Ad
Ad
Ad
Pages: « 1 [2] 3 4 ... 6 »   Bottom of Page
Print
Author Topic: The Missing Factor in Sony F828 Reviews  (Read 32992 times)
Ray
Guest
« Reply #20 on: January 29, 2004, 04:40:03 AM »
ReplyReply

Quote
I wonder why NOBODY criticizes the biggest flaw in 828 - the quality of the electronic viewfinder.
There have been lots of discussions on this forum regarding the benefits and disadvantages of electronic viewfinders. I get the impression peoples' opinions will vary depending on what they are used to. It's not a big issue for me. However, I don't think it's entirely natural to squint at the world with one eye.  Cheesy
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5086


« Reply #21 on: January 30, 2004, 01:35:52 PM »
ReplyReply

First, an aside: Ray envisions a purely "discrete" sensor with one bit output at billions of photo-sites; that is actually a description of silver halide film! All smooth tonal variation in a traditional black and white negative or print is an illusion produced by blurring of the image from a vast array of tiny, pure black or pure white pixels. (Conclusion; photsites with extremly low S/N ratio and low bit level output can produce wonderful results so long as each individual pixel prints very very small!)

    Second, indeed, "digitial cameras" actually use an analogue electronic sensor, some analogue amplification and maybe analogue noise reduction processing before converting to digital, and many image quality issues depend on performance in this analogue stage, so long as one avoids cutting down to 8-bits. (A/D converters tend to have higher accuracy that the signal going into them.) The sensor does enforce "spatial discretization" though.
The French sometime talk of "electronic cameras", and that is arguably more accurate.
Logged
Howard Smith
Guest
« Reply #22 on: February 01, 2004, 10:01:11 PM »
ReplyReply

OK Ray, now I'm really confused.  On Jan 31, you said that the advantage of LF was not necesarily its superior lens, but the superior capabilty of the larger piece of film.  Then you seem to say that large format lenses are not superior to 35mm lenses.  Whih one is it?

I still maintain that if you put a 4x5pice of that super electronic Fujichrome in a 4x5, it will shoot the socks off any 35mm, just the same as with T-Max 100 or whatever film.  The difference will still be the larger image on the 4x5 does not require the same enlargement.
Logged
Ray
Guest
« Reply #23 on: February 02, 2004, 08:39:38 PM »
ReplyReply

Quote
Quote
first, I have asked indirectly before but am still curious: how do you come to your conclusions about the resolution and noise performance of the E-1?

Only from lab test results. I accept that a slight increase in noise at 18% grey on a chart, or very slightly less resolution compared to the 10D at 100% on a screen can amount to nothing on a print. I'm trying to be objective using terms that are relative. The E-1 sensor seems to be the weak link in the chain. People often comment that the Zuiko lenses are as good as Canon L glass. In fact, the MTF charts (if we can believe them, because I know of no independent MTF tests) would indicate the Zuiko lenses are considerably better than any Canon lens that I've seen, in relation to its MTF performance at least. Some of these Zuiko lenses appear to have the same performance at 60 lp/mm as the best Canon lenses have at 40 lp/mm.

Apologies to any E-1 owners who get the impression I might be trying to disparage the performance of their new pride and joy.  Cheesy

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, sometimes very apparent, sometimes hardly apparent and sometimes not apparent at all, depending on print size and what clever strategies have been put in place, or employed in post-processing, to reduce the noise.

A fully digital sensor is the way to go, whether along the lines I've proposed or not. At the risk of stating the obvious and teaching people how to suck eggs, it must be apparent that the enlargement issue disappears with a 'true' digital sensor. The number 'four' means what it means, whether it's writ large or small.
Logged
Ray
Guest
« Reply #24 on: February 05, 2004, 09:01:26 AM »
ReplyReply

Quote
which conditions were convoluted?
I was hoping you wouldn't ask me. That was my euphemistic way of saying 'confusing and contradictory'.  Smiley

After explaining why larger numbers of smaller pixels should or could have the same performance (when summed) as an equivalent number of larger pixels, you then proceeded to give an example of binning which appears to demonstrate the opposite.

Binning only serves a purpose because it reduces the total noise of the pixels that have been binned. There's an implication in Roper Scientific's explanation of this term that Read Noise might be the same for both large or small pixels.

