igor_jarm

<p>Hm, I must say I agree with both sides.<br /> <br /> Why shouldn't one selectively obliterate parts of the sky? Not that ETTR means doing that, but brightest parts of sunlit clouds will for all practical purposes always be rendered pure white. So will the sun.<br>

I always used ETTR for preserving shadow details, just that it wasn't subjectively in the shade even though objectively it was. Using the equivalent of ND filters or other techniques I can then bring those shadows out of the dark without too much noise.<br /> Two examples:<br>

<img src="http://farm3.static.flickr.com/2523/3935120759_9a9e23c44f.jpg" alt="" width="333" height="500" /> <br /> <a href="

<img src="http://farm3.static.flickr.com/2541/3935120753_6b9ea4433e.jpg" alt="" width="500" height="333" /><br>

<a href="

But prefer to call the method I used digital zone system instead of ETTR.<br>

<br /> On the other hand it's a fact that some (if not all) digital cameras grossly underexpose. While a tipical sunlit scene will be metered at sunny 16 rule, I can safely overexpose my Canon 20d by 1.5 stops (2 stops for most scenes) and not blow any highlight. Why then wouldn't one make the best possible exposure? It's like shooting at ISO 25.</p>

<p>Dale, digital noise is not like hiss, it's actually photon noise. Only in the lowest tonal values it is due to signal gain and enhancement circuitry.<br>

D.B., you can't get around noise in high ISOs, today's sensors are actually good enough to actually count the photons collected at each sensor's pixel. At night there are just,... well not that many photons :) If you have 10 photons here and 15 there, what can you do about it?<br>

See http://www.clarkvision.com/imagedetail/digital.photons.and.qe/index.html</p>

<p>Scott,<br>

I'm actually just arguing that grossly misfocusing certain lenses has nothing to do with how tricky or delicate the autofocus process is. In these cases I'd not even call it autofocus it's something totally useless. Photographing something from distance of 50cm with accuracy of less than 5cm is hardly high expectation, isn't it? The lens is optically and mechanically decent, yet the camera can't focus it. Someone is not doing their job right, and I think it is Sigma, but then again it is the camera that does the focusing not the lens.<br>

Again this is not the problem the OP is experiencing although it seems like it is.</p>

<p>Alec,<br>

Just for you I pulled out my old EOS 1000, which as you know was the budget camera in those days.<br>

Fortunately my Sigma lens mounts this camera and guess what? In my digital viewfinder I could correct the wrong close focus by the image I saw, and in this almost 20 years old camera focus looks ok in viewfinder.<br>

You'll just have to believe me this.</p>

 

<p>Arie,<br>

I don't believe there's no feedback in the auto focus process.<br>

There are two instances, where it's obvious, there's some feedback involved.<br>

1) when the Canon camera starts with severely misfocused lens, it tries first in the direction of closer focus, if it doesn't find focus there it goes to infinity direction<br>

2) my lenses sometimes "hunt" for the best focus back and front in progressively smaller steps (my guess, never more than 3 steps)<br>

If there's no feedback involved there, I really don't know what feedback is.</p>

<p>Alec,<br>

You make these things sound way too complicated, 20 years ago cameras autofocused much better than my digital with this Sigma lens today. As far as I'm concerned their postdoctoral graduates could stop development in the 80s.<br>

I understand it's not in Canon's interest to make third party lenses focus better on Canon cameras, but I'm sure they could do it if they wanted.<br>

But this actually has little to do with the OP's inquiry, where I'm sure it's due to slop in the cheap lens mechanics. I also have this 50mm lens and sometimes it focuses ok, sometimes in front and sometimes in back. Not like my Sigma lens.</p>

<p>Alec,<br>

Sorry, I just wanted to make clear that mechanically and optically there's nothing wrong with the lens.</p>

<p>Also, if I may return to what you said earlier:<br>

<em>That's just "multiple goes", not a feedback loop.</em><br>

If an autofocus algorithm can't progressively focus from gross misfocus to acceptable focus, I'll call that algorithm dumb. In my test of repeatedly half pressing the shutter button, the camera repeatedly focuses back and forth around the correct focus and always ends with about 5cm (2") of front focus.</p>

<p>I still say, somewhere else is the problem, so I'm actually not calling Canon focus algorithm dumb, you see, I think this (all?) Sigma lens is "dumb" because I guess it doesn't communicate with the camera properly. That's all I'm saying.<br>

In the end none of us really knows how auto focus really works, we're mostly just guessing.</p>

<p>I'll be bold and suggest an algorithm improvement to camera makers: Why not just check if the lens is focused in the end and if not, focus it the "dumb" way, the way we humans focus manually with the split screen.</p>

