f5.6 with LF 9x12cm format at equivalent DoF aperture setting of f22.

The fact that, even with very imperfect film and current lenses, it is possible under certain circumstance to achieve 35mm resolution that equals 4x5, should be a strong clue that it's only a matter of time before sensor development reaches that level of noise-free quality that makes MF and LF obsolete.

I was initially disappointed in the first test results of the Olympus E-1, and that was due to the limitations of the 5MP sensor. However, I can now see that further sensor development could result in those Zuiko lenses outperforming anything the Canon 1Ds can currently deliver.
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5171


« Reply #112 on: February 03, 2004, 11:58:47 AM »
ReplyReply

Quote
On the issue of larger numbers of smaller pixels being able to do as well as smaller numbers of larger pixels, I think we're really fudging the results a bit, don't you? It's a game in which dynamic range is always going to be limted and the additional noise is always going to be there.
No, it is not fudging in the slightest; if photosite size reduction is not too extreme, visible noise levels will NOT be increased and dynamic range will NOT be decreased, as measured on prints of the same size.  This is the clear if surprising  consequence of the mathematics of randomness and noise, which is in my professional area, so I could give a rather detailed scientific and mathematical explanation; but let me try to wave my hands instead. (The 828 MIGHT have gone a bit too far in site size reduction though.)


Claim

If you makes two sensors using the same basic technology and of the same size, but with different photosite sizes, staying within the size limits for which the electron wells can occupy roughly the same fraction of the total sensor surface and so have give the same maximum electron capacity per unit area, then you will have roughly the same effective dynamic range, fineness of tonal gradations etc., when comparing on the basis of prints of the same size, or through files resampled to the same pixel count.  The worse S/N ratio and dynamic range of more numerous smaller individual pixels is exactly balanced by the greater number of them used to print each given portion of the final image.


Handwaving illustration, for the case of quadrupling the pixel count and then "binning" or downsampling

Combining the signals from a 2x2 collection of smaller sites (say 4microns) into one output pixel will give you the same total signal and total noise as if one had used one bigger photosite (say 8 microns) directly to produce that output pixel; the signal and noise values are just divided into several smaller pieces and then recombined later, instead of being combined on the sensor to start with.

   Since signal values are always positive or zero, but noise values are a mixture of positive and negative, combining the latter causes some cancellation, so the combined noise level grows less than the combined signal level, which is how binning or downsampling gives better S/N ratio than that of the values combined. (the Roper Scientific web site has some discussion of binning.)

   Printing is analogous to binning: prints of the same size might be produced at say 200PPI (8p/mm) from the sensor with fewer, bigger sites or 400PPM (16p/mm) from one with four times the pixel count, and so each small patch of paper (1/200inch square) that gets its input from one pixel in the first case blends together input from 4 pixels in the second case.
Logged
Ray
Guest
« Reply #113 on: February 04, 2004, 08:07:38 AM »
ReplyReply

Quote
Quote
Ray, even Darwin recognized that his theory of evolution was fatally flawed.

Flawed, maybe. Fatally, never!  Cheesy

Quote
You seem to think the cost of medium and large format is too high.  For whom?  You perhaps, but there does seem to a niche market out there.

Actually, I own a couple of second hand MF cameras, a Mamiya RB67 and a Fuji GSW690lll. I would never have paid the new price for these, but I'm a sucker for a bargain.

The fact is, even though I got these at a good price, I can hardly describe them as a bargain because I now rarely use them. They might eventually become collectors items. Since getting the Canon D60 a couple of years ago, I've taken more photos than in my entire life previously, many, many more. The whole process of buying film, getting it developed then scanning it is too much trouble (and expense). I'm now totally 'spoiled' by the convenience of digital.  Cheesy

It's a great feeling to know I can press that shutter as often as I like without it (necessarily) costing a cent. You should try it some time.  Cheesy
Logged
Ray
Guest
« Reply #114 on: February 06, 2004, 09:54:54 PM »
ReplyReply

BJL,
We really need to be educated on the significance of these specifications. It's so easy for manufacturers to pull the blinds over the average layperson's eyes by quoting very appealing and attractive specs without reference to other very significant and limiting factors.
Logged
georgj
Newbie
*
Offline Offline

Posts: 6


WWW
« Reply #115 on: February 09, 2004, 02:41:01 PM »
ReplyReply

something I forgot... 18,000 photons resulting from a bright (isotrope) reflecting white surface.

Georg
Logged
Ray
Guest
« Reply #116 on: February 18, 2004, 12:32:46 AM »
ReplyReply

Quote
Ray, even Darwin recognized that his theory of evolution was fatally flawed.  It's those that don't look deepr that continue to hang on to such notions, usually to prove one of their own view points.
Howard,
Sorry to drag this up again, but I just had a visit from the local pastor, and during the conversation he mentioned that Charles Darwin had admitted he was wrong, on his death bed, and experienced a conversion to Christianity before he died.

This is total misinformation perpertrated by the religious community. No such event happened, and Darwin died an Aetheist or an Agnostic. (Check the internet.)

Nothing much to do with photography, but I thought I'd mention it, just to keep the record straight.
Logged
Howard Smith
Guest
« Reply #117 on: February 19, 2004, 04:45:50 PM »
ReplyReply

Ray, you mentioned that the Pope is not an "evolution denier."  I am surprised you would add him as an expert.  Most Christians do not hold the Pope to be infallible.  In fact, it was the Pope that had Bruno burned at the stake for claiming our solar system was only one of many similar systems.  He also threatened to torture and kill Galileo if he did not renounce Capernicus and the notion the earth is not the center of the universe, but moves about the sun.  Galileo gave in and was sentenced to prison for the rest of his life.  Being Pope doesn't necessarily make you correct.
Logged
A Scottish soul
Newbie
*
Offline Offline

Posts: 9


« Reply #118 on: February 20, 2004, 07:07:05 AM »
ReplyReply

Ray

I too claim to be an agnostic as I personaly doubt large swathes of the Bible and wonder which of all the main world religions is the right one, if any! They can't all be right so are probably all wrong IMHO.

However this has nothing to do with my next question.

One of youre posts in this thread suggested an alternative method of sensor design. 4/3ds size, Foveon style in 3 layers with 256 individual micro photo sites per sensor of 4u size. Here is one possible alternative that would be possible if the technology exists for you're suggestion.!

Instead of using the 256 micro sensors, which are still analog by design, just use one and use it to count photon strikes via a digital counter capable of counting up to the 15000/3 layer = 5000 photons expected for a bright day or more to add extra dynamic range. This would be a 13 bit counter giving 8192 levels ov brightness.  Then you would have a 100% digital system with unlimited dynamic range, depending on the max value possible in the counter. The digital value in the counter would then be read out after the exposure. Speed of operation of the counter would be key to its possibility and also squeezing the small detector and counter registers into a 4u square area. This is unlikely at the pressent but who knows what the nano technologists may be able to do on silicon in the not to distant future!


What do you think about this one?


John
Logged
Pages: « 1 ... 4 5 [6]   Top of Page
Print
Jump to:  

Ad
Ad
Ad