Unfortunately, Roper Scientific seems to deal with CCDs rather than CMOS devices, so their figures might be out of date. They give examples of Read Noise figures as low as 5 to 11 electrons for CCDs with a full well capacity of 45,000 to 375,000 electrons. That's certainly insignificant for large pixels at full saturation but could be a problem for small pixels in low light situations.

I'm tempted to speculate that the F828's high noise at ISO settings of 200 and above is due to the prominence of Read Noise relative to the signal. On the other hand, I read somewhere that CMOS sensors can now have a Read Noise as low as 1 electron.  Huh

Thanks for pointing out that mistake of F13. That is of course equivalent to the F828 at F5.6. F20 it is then (or nearest setting).  Cheesy
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5086


« Reply #25 on: February 09, 2004, 10:44:18 AM »
ReplyReply

Quote
I've come across statements that nothing can be done about photon shot noise. We're up against fundamental laws of Physics. But that doesn't seem quite true to me. You can always reduce photon noise by letting in more light, ie. reducing ISO, and at the same time effectively reduce all (?) the other types of noise.

Do you know of any fundamental limits?
Fundamental limits are the easy ones! All one needs to know is the appropriate physics, not the engineering trade secrets of the camera companies.

    The most famous fundamental limit is what I call quantum mechanical counting error noise, both photon noise and other sources based on counting photons and electrons. This limits S/N ratio to the square root of the number of items counted. It is the one that says, if photosites have a maximum electron count proportional to area, decreasing photosite size will decrease photosite S/N ratio in proportion to the linear dimension of the photosite, so shrinking sensors reduces possible dynamic range for a given amount of spatial detail [i.e. a given pixel count].

   The only way to mitigate this problem is your idea of allowing the sensor to handle more light (quantum mechanics says that more light = more information = greater total combination of spatial detail and tonal gradations). For example, deeper electron wells, or Fuji's twin photodetector idea, or an idea that has been floated of measuring how long it takes for highlight pixels to blow out and hence determine highlight luminosity levels at blown-out pixels by extrapolation, or my new idea of repeatedly reading out and emptying photosites at high rates to avoid any overflows.

   A possible second almost fundamental limitation is optical limits on minimum aperture ratio, related to optical distortion and maximum deviation from perpendicular of the angle of incidence of light on the sensor. Digital cameras will probably never be able to make much use of aperture ratios faster than f/1, and f/1.4 is a more plausable value: with lower aperture ratios, the extra light coming from the outer edges of the wider aperture will scarcely be detected.  This is the one that is likely to cause maximum usable system speeds (lens plus sensor) to decrease in proportion to sensor/photosite area.

P. S. The trend of high end digicams with lots and lots of tiny, roughly 2.7micron pixels has now spread to almost every major player: Canon, Konica-Minolta, Nikon and Sony with 8MP 2/3" format, Fuji with 6MP 1/1.7". Is it possible that none of them understands the inherent undesirability of such small photosites, or that Canon, Nikon and Konica-Minolta have decided to produce raer expensive products that are both inherently flawed and enrich major competitor Sony by using their 8MP sensor. Canon has stuck their neck out on the 8MP PowerShot Pro1 by calling it a "pro" product and giving the "L" designation to the lens.
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5086


« Reply #26 on: February 09, 2004, 05:11:27 PM »
ReplyReply

As I said a while  ago, in principle a larger format can do everything that a smaller format can, if you are willing to bear the price and weight: the basic stategy is for the larger sensor to have the same pixel size as a smaller one, allowing one to effectively crop down to the smaller sensor when that gives the best results. Maybe some shot focal length lenses of smaller image circle and faster maximum lenses (like the new wave of "digital LF lenses") would also be useful when doing such cropping.

    However, this seems about as relevant as when some large format zealots ridicule 35mm and even medium format, or when LF users and writers like Roger Hicks in Popular Photography manage to ignore the more restrictive shutter speed needs of making larger format images. It is certainly not relevant to the original, long lost topic of this thread; deciding whether the Sony 8MP 2/3" sensor is a "Good Thing" for some photographers at least.
Logged
Howard Smith
Guest
« Reply #27 on: February 10, 2004, 11:20:55 AM »
ReplyReply

Ray, I can live with the 828 being a 2/3" camera with a 7-50mm f/2 lens because I am pretty sure that is what it is.  BJL, I will freely admit that the 828 is a "good thing" for at least some people.  In fact, it is too much of a good thing for my Aunt Tilly, and not so much of a good thing for others.  Michael carries one around and uses it sometimes, but not all the time.