<p>Alec, manual focusing via split prism is trivial, there's no need for a Ph.D. thesis to describe what I'm doing when I correctly manualy focus my old manual Canon TX. Same with phase-detection autofocus. My profession is computer programming and I know a bit about ways of finding acceptable results. Making a lens with tight enough tolerances for this is another matter, but my sigma is an "USM" lens with "no" play in the mechanics, and certainly orders of magnitude lower than the amount of misfocus would make you think.<br>

The error my camera makes when focusing this sigma lens has nothing to do with that. It is grossly misfocused, the problem lies somewhere else. The lens is also way too sharp that you could blame optical misalignment for this.</p>

<p>I have a Sigma 30mm f1.4 which is notorious for bad focus and a canon camera.<br>

If autofocus is a process with no feedback, than I can introduce this feedback with multiple half presses before I take a picture. It doesn't work, my lens always front focuses.<br>

Even If I set focus to manual and rely on focus confirmation light it still front focuses.<br>

If I rely on focusing by sight I can always focus better than autofocus.<br>

This is realy only a problem when focusing close.<br>

I'm not sure where I read it or if I read it at all but canon lenses send additional data to the camera to microadjust focus sensors. That's because of different aberrations in different lenses. If the lens does not send any data or if it sends incorrect data, focus sensors will be off.</p>

 

<p>Rishi, like I said, I don't do any of this, except for occasional test. I do it the only way I know how, by sequentialy merging 50% transparent layers. That's why I merged 16 exposures: 16>8>4>2>1. I stopped at 16 because 32 would be far too much work for me :)</p>

<p>Rishi, a few ideas, how to remove the horizontal banding:<br>

1) take 4 or more exposures, average them so that temporal variations cancel out, and subtract an average of a number of dark frames to get rid of fixed noise pattern. I believe most of what will remain in the shadows will be foton noise (at high ISOs).<br>

Obviously you can only do this with static scenes.<br>

Actualy, doing that at high iso, is somewhat equivalent to taking a long exposure at low iso, except you can get even more dynamic range by not averaging, but by making a sum of exposures. You'd of course need a sum of the same number of dark frames made at the same ISO to effectively get rid of all non image forming noise.<br>

2) You could also make only 2 exposures in manual exposure mode one at say ISO 100, the other at ISO 1600. ISO 1600 would be used to extend the dynamic range by 4 stops in the shadow area. This way you'd push the banding 4 stops lower.<br>

3) Get rid of vertical banding by just subtracting the average of dark frames.<br /> You also get rid of some horizontal banding this way, but unfortunately you also lose information where there was no banding noise.</p>

<p>Needless to say, you'd get the best results doing all this in linear 16bit rgb space before you apply other corrections.</p>

<p>p.s. I don't do any of this, as I said earlier, for my needs out of camera raw to jpeg conversion is good enough.</p>

<p>I redid the test as I said. The fixed pattern noise was the same.<br>

My findings are as follows:</p>

<ol>

<li>Horizontal bands are random from shot to shot </li>

<li>Vertical bands are random but change very slowly - after two days I can't tell the difference, but a year earlier it was totally different</li>

<li>The lens being used or the changing of batteries don't seem to have an effect on read noise</li>

</ol>

<p>The tests were done on a Canon 20d.</p>

<p>Out of curiosity I just did the test again. One day later the fixed pattern noise is the same. Thru all the noise I can't spot the difference. But it is totally different than the pattern from a year ago. The fixed pattern is obviously changing but this is a slow process.<br>

I have two more ideas to try. I will change the lens and redo the test, and then take the battery out and again redo the test tomorrow. I already changed the card so this is obviously not a factor.</p>

 

<p>I just redid the test on 20d. The banding looks the same, but the fixed pattern noise I got out of averaged 16 exposures is totaly different than the pattern from a year ago.<br>

Obviously to reduce banding it only makes sense to subtract average of dark frames from about the same time as the photos were taken.<br>

One important difference I noticed is noticable amplifier glow which was not present a year ago. Exposure was the shortest possible so it's not from long exposure. I'm not sure what exactly this means, but it seems electronics degrade as time passes. This may affect the photos.<br>

I might repeat the test tomorrow, to see if it changes from day to day.</p>

<p>Rishi, I'm here.<br>

Just wanted to say, no need to repeat the test without the lens, it was tested with an EF 50 1.8 II, so no USM.<br>

It might be interseting to repeat the test thou, just to see if fixed pattern noise changes with time. The test was done over a year ago.<br>

At least on 20d in-camera jpeg conversion is realy bad and the banding becomes evident even in relatively bright areas shot at ISO 800.<br>

For me the problem was removed almost completely just by shooting raw and converting to jpeg out of camera. For example see this <a href=" at equivalent of ISO 25600</a> .</p>

 

<p>Rishi said:<br>

<em>One thing that confuses me though -- sometimes it seems like the direction of the banding changes from image to image, by 90 degrees (and no, not because the image was portrait vs. landscape). I'm confused...</em><br>

<em><br /> </em><br>

I tested my Canon 20d for low light performance and made an average of 16 black frames containing only read noise. While individual shots contained horizontal bands, the average contained only vertical bands with random noise averaged out. So it seems vertical bands are always present but horizontal are more obvious.</p>

<p><a href=" exposure</a><br>

<a href=" of 16 exposures</a></p>

<p>Detail of 1 exposure :<br>

<img src="http://farm4.static.flickr.com/3244/3095714886_288f3c9e3a_o.jpg" alt="Detail of 1 exposure" width="800" height="800" /><br>

Detail of average of 16 exposures:<br>

<img src="http://farm4.static.flickr.com/3235/3094873179_5fda1afd21_o.jpg" alt="Detail of average of 16 exposures" width="800" height="800" /><br>

This does not explain what you're seeing but it is what 20d and perhaps all canon sensors record.<br>

The value of majority of this noise was around 4 in recorded raw at all ISOs.</p>

 

Pankaj, taking sensor size into account, point & shoot cameras already reached physical limit of resolution. For similar resolution/limit you'd have aps sized sensors with 100mp and full frame with 250mp. I'm pretty sure things will never go that far, but while point & shoot cameras sell on lies and "features" dslrs leave a lot more possibilities for future development.

 

If I were you, i'd rather watch for new emerging compact cameras with large sensors like sigma dp1. But even these won't be a match for dslrs. If you think about such a camera with a fast or long lens, it's bulk is no different than a dslr.

The adapter should be 1.5mm thick on the outside. If it's thicker the lens won't focus to infinity.

 

I have 4 russian adapters, of which two are actualy too thin but that's not such a problem.

If you were thinking about reducing the existing lenses to rectangular shape:

1) aperture is idealy circular, with fast lenses for example all of the area of the front element is used, hence almond shaped bokeh off center - the front element is too small as it is, now if it was rectangular...

2) weird bokeh with rectangular openings

3) you'd have even more light falloff

3) mounting the lens elements with circular retaining rings as oposed to what?

4) unnecessary difficulty in grinding the rectangle

 

It would make a little sense to appropriately enlarge the existing lens to rectangular shape, but that would mean a lot more weight, size, cost, difficulty in manufacturing etc.

 

PS cameras that seem to have rectangular lenses I believe only have rectangular shades over circular glass.

Press the DRIVE-ISO button, with the front wheel change the small image to circle or clock. That's it. The camera will blink for 10 secs and take a picture.

If you use mirror lockup it will wait only 2 secs, after it lifts the mirror.

I haven't tried this yet, but it's worth to look at "flat frame" astro photographers use to get rid of dust on sensor, vigneting etc.

Your fungus realy looks horrible but it's realy not much different than dust on sensor.

I'm not sure it can be done in photoshop, but what you do is divide the photo by flat frame.

Thanks for the answers all is clear now.

 

I know what I will do is pretty much irreversible damage to a perfectly good lens, but I realy don't plan to buy any FD camera.

 

I successfuly took apart the mounting part of a lens and isolated the part that needs the change. The change needs some planing and precision but it realy doesn't seem that difficult.

 

My plan is simply to remove the chrome ring, mill the remainder for a bit more than 2mm and attach on it the EF bayonet. What I don't like is I will probably damage or destroy the red dot, I realy like it :)

 

The only thing that worries me a bit is I have to use the part with the aperture levers because it contains a glass element and there's not enough space for the EF bayonet. I'll probably have to mill this one too. I'll use an m42 to EOS converter and bore the 42mm hole to a larger diameter. I'll have to shorten the levers too.

 

I will post my progress as I go, but it may take a little while.

 

Thanks again.

Hi,

 

I have a FD lens that I want to use on my EOS digital camera. I have a small

workshop so I can make all the parts needed for the makeover. I know the

flange to film distance for the FD is 42mm and for EF 44mm. What I don't know

is what exactly is the flange on a FD lens. The inner part where the aperture

levers are, or the outer ring that looks like a female side of a bayonet?

I know next to nothing about mounting the FD lens, but it seems to me the

inner part is what actualy doesn't rotate - here I'm thinking of the things I

read about FD breech lock system that has no wear on the flange precisely

because of not rotating, but the outer ring simply looks like a flange to me.

So, which is it?

 

Thanks.