You certainly didn't need to get into photon counting to convience me of that.  And why level the field?  If I can take "better" images with my Super Zoomoflex, why do I have to crank up the ISO and stop the lens way down to show it doesn't take "better" images than an 828?
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5086


« Reply #28 on: February 18, 2004, 03:35:40 PM »
ReplyReply

Quote
BJL, why can't a debate of God be rational? ...
I am sure it can be: I was an enthusiastic student of the philosophy of religion in my younger days, and still follow various religous debates, and not just ones about the infalibility or evil of various camera makers. On the other hand, it seems that on the internet as a whole, it is hard enough to keep even a debate of photography rational!

(I suppose that on page ten of a thread, we can discuss almost anything, since almost nobody is listening any more!)
Logged
Ray
Guest
« Reply #29 on: February 19, 2004, 06:02:55 PM »
ReplyReply

Quote
Ray, I certainly don't know you are right.

For natural selection to work, the mutation must be beneficial or attractive for the organism - give it an edge.
Well, I thought I'd already explained that  Cheesy . Mutation is not the driving force or primary mechanism of Evolution. I'll give you an example. A species of moth in pre-industrial England is a pale brown colour that blends in nicely against tree trunks etc. This protects the moths from being gobbled up too frequently by predatory birds. There's a balance created.

Along comes the industrial revolution and covers everything in a layer of soot (no exaggeration! We've got it good with our clean air policies). The once brown tree trunk is now grey. This species of moth almost gets wiped out. It can't hide itself and is picked off too easily by the birds.

Thank God for variation within the species! Just a few moths are a slightly darker shade of brown. They are the ones who tend to survive and their offspring are also a slightly darker shade of brown. And amongst those offspring there is variation and some are an even darker shade of brown. Within 50 years or so, that species of moth has completely changed colour and is thriving as strongly as ever.
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5086


« Reply #30 on: February 20, 2004, 08:31:10 AM »
ReplyReply

Quote
Keep in mind that ONE misplaced acid is usually fatal. Sickle cell anemia is an example.
I will not even begin to try to change anyone's religious views based on reading or religious writers, but simple errors of fact I will address. it is far from true that "One misplaced acid is usually fatal"; on the contrary, there are considerable variations in DNA that cause no phenotypic effect, and considerable genetic variations in the the details of proteins that have little or no effect on their functioning.
   Also, sickle cell anemia is a very poor example; it is related to a genetic variation that is overall beneficial in Africa, where it is most common, by imparting a degree of resistance to malaria to those who inherit that version of the gene from only one parent; only the smaller number of people who inheret it from both parents suffer sickle cell disease. And I believe it is far more than a one acid change away from regular haemoglobin.
Logged
Scott_H
Sr. Member
****
Offline Offline

Posts: 331


« Reply #31 on: January 22, 2004, 08:52:52 PM »
ReplyReply

Typically CA is only going to show up in extremely high contrast situations.  In those situations the image probably isn't going to be acceptable anyway; since you are probably exceeding dynamic range, and chances are your highlights are blown.
Logged

BJL
Sr. Member
****
Offline Offline

Posts: 5086


« Reply #32 on: January 23, 2004, 04:06:31 PM »
ReplyReply

Quote
Lenses have normally their best performance at 2-3 f-stops from being wide open. After that their performance slightly decreases.
I wonder if that guideline carries over the the world of far smaller image circles and focal lengths?
   One suggestion at Norm Koren's site is that lens abberations starts to be the dominant factor in image quality at f/4 and wider, and he implies that this is an inherent limitation in optical design (dealing with coma, spherical abberation and all those other things that I know little about except their names).
   So maybe we should be looking for small 2/3" format digicam lenses to be at their best roughly between f/4 (to limit aberrations) and f/8 (to limit diffraction).
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5086


« Reply #33 on: January 25, 2004, 08:07:41 AM »
ReplyReply

Quote
Who needs great (or greater) depth of field at focal lengths of 7 - 50 mm?
In the 2/3" format of the Sony 828, those focal lengths cover the same angular field view range as about 28mm to 200mm in 35mm format (as I am sure you know, but for some reason overlooked), which covers the FOV choice in a great majority of all photgraphs. So the answer would seem to be "almost anyone who wants lots of DOF in a photo". Like me with almost every photograph I take of landscapes, urban scenes, and close ups. (It seems to be that the DOF characteristics of small digital sensor formats are particular attractive for close-up/macro photography.)
Logged
Howard Smith
Guest
« Reply #34 on: January 26, 2004, 12:00:14 PM »
ReplyReply

Perhaps this would make more sense and be easier to understand if everyone used the same definitions and the one most commonly accepted.  Depth of field is simply the distance in front of plus behind the plane of focus that prodcues a circle of confusion defined by the user as "acceptable" for sharpness.  Frm that, you can decide for yourselfwhat is sharp and what isn't.  Somewhere someone said something about at what f/stop does a Canon50 mm f/1.4 lens bwcome "razor sharp."  Well, that is a very qualitative staement.  Maybe for some it is "razor sharp" at every f/stop and for some it fails miserably at every f/stop.  It is dangerous to make critical comparisons using ill-defined terms to describe qualitative traits.
Logged
Ray
Guest
« Reply #35 on: January 26, 2004, 09:29:55 PM »
ReplyReply

And I thought I could be crystal clear on this matter. I don't know what came over me!  Cheesy

I'm afraid BJL is right, Howard. It's not just the size of the circle of confusion that counts, but its size in relation to the whole image.  A CoC of diameter 0.01 inches might be insignificant on an 8x10" film, but very significant on a 2/3rds format sensor.

However, you are quite right that in order to calculate the diameter of the CoC (on the film or sensor) at a particular distance from the camera, say with the lens focused at infinity, the format or size of the film or sensor is irrelevant and doesn't come into the equation.

As I understand it, the fundamental formula for such calculations, with lens focussed at infinity, is CoC = focal length squared, divided by (f stop x distance of subject from camera). Focal length and distance should be in the same units of course.

What I'm trying to find out is; can the small format camera do anything that the larger format cannot? We know that the small format digicam is at a disadvantage as regards shallow DoF. Is this counterbalanced by a greater DoF capability whilst still maintaining good resolution?

According to Fabio, the F828 Zeiss lens is sharpest at F5.6. This seems to accord with the aperture Michael chose in his test review of the F828, comparing it with the 10D.

Now, according to my calculations, the F828 at F5.6 gives the same DoF as the 10D, at equivalent focal lengths, using an aperture of f13.5 (2 1/2 stops difference). I don't think any of my 35mm lenses would have their best performance at f13.5. At that aperture, it probably wouldn't make any difference what type of lens one used, a $5,000 Canon prime or a $500 Sigma zoom.

I'd be really interested to see how these two cameras perform at these relative apertures. Will the F828 really shine, or will resolution still be roughly on a par with the 10D?

What about comparing the F828 at f8 with the 10D at f19?

In the meantime, I'm going to work on that basic formula and see what the CoC's should be. However, my experience tells me there's often a discrepancy between theory and practice.
Logged
Howard Smith
Guest
« Reply #36 on: January 26, 2004, 09:30:53 PM »
ReplyReply

Here's an exercise for you, BJL.  I will cut a 1x1 inch piece of film from a negative I have taken.  I will tell you the lens focal length, the f/stop used and the focus distance.  Please calculate the format for me before I cut the film.  If you can tell a 1x1 inch piece from the middle of a 4x5 taken with a 150 mm lens @ f/8 from a 1x1 inch piece taken cut frm a 6x6 cm taken with a 150 mm lens @ f/8, I will be completely amazed.  You can't.  Format has no input.  Yes, I suppose you might get a hint from the thickness of the film base, but that doesn't count.

Here's the truth.  I you want to compare a print from a small digital sensor to a print from  a large format negative, you will likely choose a much smaller circle of cofusion for the digital image because you will likely need to enlarge it more.  BUT, if the image size (the size of the bord on the wire) i the same for both prints, then only the f/stop matters, given the same focus distance.  The format and focallength have no input.

Last excercise.  Read Michael's tutorial on Understanding Depth of Field.
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5086


« Reply #37 on: January 27, 2004, 08:43:42 AM »
ReplyReply

Quote
Here's an exercise for you, BJL. I will cut a 1x1 inch piece of film from a negative I have taken. ...
Howard, some comments and questions.

Comments:
1) You are the one who needs to reread Michael's essay of DoF. He points out that the CoC choice is computed backward from a roughly one minute of arc resolution limit fort the eye, and a resulting 1/6mm resolution limit on prints viewed from about 15" [50cm], and thus CoC choice depends on the expected maximum magnification (about 5x traditionally for 35mm format). He goes on to mention that for larger formats like 8"x10", CoC is typically computed using a larger CoC value due to the lower magnifications factors likely to be used in printing. It is exactly the same reasoning which says that for digital cameras with very small sensors, small pixels and hence large magnification factors often used in printing that smaller CoC values should be used.

2) You comparison of equal areas of film is mostly relevant if one prints with the same magnification regardless of format and view the prints from the same distance despite their different sizes. In our digital example, it would apply to making a 12"x9" print from a 2/3" camera and a 50" by 33" print from 35mm format, and viewing them both from the same distance. Even if one were to make such a huge print from 35mm format, it would typically be viewed from far further back, which increases the perceived depth of field because of that roughly one minute of arc resolution limit mentioned in Michael's essay.


Questions

Have you noticed that the DOF calculator I mentioned adjusts the circle of confusion roughly in proportion to format size, (as does every DOF calculator covering digital formats that I have seen)? If so, do you claim that they are all wrong, and that instead they should be using the same CoC value regardless of format?
Logged
Ray
Guest
« Reply #38 on: January 27, 2004, 06:32:38 PM »
ReplyReply

Quote
you seem to working harder than necessary to prove an obvious advantage of larger formats.
BJL,
That's odd!  Smiley  I thought I was doing the opposite. I'm trying hard to find a 'real' advantage of the smaller format, apart from the obvious fact that smaller format cameras are, or can be, lighter and less bulky.

I'm still a bit uneasy about the way Michael has compared the F828 to the 10D. I get a sense of apples and oranges. There doesn't appear to be a level playing field. On the other hand, it might be (more likely  Smiley ) I just don't understand the significance of his methodology.

When taking any photo, major considerations are; what angle of view? what depth of field? what shutter speed?

With modern digital cameras with adjustable ISO, selection of shutter speed is also going to be influenced by increased noise considerations. (Perhaps more than is necessary in my case, but that's another matter.)

When comparing two quite different cameras such as the F828 and 10D, in order to keep the playing field level and compare apples with apples, it seems to me that one should keep equivalent focal lengths (same angle of view), keep equivalent DoF (different f stops), and keep the shutter speed the same (different ISO settings).

Michael doesn't appear to have done any of this (at the camera level) in his comparisons. He seems to have used the same aperture for both images (f5.6) which ensures that the 10D image will have less DoF. He seems to have used a Canon lens at 200mm, which is the full frame 35mm equivalent focal length, not the 10D focal length equivalent, and by comparing the F828 at ISO 64 and/or ISO 100 with the 10D set at ISO 100, he's given the F828 simultaneously a shutter speed advantage and a noise disadvantage, which sort of confuses the issue.

If I were setting up these two cameras for comparison, I'd be comparing the F828 at 200mm and F5.6 (actually 50mm and f5.6) with the zoom on the 10D set at 120mm and f13. Or, I'd use f4 on the F828 and f9.5 on the 10D. This would ensure equal DoF for both images.

I would also set the F828 at ISO 100 and the 10D at ISO 400 to ensure equal shutter speeds, as well as ISO 64 and ISO 250 (if that were possible). Clearly, exact equivalence is not always possible.

Now, before anyone jumps on me for criticising Michael's methodology, let me say that his methodology appears to understate the performance of the F828, and this might be quite deliberate in order to deflect criticism.  Smiley
Logged
Ray
Guest
« Reply #39 on: January 28, 2004, 05:46:20 PM »
ReplyReply

Quote
P. S. want to redo this whole debate with the next 8MP 2/3" format digicam, the Nikon 8700 with is 8.9-72mm [35-280 equiv.], f/2.8-4.2 lens?
No need, if Michael includes my testing recommendation of level playing field  Cheesy .

We need to know the maximum image quality attainable, despite different settings and different equivalent settings, for each camera, as well as image quality at equivalent DoF and same shutter speed.

What seems to be happening at present is a state of confusion where someone who owns a small digicam might be under the impression that he/she can take a photo in a nightclub with no flash, at f2 and get reasonable DoF and results that cannot be achieved with, say the larger 35mm camera which would produce an undesirable shallow DoF at F2, and that therefore the small digicam has some advantage in this respect.

This is, I believe, illusory. One simply stops down with the larger format to get equal DoF and goes up in ISO to maintain the same shutter speed, and the quality of the images should then be roughly the same, at least in terms of noise.
Logged
Pages: « 1 [2] 3 4 ... 6 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad